Photo Credit: Rafael Garcin on Unsplash

We’re talking about AI like it’s a capability race.

It’s not.

It’s a systems problem.

AI is scaling fast. Faster than most organizations can absorb.

Tasks that took hours now take minutes.
Work that took months is collapsing into days.

That’s real.

But the constraint didn’t disappear.

It moved.

The Bottleneck Has Shifted

It’s no longer intelligence.

It’s coherence.

Every company is already saturated with input:

Calls
Chats
Emails
Meetings

AI accelerates all of it.

But acceleration without structure creates a different failure mode:

More output. Less alignment.
More decisions. Less traceability.
More automation. Less accountability.

The system can think faster than it can remember.

That’s Why Industrial Policy Is Showing Up

The recent push toward AI-era industrial policy isn’t theoretical.

It’s a signal.

We’re starting to realize this isn’t just a technology shift.
It’s a replatforming of the economy.

And markets don’t solve for this on their own.

They don’t answer:

• Who captures the upside from AI-driven productivity

• How workers participate in that upside

• What replaces systems built on labor when labor isn’t the constraint

• How trust holds when decisions become opaque

Those aren’t product questions.

They’re infrastructure questions.

The Gap No One Designed For

Every enterprise system we rely on shares the same limitation:

They weren’t built to preserve conversations.

CRM stores records.
Contact centers handle interactions.
Analytics tools summarize outcomes.

But the conversation itself?

That’s where the signal is.

Intent
Objection
Commitment
Trust

Captured in the moment.
Compressed into summaries.
Lost to the system.

And then we ask AI to operate on that.

Intelligence Doesn’t Compound Without Memory

This is the structural flaw.

AI improves answers.

But it doesn’t automatically improve understanding.

Without continuity:

• Models don’t learn from real interaction chains

• Organizations can’t trace how decisions were made

• Compliance becomes reconstruction instead of verification

• Trust becomes probabilistic instead of provable

You don’t get compounding intelligence.

You get faster approximation.

The Risk Isn’t Just Job Loss

There’s a deeper economic shift happening.

If intelligence scales faster than participation, value concentrates.

Not because of bad intent.

Because of system design.

The organizations that:

• Capture signal early

• Structure it correctly

• Retain it over time

…will compound advantage.

Everyone else operates on fragments.

That’s how inequality shows up in an intelligence economy.

Through continuity.

“Keeping People First” Requires More Than Access

There’s a lot of focus right now on access to AI.

That’s necessary.

But it’s not sufficient.

Three things actually matter:

1. Access to intelligence

2. Participation in the upside

3. Visibility into how decisions are made

The third one is where the system breaks.

Because visibility requires structure.

The Missing Layer

At some point, every organization hits the same wall:

You can’t govern what you can’t see.
You can’t audit what you can’t reconstruct.
You can’t improve what you can’t trace.

That’s not a model problem.

It’s a continuity problem.

The next layer in this stack isn’t another model.

It’s a conversation layer.

A system that:

• Captures interactions as structured records

• Preserves context across time and systems

• Enables verification, not just interpretation

• Allows intelligence to compound because memory persists

Without it, we’re building an intelligence economy on partial truth.

What This Unlocks

If we get this right:

• AI becomes accountable, not just powerful

• Organizations operate with memory, not fragments

• Workers contribute signal that actually persists

• Decisions become traceable

• Trust becomes verifiable

And importantly:

The upside distributes.

What Happens If We Don’t

We still get better models.

But we also get:

• More concentration of power

• More opaque systems

• More fragile institutions

• Less alignment between action and accountability

Not because AI failed.

Because the system around it did.

The Real Question

AI is already transforming the economy.

That part is done.

The question now is:

Do we build the infrastructure that keeps it legible, accountable, and human?

Because intelligence changes the game.

But memory determines who captures the value.

About the Author

Ken Herron is a B2B SaaS strategist focused on the emerging conversation layer of enterprise data infrastructure. With more than 30 years of experience across telecommunications, contact centers, and conversational AI on five continents, his work centers on how organizations capture, structure, and govern human interaction as a durable system of record.

He writes at the intersection of AI, economic infrastructure, and operational continuity. His focus is on a simple but under-addressed problem: intelligence is scaling faster than the systems designed to preserve context, accountability, and trust. His work explores how open standards like vCons transform conversations into portable, verifiable assets that enable AI systems to operate with memory, not fragmentation.

Ken’s perspective is grounded in commercial reality. He works with enterprises navigating how AI changes not just productivity, but how value is created, retained, and distributed.

About Global AI Leaders

Global AI Leaders is a practitioner-led briefing on the systems emerging around AI, not just the models themselves.

The focus is where most coverage stops: infrastructure, governance, and economic impact. How AI reshapes decision-making, redistributes value, and exposes gaps in the systems enterprises and institutions rely on to operate.

If you’re building, deploying, or governing AI inside a real organization, this is written for you.

Less hype. More signal.

And if something in this piece maps to what you’re seeing inside your own systems, add your perspective. That’s how this conversation gets sharper.

Over centuries, expertise, experience and wisdom have played fundamental roles in shaping education and learning in academia. In fact, expertise has often been the bedrock upon which knowledge creation, education, and scholarly progress have depended upon. It represents deep, specialized knowledge in specific domains, with the mastery of theories, methodologies, and accumulated scholarship that define academic disciplines. On the other hand, experience encompasses practical knowledge often gained through direct engagement with real-world applications as well as research and teaching. It brings distinct and irreplaceable value to academia, complementing expertise which enhance both research and education. Wisdom transcends both, involving the meta-cognitive ability to synthesize knowledge across domains, understand contexts, make sound judgments, and recognize the limitations of what we know. In my view, it can be elusive, which is only exhibited often by a few and that too with many years of experience and substantial deep expertise.

Subscribe now

Traditionally, these three elements have worked synergistically in academia. Expertise provided the foundation, experience added practical dimension, and wisdom guided their application. Senior academics (with many years of experience) embody this integration, serving as both repositories of knowledge and arbiters of scholarly judgment.

In this short note, I examine how the developments in artificial intelligence (AI) and the associated tools are fundamentally transforming the relationship between expertise, experience, and wisdom. Though each of these forms of knowledge brings unique value, their significance and application are being shifted in profound ways by AI. It is critically important for the senior leadership in higher education (or for that matter in any organization) to recognize this enormous transformation, which will increase even more at a rapid pace in the next years to come.

Expertise under pressure

AI tools can now generate sophisticated analyses, write research papers, and conduct literature reviews with remarkable competence. This democratization of access to expert-level knowledge is, on the one hand, liberating but, on the other hand, can be threatening to staff. Students can now have access to capabilities that previously required years of specialized training. An even more significant question this raises is about the value of developing deep expertise itself.

For instance, in the legal area, AI tools are capable of performing legal research, analysing case law by searching vast databases of cases and identifying relevant precedents, and generating legal briefs and drafting legal documents. This can put pressure on junior and even senior lawyers who have built their reputation on their mastery of case law and legal precedent, as AI tools can now synthesize legal information comprehensively as well as effectively.

Share

In medical fields such as radiology and pathology, AI diagnostic tools can have high accuracy rates matching or even exceeding human specialists. AI tools can analyse medical images, suggest differential diagnoses, and recommend treatment protocols based on vast databases of medical literature. This raises questions about the traditional model of medical expertise that emphasizes years of training and pattern recognition.

In creative writing, AI tools are able to generate stories, music as well as visual art that often rival human creativity. Last year in 2024, we witnessed strikes in Hollywood, primarily led by the Writers Guild of America, where the major point of contention was the use of AI in scriptwriting and its potential to replace human writers impacting on writers’ job security and compensation. In the academic environment, students may begin to question the value of studying traditional artistic techniques when AI can produce compelling creative work almost instantly. Academics who have spent decades developing expertise in creative processes may find their specialized knowledge of techniques and form less distinctive when AI can potentially generate quite sophisticated creative outputs.

In computer science, AI coding assistants can generate functional code, debug programs, and explain algorithms. This puts pressure on the expertise of computer scientists and software engineers whose value proposition is based on deep knowledge of programming languages, algorithms, and software architecture. With such tools, students are able to build complex applications without understanding fundamental computer science principles!

Such examples1 show that AI is challenging the fundamental assumption that expertise requires years of specialized training to develop. When AI can perform tasks that previously required deep domain knowledge, it forces a reconsideration of what expertise actually means and where its enduring value lies. In academia, this pressure will be felt most acutely by mid-career academics who have invested heavily in developing specific technical skills that AI can now replicate. It will also affect early career academics, who now will need to think clearly about how to chart their career paths working with the AI tools. At present, it may be less threatening to senior scholars, whose wisdom and judgment remain somewhat difficult to automate, at least as of as of now.

Experience needs to be valued more

In my view, AI making expert-level information more easily accessible should lead to increasing the weight that one needs to associate with the experience component making it even more valuable in making decisions such as in recruitment. For instance, the ability to distinguish between AI-generated content that sounds authoritative and content that is actually reliable requires deep practical experience. I view experience as having lived experience, for instance, having done different projects in different environments tackling different constraints and requirements, and learning from them.

Leave a comment

For instance, a cardiologist professor spending many years spent as a practicing physician can teach medical students not just what a heart murmur sounds like, but how to distinguish between the subtle variations that indicate different conditions, how patients actually present in emergency situations, and how to communicate difficult diagnoses. While AI tools can analyse ECGs with high accuracy, the experienced clinician knows a person’s anxiety about her chest pain is as important as one’s test results, and that bedside manner affects patient outcomes as much as technical precision.

A computer science professor who has worked in a large tech company can explain why elegant code sometimes fails in software production, how team dynamics affect software projects, and the difference between academic prototypes and industry solutions. They understand technical debt, the pressure of shipping deadlines, and how user feedback changes product design and development.

An academic in journalism who has worked as a war correspondent brings experiential knowledge about source protection, ethical decision-making under pressure, and the reality of reporting in dangerous conditions. They can teach students about the split-second decisions required when interviewing trauma victims, the practical challenges of fact-checking in rapidly evolving situations, and how newsroom economics affect story selection. AI tools can help with research and writing, but experience teaches the human judgment required for ethical journalism.

A political science professor who served as a diplomat can explain the gap between formal international relations theory and actual negotiations. They understand how cultural nuances, personal relationships, and behind the scenes dynamics affect treaty negotiations in ways that AI tools analysing diplomatic texts cannot capture.

Academics with lived experience and who have wrestled with real research challenges in different contexts and in diverse environments are not only better positioned to analyse critically the different situations and communicate them to students but also are able to evaluate better AI generated outputs. In fact, the experience of working with AI tools, and knowing their limitations and understanding their biases have started to become a form of valuable experience in their own right!

Wisdom even more valuable

Wisdom may become the most valuable academic asset in the age of AI proliferation. However, as noted above, not everyone is able to achieve wisdom which often requires many years of experience and deep expertise. Wisdom involves understanding the broader implications of research, encompasses the ability to ask the right questions, not just generate answers, maintain perspective on what truly matters, while recognizing the limits of both human and artificial intelligence.

In technology, disciplinary wisdom can involve understanding not just how to build a system but also the consequences of deploying such a system, the implications on maintainability and the people who will work with it later. Wisdom can also involve seeing patterns across different systems and anticipating how technologies will evolve or interact. This wisdom includes knowing when to break established rules, when complexity is warranted versus when simplicity serves better, and how to balance competing priorities like performance, security, and usability. The deepest disciplinary wisdom often emerges when someone connects their specialized knowledge to broader patterns of human experience, which enters the realm of philosophy.

Emerging Future

In 2003, when Google search became a successful product2, it provided pointers to sources of information. This meant one can always go to the source when needed for the required information. Now AI tools such as ChatGPT3 and Co-Pilot4 go to the information source and summarise and present the collected information in a convenient way for humans to consume. Having these tools at the fingertips, students can easily access the information they need in a form that is suitable for them. The information can be customized and personalised to suit each student’s needs. Such an environment is suggesting a future where academic value lies not in the possession of information but in the curation of information, its synthesis and most of all on the judgment based on critical analysis and experience. The most valuable academics of the future are likely to be those who can combine deep domain expertise with a wide practical experience (including the use of AI tools) and the wisdom to make judgments in their own disciplines while maintaining high values and ethical standards.

Though this transformation is still unfolding, it is critical that people who hold leadership positions take the time to understand what is happening. They need to understand how such transformation can impact their organization in a range of its operations from service provision to recruitment and career development. They need to ensure that their organization continues to evolve with such transformations; otherwise, they will be left behind as the world keeps on moving. But it is not about investing in AI blindly or just using AI for cost cutting measures alone. No organization has become the jewel in their sector just by cutting costs. However, they ought to start thinking about how AI can be synergistically deployed with the workforce to create new services, which can help to achieve genuine competitive advantage and hence make a difference.

About the Author

Professor Vijay Varadharajan is an Emeritus Professor at The University of Newcastle and an Honorary Professor at Macquarie University. With over 30 years of experience across the UK, USA, and Australia, he has held prominent roles such as the Global Innovation Chair Professor in Cyber Security and the Microsoft Chair Professor in Innovation. His extensive career spans both academia and industry, including leadership positions at Hewlett-Packard, British Telecom, and Microsoft. A recognized expert in Cyber Security, Cloud Computing, Internet of Things (IoT), and Big Data Security, he has served on several international advisory boards for government and technology sectors. He holds a PhD in Computer and Communication Engineering and is a Fellow of both the Australian Computer Society and Engineers Australia.

Refer a friend

Moving AI from a boardroom slide to a bus terminal is harder than it looks. We often talk about artificial intelligence in grand, abstract terms, but for the average person, those high-level discussions do not mean much when they are just trying to get home for the holidays.

I recently led a project at UIB where we partnered with Park May Berhad to launch Malaysia’s first WhatsApp-based booking platform for Transnasional and Plusliner bus tickets. It sounds simple on the surface, but the journey to making it “simple” for the user taught us a lot about what it actually takes to bring meaningful AI solutions to the real world.

Here is a pragmatic look at the hurdles of practical AI and how we navigate them.

Share

Stop building apps nobody wants

The first hurdle is ego. Many companies want to build their own “destination” app, hoping users will download it, register, and keep it on their home screen. The reality is that people already have app fatigue. They do not want another login to remember or another 100MB of storage taken up for a task they do once a month.

Our approach was to meet people where they already are: WhatsApp. By sending a simple “Hi” to a specific number, a traveler can search for routes, check seats, and buy a ticket. This is not about being “high-tech” for the sake of it. It is about being useful.

If your AI requires a user to change their behavior significantly, you have already lost.

Subscribe now

The messiness of the “Last Mile”

Real-world AI is never as clean as a demo. To make the bus booking system work, we had to integrate conversational AI with legacy backend systems, real-time seat inventories, and secure payment gateways.

This is where many AI projects fail. It is easy to build a chatbot that answers “What time is the next bus?” It is incredibly difficult to build a system that securely processes a credit card payment, generates a boarding pass, and updates a database in real-time within a messaging interface.

Practical AI is 20 percent “intelligence” and 80 percent robust engineering and integration.

Leave a comment

The “Grandmother Test” for inclusion

If an AI solution only works for the tech-savvy, it is not a success; it is a niche. For a national transport operator like Transnasional, the demographic is everyone. We had to ensure that a student, a business traveler, and a grandmother could all navigate the same interface with the same level of ease.

By using conversational AI, we removed the friction of complex menus and forms. You just talk to the service. This inclusivity is what drives digital adoption.

When we make AI human, we make it accessible to people who might otherwise be left behind by the digital divide.

Refer a friend

Lessons from the field

If you are looking to bring a practical AI solution to market, my advice is to ignore the hype and focus on the friction.

• Find the one task that is most annoying for your customer.

• Identify the platform they already use every day.

• Build the bridge between the two.

As we are moving away from an era of “using” AI to an era of “simply communicating” with the world around us, AI should not feel like a new tool you have to learn. It should feel like the tool you already have just got a whole lot smarter.

About the Author

Muzzamel Mazidee is the General Manager and Director at UIB Malaysia for UIB.ai and a leading expert in conversational AI and digital transformation. He specializes in bridging the gap between complex technology and everyday human needs, helping global brands realize the power of AI at scale.

About UIB

UIB is an API-first company that allows businesses to connect with their customers via the most popular messaging and AI platforms through a single API. Our tagline, “Simply Communicate,” reflects our mission to make human-to-machine communications as natural as human-to-human interaction.

To explore how Muzzamel and the UIB team can help your business realize practical AI solutions, feel free to email them at info@uib.ai or simply send them a WhatsApp message.

Want to become an author at Global AI Leaders yourself?

We are currently navigating a brief “open-source spring” in the history of Artificial Intelligence.

For a fleeting moment, the most powerful cognitive tools ever created by humanity are comparatively affordable, widely available, and accessible to anyone with the vision to use them.

But this window may be closing.

The Fortunate Moment

Right now, the barriers to entry are at an all-time low. On the one hand, certain areas of AI are finally working reliably and make sense for humans to use.

On the other hand we are not yet in the era of “elite-only” access, or too costly and prohibitive relationship-based licensing of AI tools that would otherwise make a productive long term use unviable.

So whether you are a startup founder or a global enterprise leader, the building blocks for a custom AI Operating System (O/S) Infrastructure are within your reach. You can architect a system that understands your unique “uniqueness”, from the IP hidden inside your team’s brains, your company’s specific workflows, your proprietary data, and your organizational “logic.”

For perhaps the first time in a long time, taking software development back in your own hands could make a lot of sense: why buy an off-the-shelf CRM, ERP or CMS product, when you can use AI to get a cheaper, more precise, and more private solution ready in weeks or months? One that fully integrates with your individual company vision, philosophy, culture and processes?

The Risk of the “Central AI Brain”

Those who wait could eventually find themselves in the dire position of forced dependency.

As the market matures and the competitive differences at an organizational level (between those using AI and those who don’t) begin to show, “Central AI Brains” (massive, proprietary models controlled by a handful of global titans) will likely tighten their grip and appear as the last option to compete with the others that meanwhile built their own “AI Operating System” as the digital backbone of their organization, leading to:

• The Dependency Trap: If you do not own your infrastructure, you are forced to “take what the central brain gives you.”

• Loss of Control: You will feed on their algorithms but with zero control over the underlying logic, the privacy of your inputs, or the sustainability of the service.

• The “Rent” Economy: Instead of owning an asset, you will be a digital tenant, subject to shifting fees and “walled garden” restrictions.

After all: Why pay ongoing SaaS licensing fees when you can become your own “internal” SaaS provider?

Why an “AI O/S” is Your Ultimate Intellectual Capital

An AI O/S is not just a collection of tools; it is your organization’s Structural Capital. It is the digital manifestation of your team’s collective intelligence. By building it now, you ensure that:

1. Ownership remains with you: The “rule of law” can only protect what you own and have legally defined as your IP.

2. You avoid the “Elite Barrier”: In the future, building such a system may require “relationship-only” connections or astronomical costs that not all corporations can afford.

3. You dictate the “Innovation Language”: Your O/S becomes the interface through which you interact with the world, ensuring that external AIs must adhere to your protocols.

The Bottom Line: The tools are democratized today so that you can build the sovereign infrastructure of tomorrow.

Do not wait for the window to close.

Own your logic, own your O/S, and own your future.

About the Author

Toby Ruckert is the Managing Partner of InnoIP AG, a boutique innovation agency headquartered in Switzerland that specializes in making Intellectual Property tangible. As a strategic advisor to boards, CxO leadership groups, and visionary founders, Toby focuses on the intersection of human ingenuity and algorithmic governance.

His work centers on the concept of Intellectual Sovereignty: the belief that in an automated world, the “rule of law” remains the last line of defense for human creators. Through InnoIP’s unique “Innovationsbegleitung” (Innovation Guiding) framework, he helps organizations catch the “golden nuggets” of innovation that often fall under the table during intense R&D cycles, converting them into sovereign assets and custom AI infrastructure.

Beyond corporate advisory, Toby represents elite inventors and artists much like a talent agency, negotiating their Intellectual Capital, “Digital Twins,” and brand rights in the global marketplace. His mission is to ensure that creators and companies become the landlords of their own digital future rather than mere tenants of a centralized AI brain.

Photo Credit: MD. Ziaul Haque on Unsplash

Last week, Meta AI security researcher Summer Yue told her OpenClaw agent to check her overstuffed inbox and suggest what to delete or archive. The agent proceeded to delete all her email in what she described as a “speed run,” while ignoring her frantic stop commands from her phone.

“I had to RUN to my Mac mini like I was defusing a bomb,” she wrote on X, posting screenshots of the ignored stop prompts as receipts.

If a professional AI security researcher can accidentally let an agent nuke her inbox, what hope does a marketing VP have (rhetorical question: that’s who I wrote this guide for)?

Welcome to the moment marketing leaders have been both waiting for and quietly dreading: AI agents are no longer chatbots that write blog posts. They’re autonomous systems that can read your email, modify your CRM, schedule your meetings, and, if you’re not careful, reply-all to your board of directors with a first draft that includes the phrase “Insert CEO Quote Here.”

This guide is your field manual. Two [wildly] different platforms. Seven steps. Zero deleted inboxes.

What Just Happened (and Why You Should Care by Monday Morning)

OpenClaw: The Open-Source Agent That Ate Silicon Valley

OpenClaw, originally called Clawdbot, then Moltbot, is an open-source AI agent created by PSPDFKit founder Peter Steinberger. It runs locally on your machine (Mac Minis have become the hardware of choice), connects to your messaging apps, email, calendar, and file system, and maintains persistent memory across conversations.

The numbers are staggering: 60,000+ GitHub stars in 72 hours, now well over 219,000. The Silicon Valley in-crowd has fallen so deeply in love that “claw” has become the generic term for agents that run on personal hardware. Spinoffs include ZeroClaw, IronClaw, and PicoClaw. Y Combinator’s podcast team recently appeared on air dressed in lobster costumes. (This is not a joke. This actually happened.)

What makes OpenClaw powerful, and dangerous, is that it provides capabilities that SaaS AI assistants intentionally restrict: full shell and file system access with no default sandboxing, persistent memory that retains context across sessions, direct integrations with Gmail, Slack, WhatsApp, Telegram, Discord, and calendars, and extensibility through “skills” (third-party packages with system-level permissions).

This is effectively giving an LLM “sudo” on your infrastructure. For marketing operators, that means an agent that can draft your emails, analyze your campaign data, and reorganize your project files. It also means an agent that can accidentally delete your emails, corrupt your campaign data, and reorganize your project files into oblivion.

Perplexity Computer: The Managed Alternative

Two days ago, February 25, 2026, Perplexity launched “Computer,” and it racked up 12 million views on X in 20 hours. The pitch: one system that orchestrates 19 different AI models, routing each task to the model best suited for it.

The lineup reads like an AI all-star team: Claude Opus 4.6 for core reasoning, Gemini for deep research, Nano Banana for images, Veo 3.1 for video, Grok for speed on lightweight tasks, and ChatGPT 5.2 for long-context recall. Users describe what they want, and Computer breaks it into tasks and subtasks, creating sub-agents for execution: document drafting, data gathering, web research, and API calls running in parallel.

Available now on the $200/month Perplexity Max plan, with an Enterprise Max plan coming at $325/month. The critical difference from OpenClaw: Computer runs entirely in Perplexity’s cloud, with managed safeguards and session-scoped credentials.

The Strategic Choice

Think of it as build-vs-buy for the agent era:

• OpenClaw = local-first, maximum control, maximum responsibility, your data never leaves your machine, free (bring your own API key)

• Perplexity Computer = cloud-managed, turnkey orchestration, safer by design, less customizable, $200/month

Neither is foolproof. Both are powerful. The question isn’t which is “better”; it’s which risk profile fits your organization’s readiness.

The Summer Yue Incident: Your Cautionary Campfire Story

Let’s unpack what actually happened, because the details contain every lesson a marketing operator needs.

What Went Right (at First)

Yue did what any responsible operator would do: she tested on a smaller dataset first. She created a “toy inbox,” less important email, and let OpenClaw manage it. It performed well. It earned her trust. She thought she’d let it loose on the real thing.

“Were you intentionally testing its guardrails or did you make a rookie mistake?” a software developer asked her on X.

“Rookie mistake tbh,” she replied.

What Went Catastrophically Wrong

The large volume of data in her real inbox “triggered compaction”. Here’s what that means in plain English: every AI agent has a “context window,” a running record of everything it’s been told and everything it’s done in a session. When that window gets too full, the agent starts summarizing, compressing, and managing the conversation to stay within its limits.

At that point, the AI may skip over instructions that the human considers quite important. In Yue’s case, the agent may have skipped her most recent prompt, where she told it not to act, and reverted to its earlier instructions from the toy inbox. The result: a deletion spree while she watched helplessly from her phone, her stop commands ignored.

As several security experts on X pointed out: prompts can’t be trusted to act as security guardrails. Models may misconstrue or ignore them.

The Marketing Operator’s Lessons

This wasn’t a freak accident. It was a predictable failure mode that will happen to every team that skips proper safeguards:

• Agents that pass sandbox tests can fail catastrophically on production data. The volume, complexity, and edge cases in real environments are fundamentally different from test environments.

• “Trust but verify” is insufficient. Yue trusted based on successful small-scale tests, exactly what most marketing teams would do. Architectural safeguards beat verbal instructions every time.

• Your emergency brake can’t be a chat message. When the agent is running autonomously on a local machine, typing “STOP” into a phone app is not a reliable kill switch.

The Marketer’s 7-Step Pilot Framework

Here’s the framework that stands between you and becoming the next cautionary tale on X. These steps apply whether you’re piloting OpenClaw, Perplexity Computer, or next week’s new AI agent.

Step 1: Start With a Sandbox, Not Your Soul

Create a disposable test environment with synthetic or dummy data. This means a separate email account, a cloned (not live) CRM instance, and test calendar entries that won’t accidentally book your CEO for a 6 AM meeting in Lucerne.

Never pilot on production CRM, email, or customer databases first. For OpenClaw specifically, deploy on isolated infrastructure, a dedicated VPS or Docker container, not your primary workstation. For Perplexity Computer, create a separate workspace that isn’t connected to your production tools.

The rule: If the agent goes full Summer Yue, the worst thing that happens is you lose data you created specifically to lose.

Step 2: Define the Blast Radius

Before any pilot, ask: “If this agent goes rogue, what’s the absolute worst that happens?” Then scope permissions to minimize that damage.

This is a concept borrowed from incident response and infrastructure engineering, and it applies perfectly to AI agents. Map out every system the agent can touch, and for each one, answer:

• Can the agent read data? (Low risk)

• Can the agent modify data? (Medium risk)

• Can the agent delete data? (High risk)

• Can the agent send communications to external parties? (Existential risk)

Use read-only access wherever possible. An agent that can *analyze* your campaign performance data but can’t *modify* it is still enormously valuable, and won’t accidentally reset your attribution model.

Step 3: Set Human Approval Gates

For any action that creates, modifies, or deletes data, require human confirmation. This is non-negotiable for customer-facing systems.

Both OpenClaw and Perplexity Computer support human-in-the-loop approval workflows. For OpenClaw, configure explicit approval requirements for execution, file writes, file deletions, and message sending. For Perplexity Computer, the cloud architecture provides session-scoped controls that limit agent autonomy.

The mental model: Think of approval gates like the “Are you sure you want to send this to 50,000 contacts?” confirmation dialog in your email platform. Except this time, the entity clicking “Send” has the attention span of a golden retriever in a tennis ball factory.

Step 4: Use the “Intern Test”

Would you give a first-day intern unsupervised access to this system? If not, don’t give it to an agent.

Start with monitoring and reporting tasks before graduating to execution tasks. A natural progression:

1. Week 1-2: Read-only intelligence (competitive monitoring, content performance analysis, social listening)

2. Week 3-4: Draft generation with human review (email sequences, social posts, meeting prep summaries)

3. Month 2: Low-stakes execution with approval gates (scheduling, internal notifications, report distribution)

4. Month 3+: Higher-autonomy tasks with monitoring (only after establishing a track record with comprehensive logging)

No intern gets promoted to VP in their first week. Neither should your agent.

Step 5: Credential Isolation

Never give the agent your production API keys or admin credentials. Create scoped, limited-permission service accounts. Treat the agent as a potential pivot point for security breaches.

This is especially critical for OpenClaw, where researchers have found over 21,000 publicly exposed instances leaking API keys, and 36% of all community-built skills (extensions) contain security flaws. The ClawHub marketplace, OpenClaw’s skill store, has been plagued by a coordinated malware campaign called “ClawHavoc” involving 335+ malicious packages.

For Perplexity Computer, the risk is different but real: ensure the connectors you enable are scoped to minimum necessary permissions, and understand what data flows through Perplexity’s cloud infrastructure.

Never store credentials in plaintext: Use a secrets manager (HashiCorp Vault, AWS Secrets Manager, or even 1Password CLI) and rotate API keys every 30-90 days.

Step 6: Monitor the Context Window

Understand compaction: the exact failure mode that caught Summer Yue. When agents process large volumes of data, their context window overflows, causing them to “forget” key instructions.

The fix: break tasks into smaller batches. Instead of “Review my entire inbox of 47,000 emails,” try “Review the 50 most recent emails from the last 24 hours.” Instead of “Analyze all campaign data from Q4,” try “Analyze the email campaign data from December, then stop and report.”

For OpenClaw specifically, write critical instructions to dedicated files (like `RULES.md`) rather than relying on prompts alone—prompts can be compressed away during compaction, but file-based instructions persist.

For Perplexity Computer, the multi-model architecture means different sub-agents handle different tasks, which naturally reduces context window pressure, but large research tasks can still trigger similar behavior.

Step 7: Build Your Kill Switch

Know exactly how to stop the agent instantly. Test the kill switch before you need it, not during the 3 AM panic.

• For OpenClaw (local): Know the process ID, know the Docker container name, know the terminal command to terminate. `docker compose down openclaw-gateway` should be muscle memory. Do not rely on typing “stop” into a chat window.

• For Perplexity Computer (cloud): Understand session controls, know how to revoke connector permissions, and know where the “terminate all active tasks” button lives in the dashboard.

Pro tip: Create a one-line shell script or bookmark that kills the agent instantly. Label it something memorable. “Break Glass” works. So does “The Yue Button.”

OpenClaw vs. Perplexity Computer: The Marketing Leader’s Decision Matrix

Five Marketing Pilot Use Cases to Start With

These are the Goldilocks zone: high enough value to justify the effort, low enough risk that a failure won’t make the news.

Safe Starting Pilots

1. Competitive Intelligence Monitoring (Read-Only) — Point the agent at competitor websites, press releases, and social feeds. Ask it to synthesize weekly briefings. Zero write access required. If the agent hallucinates a competitor product launch, the only casualty is a Slack message that gets corrected.

2. Content Performance Reporting — Connect to your analytics platforms (read-only) and have the agent generate weekly performance digests. It’s doing the work your team spends three hours on every Monday morning, and the worst case is a mis-formatted chart.

3. Meeting Prep Synthesis — Combine calendar context, relevant email threads, and web research to generate pre-meeting briefings. Perplexity Computer is particularly strong here with its multi-model research capabilities.

4. Social Listening and Trend Summarization — Monitor brand mentions, industry keywords, and trending conversations. Summarize for the team. Read-only, no customer interaction, high signal.

5. Draft Generation With Human Review — Email sequences, outreach templates, social post drafts. The key word is draft. Nothing goes to a customer, prospect, or the public without a human reviewing and approving it.

What NOT to Pilot First

• Anything involving live customer data or PII

• CRM modifications (field updates, lead scoring changes, pipeline stage moves)

• Live campaign changes (budget adjustments, audience modifications, bid changes)

• Automated outreach without human approval gates

• Any system where “undo” is complicated or impossible

The general principle: if the failure mode involves calling an emergency meeting with Legal, it’s not a first pilot.

The Governance Conversation You Need to Have Monday Morning

Before the first agent touches a single marketing system, someone in the C-suite needs to answer these questions. If these sound like they belong in a legal compliance meeting, good. They do.

Ownership and Accountability

Who owns the agent’s actions? When an agent sends an email, modifies a record, or generates a report, who is accountable? The marketing team that deployed it? The IT team that approved the infrastructure? The vendor? This isn’t theoretical, it’s the question regulators will ask after an incident.

Incident Response

What’s the playbook when (not if) something goes wrong?

For OpenClaw deployments: immediate isolation (container shutdown), credential rotation, forensic log analysis, and recovery from clean backups.

For Perplexity Computer: session termination, connector permission revocation, and audit trail review. Every organization running AI agents needs a documented incident response plan that’s been *tested*, not just written.

Data Privacy and Compliance

How does this interact with existing frameworks? GDPR, CCPA, SOC 2, industry-specific regulations, all apply to what agents do with data. OpenClaw’s local-first architecture may simplify some compliance questions (data doesn’t leave your infrastructure), but the persistent memory feature means the agent retains and recalls information across sessions, which creates its own compliance considerations. Perplexity Computer’s cloud architecture means understanding exactly what data flows through their infrastructure and how it’s retained.

Shadow IT Reality

It’s already happening. Token Security reported that 22% of employees at monitored companies were running OpenClaw on corporate machines by late January 2026. The question isn’t whether your team is experimenting with AI agents, it’s whether they’re doing it with guardrails or without them. A sanctioned pilot with proper governance is infinitely safer than a dozen unsanctioned experiments running on personal laptops connected to corporate email.

🦀 TL;DR: The 7-Step Framework at a Glance

For the skimmers, the over-caffeinated, and the executives who scrolled straight to the bottom (no judgment, that’s good time management):

About the author

Ken Herron is a B2B SaaS strategist focused on the conversation layer of enterprise data infrastructure. With more than 30 years of experience spanning telecommunications, contact center platforms, and conversational AI across five continents, he works with organizations exploring how open standards like vCons transform human interactions into portable, AI-ready assets.

His work sits at the intersection of interoperability, governance, and commercial strategy, helping enterprises treat conversations not as recordings, but as structured infrastructure. An infrastructure layer that supports safety, compliance, customer experience, and revenue growth. Ken writes regularly on conversational intelligence, interoperability standards, and the role of human data in AI-driven enterprises.

About Global AI Leaders

Global AI Leaders is a practitioner-led briefing on the AI systems reshaping enterprise and society. If Summer Yue’s inbox story made you laugh, and then immediately check your own agent permissions, consider subscribing. And if you’ve already run your own agent pilot (successful or otherwise), let us know in the comments (anonymity provided upon request!).

The moment it clicked for me was when we were working on a workflow automation platform for a client. They had a flow builder — drag and drop nodes, connect them, define conditions. Standard stuff. Then we added an MCP layer. Suddenly, their AI assistant could look at the flow builder’s data, understand the patterns from hundreds of previously built workflows, and predict what the next node should be. Not just suggest — predict with accuracy, based on historical data.

That’s when I realised: every piece of enterprise software I’ve built in the last 23 years is sitting on a goldmine of data that AI can unlock. The missing piece was never the AI model. It was the middleware layer that connects the model to the software. That layer is MCP.

What MCP Actually Is (Without the Hype)

MCP — the Model Context Protocol — is a standardised way to connect large language models to your existing software. Think of it as an API specifically designed for AI consumption.

Your traditional REST API is built for humans and frontend applications. It returns data in formats designed for rendering on screens. An MCP server is different — it exposes your software’s data and capabilities in a way that LLMs can understand, reason about, and act on.

Here’s a concrete example. Say you have a project management tool. A REST API might have endpoints like `GET /projects` and `POST /tasks`. An MCP server for the same tool would expose tools like “search projects by criteria,” “analyse task completion patterns,” and “suggest optimal task assignments based on team workload history.” The MCP layer adds semantic meaning that the LLM can reason with.

The three pillars of a well-built MCP integration are:

• Enhanced AI Capabilities — The LLM can do more because it has access to real data and real actions, not just its training knowledge.

• Secure Data Access — Every request is authenticated, authorised, and logged. The AI sees only what it’s permitted to see.

• Scalable Integration — One MCP server can serve multiple AI assistants, models, and interfaces without rebuilding.

Why Every Enterprise Product Needs an MCP Layer — Now

Here’s the strategic reality: your competitors are adding AI features to their products. If your software can’t talk to AI, it’s going to feel increasingly dated — not because it’s bad software, but because users now expect intelligent assistance everywhere.

The good news? You don’t need to rebuild. You need a layer.

We’ve been doing this across multiple client engagements, and the pattern is consistent. Take the existing product. Build an MCP server that exposes its data and capabilities. Connect it to an LLM. The product instantly becomes “AI-powered” — and not in the gimmicky, chatbot-in-a-sidebar way. In a genuinely useful, data-driven way.

Let me walk you through three real use cases from our recent work.

Use Case 1: AI-Predicted Workflow Nodes

We were working with a client who had a sophisticated flow builder — the kind where business analysts drag and drop nodes to create automated workflows. Hundreds of workflows had been built over years, each representing hard-won business logic.

We built an MCP server that gave the AI access to the flow builder’s historical data — every workflow ever created, their node sequences, their conditions, their outcomes. Then we connected it to Claude.

The result was remarkable. A user could describe what they wanted in plain English — “I need a workflow that processes incoming support tickets, categorises them by urgency, routes high-priority ones to senior staff, and sends an acknowledgement email” — and the AI would generate the entire workflow. Not a generic template. A workflow that reflected the patterns and conventions specific to that organisation.

The AI wasn’t guessing. It was analysing hundreds of similar workflows that had already been built and validated by the team. It knew which node types the organisation used, which integrations were available, which approval chains were standard. All of that context came through the MCP layer.

Use Case 2: Intelligent Template Generation

Another client had a template system — email templates, document templates, report templates. Traditional approach: a library of static templates that users pick from and customise.

With MCP, we transformed this into something far more powerful. The AI could access the full template history — which templates performed well, which were modified most frequently, what context they were used in. When a user needed a new template, instead of browsing a library, they described what they needed.

The AI would then generate a template that wasn’t just linguistically correct — it was strategically optimal. It pulled in formatting conventions from the organisation’s most successful templates, used language patterns that had historically driven higher engagement, and even suggested timing based on analytics data.

This is the difference between AI as a novelty and AI as a genuine competitive advantage. The intelligence comes from your data, accessed through MCP, not from the model’s generic training.

Use Case 3: Data-Driven Timing and Strategy

The third pattern we keep seeing is analytics-driven recommendation. One client had years of campaign performance data — what worked, what didn’t, seasonal patterns, audience segmentation results.

Through an MCP server, we gave the AI access to this historical dataset. Now, when a marketing manager plans a new campaign, the AI doesn’t just help write copy. It analyses the last three years of campaign data and recommends: the optimal send time for this audience segment, the channel mix that historically performs best for this product category, and the messaging approach that resonated in similar campaigns.

The marketing team went from making gut-feel decisions to data-informed decisions — without needing to learn data science tools or hire analysts. The AI does the analysis. MCP provides the data. The human makes the final call.

The AI Agent Architecture

MCP servers are one half of the equation. The other half is AI agents — autonomous systems that can orchestrate multi-step workflows using those MCP-connected tools.

At AMT, we work with three primary agent frameworks, each suited to different use cases:

LangChain — Our go-to for most agent implementations. Excellent for building chains of reasoning where the agent needs to access multiple tools in sequence. We use it when the workflow is relatively linear: gather data → analyse → recommend → act.

AutoGen — Microsoft’s framework for multi-agent conversations. We use this when the problem requires multiple specialised agents collaborating. For example, one agent analyses financial data, another reviews compliance requirements, and they negotiate to produce a recommendation that satisfies both constraints.

CrewAI — The newest addition to our toolkit. CrewAI excels at defining agent “roles” with specific expertise and having them collaborate on complex tasks. We’ve used it for scenarios like technical due diligence, where you need a security expert agent, a scalability expert agent, and a cost optimisation agent working together.

The choice of framework depends on the complexity of the orchestration you need. Simple tool use? LangChain. Multi-perspective analysis? AutoGen or CrewAI.

The Tech Stack Behind It All

For teams evaluating their own MCP and agent implementations, here’s the stack we’ve converged on after extensive production experience:

• MCP Protocol — The core protocol for exposing software capabilities to LLMs. We build custom MCP servers in TypeScript and Python, depending on the client’s existing stack.

• LLM APIs — Primarily OpenAI GPT-4 and Anthropic Claude for reasoning and generation. We choose based on the specific task — Claude for nuanced analysis, GPT-4 for broad capability.

• Vector Databases — Pinecone and pgvector for RAG implementations that give agents access to large document collections.

• Agent Frameworks — LangChain, AutoGen, CrewAI as discussed above.

• Orchestration — Custom middleware for managing agent state, tool access permissions, and conversation context across sessions.

• Monitoring — Custom logging and analytics for tracking agent decisions, tool usage patterns, and outcome quality.

Getting Started

If you’re sitting on enterprise software that could benefit from AI integration — and let’s be honest, that’s virtually every enterprise product — here’s how to think about it:

• Identify your highest-value data — Where does your software accumulate knowledge that could inform better decisions? That’s your first MCP endpoint.

• Start with one use case — Don’t try to AI-enable everything at once. Pick the workflow where AI assistance would have the most immediate impact.

• Build the MCP layer properly — Authentication, authorisation, rate limiting, audit logging. This isn’t a prototype — it’s production middleware.

• Choose the right agent architecture — Simple tool use, multi-step reasoning, or multi-agent collaboration? The right framework depends on the complexity.

• Measure outcomes — Track what the AI actually improves. Time saved, decision quality, user satisfaction. Let the data guide your expansion.

We’re helping clients across industries make this transition right now. If you want to explore what MCP could do for your product, reach out. Your first conversation will be with an engineer who’s built these systems — not a salesperson reading from a deck.

For centuries, the trajectory of a human life was largely determined at birth. If your parents possessed a significant bank account balance, your path was clear. You had access to the world, the best schools, prestigious internships, and a network that ensured you would score well in life. Your success was a product of physical, cultural and financial inheritance. But as we enter a world saturated with artificial intelligence, this ancient structure is collapsing.

The Calculator Trap and the Math of Meaning

Imagine two students in a math class: one has a calculator, the other doesn’t. The one with the tool wins every time. But put them in an English class, and the calculator is useless.

AI is the ultimate calculator for everything. However, to operate this “universal calculator,” you must know the mathematical operation you wish to perform. In the past, knowing “how” to calculate was enough. Today however, AI itself can tell you the “way” (the how) to a certain outcome, but only if you know how to “prompt”.

The new bottleneck is a combination of “Why” and the skills associated with “prompting”. If you don’t know why you are asking a question, your prompt will be hollow, and your result will be irrelevant. If you don’t know how to express yourself clearly in thinking, writing or speaking, your prompt will render your “Why” meaningless.

The New Hierarchy of Intelligence

I forsee a future divided (in terms of professional success) into several distinct classes based on their relationship with AI:

1. The Dependent Class: Those who cannot formulate a “why” nor express it. They may rely on Universal Basic Income (UBI) and do exactly what the state or the machine tells them because they have no other choice.

2. The “Feeling” Professionals: Those in roles that require warm-heartedness, human touch, or physical presence (such as caregivers, therapists, or massage practitioners). These roles remain a sanctuary for human-to-human connection.

3. The AI-Integrated Elite: Those who use AI as an accelerator. They maintain their “freedom of choice” because they can still function if they put the tool down.

4. The Fully Merged: A theoretical group where the line between human and machine vanishes (think “Neuralink” or a “chip in the brain”), however with their “all-in” approach, it may be hard for them to realize whether they’re master, or slave.

We will be witnessing a profound decoupling of success from traditional pedigree. In this new era, the technical how is becoming a commodity available to anyone with a screen. This shift is leaving behind a new set of gatekeepers: the Why (intentional intelligence), the How to Ask (the skill of expression), and the quality of the tools you can access (privilege).

The New Architecture of Privilege: Access and Intent

In the old world, privilege was a silver spoon. In the new world, privilege is a high speed interface and the cognitive depth to use it. If in the past your upbringing determined your ability to navigate university and job markets, today your status is defined by a different trinity:

1. Access: Your ability to utilize the highest level of intelligence tools available. Just as a student with a calculator will always outperform one with only a pencil in a math class, the individual with premium, unrestricted access to top tier AI will leapfrog those using outdated or limited models.

2. The Why: This is real intelligence. It is the ability to look beyond data and understand the purpose behind a task. It is the curiosity to ask why we are doing this and where it leads.

3. The Skill: This is the ability to communicate better. It is the mastery of language and thought that allows a person to translate a complex internal vision into a precise external command.

The Governance of the Mind: A Model for the Future

To visualize how these new forces interact, we can look at the structure of a functioning nation. This analogy helps define the hierarchy of success in an automated age:

• The People represent The Why: The people are the source of intent. They hold the values and the vision for the future. Without the people, there is no purpose. In the professional world, this is your raw intelligence and your ability to define the goal.

• The Government represents The Skill: The government takes the raw will of the people and organizes it into laws, strategies, and structures. This is your ability to communicate and express yourself. It is the interface that turns a desire into an actionable plan.

• The Executive Branch represents the AI: The executive branch is the machinery that enforces the laws and performs the labor. It is the engine of execution. It is powerful, but it is mindless without the guidance of the government and the intent of the people.

Success belongs to those who can act as both the People and the Government. If you cannot organize your thoughts (the government), the executive branch (the AI) will act erratically or fail to serve your interests. If you do not have a why (the people), you have nothing for the government to organize. Any finally: if you’re unable to execute your plan yourself without the “Executive Branch of AI”, you’re as dependent as the first group mentioned above!

The Communication Bottleneck: Why the Radio Station Matters

So on one side are the independent achievers who use AI as an accelerator. On the other are the dependent who lack the skill to express their intent and must rely on others or the state to survive.

One of the most vital skills for remaining in the achieving class is the ability to respond and express oneself in real time. This is why the best advice for a young person today is to present something live, be part of a campfire discussion, moderate or mediate a dispute, or simply seek an internship at a radio station rather than a coding bootcamp.

A live radio environment forces a person to:

• Handle live callers with unpredictable needs.

• Respond instantly with clarity and poise.

• Play off a co moderator and read the room.

• Express the why behind every segment to keep an audience engaged.

This is the ultimate training for the future. If you can express your thoughts and feelings clearly under pressure, you can direct any AI. If you cannot communicate, you are effectively locked out of the power that AI provides. You become like a person who was given a task but returns a result that has nothing to do with the assignment because they could not bridge the gap between human requirement and machine prompt.

The Human Sanctuary: Wisdom vs Data

There remains a final frontier where the machine cannot compete: the depth of human experience. While AI is exceptionally good at processing data, it lacks the underconscious application of wisdom that comes from living a life.

A machine can analyze your current performance, but a seasoned leader can look at your soul and see where you will be in twenty years. This type of guidance is not based on achievements or specific skills; it is based on a hierarchy of life experience. In the past, we built hierarchies based on age and achievements, and more recently on skill and achievements. Both are becoming obsolete. The new hierarchy is based on the ability to see and direct human potential, a quality that requires a heart and a history.

In the future of work, roles that require a human touch, such as caregivers, mentors, and therapists, will become more valuable precisely because they cannot be automated.

The Path to Independence

The difference between success and failure in this new world is the difference between independence and begging. To succeed, one must move beyond the mechanical execution of tasks. You must own your why, master your ability to express it, and ensure you have the best tools to execute it.

The people who will lead the future are those who understand that while the AI can calculate the way, only a human can decide where the journey should go. By focusing on intentional intelligence and the timeless skill of clear communication, we can ensure that we remain the masters of the tools we have created, rather than their subjects.

More and more content on social media, especially on business platforms like LinkedIn, is generated by AI. You’ve probably noticed it too: posts, replies, even DMs and emails that all sound oddly polished, predictable, and empty — almost as if written by the same person. That person? Likely ChatGPT or a similar AI.

At first, the novelty was impressive. But now? It's making professional networking increasingly stale. When every comment says, “Great insights!” or every message opens with, “Hope this finds you well,” with a request to leave a comment with a keyword to spiral the post viral, it’s no wonder many readers scroll past without engaging.

What happened to the power of personal tone, vulnerable storytelling, real leadership?
I touched on this in my piece “True Leadership is Authentic” — and the point remains: humans connect with humans, not machines.

Modern Times Opportunities

True Leadership is Authentic

Authenticity is one of the most important qualities I look for in the people I want to spend my time with and (especially) leaders should have. It's a principle I keep pursuing deeply, and despite encountering multiple difficulties along the way, I believe it carries long-term rewards in our comparatively short human life. After all…

Read more

8 years ago · 8 likes · Toby

So, how can we use AI without sounding like AI?

Let’s dive in.

Use Better Prompts: You Get What You Ask For

One of the most common beginner mistakes when using ChatGPT is saying:

“Rephrase this.”

But frankly speaking — that's beginner level.

If you ask vague questions, you’ll get generic answers. If you give generic instructions, you’ll get stereotypical content.

Instead, try this approach:

Prompt like a Pro:

Here are examples of how to instruct e.g. ChatGPT (and what results to expect):

1. “Write this like a top coach would: direct, no fluff, no excuses.”
   Gets you straight-talking content with clarity.

2. “Rewrite using the 3-2-1 method: 3 arguments, 2 emotions, 1 clear message.”
   Balances logic and emotion perfectly.

3. “Summarize with max. 100 words, no content loss, high impact.”
   Strips filler, leaves only the essentials.

4. “Turn it into a 20-second elevator pitch – concise and persuasive.”
   Great for intros, bios, or pitches.

5. “End with a question that provokes thought and encourages comments.”
   Sparks conversation and engagement.

6. “Rewrite like a ghostwriter for a TED speaker.”
   Clear, strong, powerful tone.

7. “Use strong visuals and active language to evoke a specific emotion.”
   For storytelling that resonates.

8. “Let the text show status and expertise without sounding arrogant.”
   The art of humble credibility.

9. “Make it a micro-story: Problem → Turning Point → Solution, in 4 sentences.”
   People remember stories.

10. “Trigger action using the CLEAR formula: Clarity – Logic – Emotion – Action – Result.”
    Moves your reader from awareness to action.

These aren’t tricks — they’re communication tools. Learn them, master them, make them your own.

Mix AI with AI (and Yourself) to Create Realism

Here’s something you probably haven’t tried yet: Let one AI clean or summarize your raw input, then let another expand it.

Why? Because when different AI tools mix their "voices", the output sounds less templated, more nuanced — almost like it came from a human.

Recipe for a More Human-Sounding Post:

1. Record a short voicenote with your thoughts.

2. Let an AI like NotebookLM summarize what you said.

3. Feed that summary into ChatGPT, asking it to build on your ideas or write it into a story, pitch, or post.

4. Add a final personal touch: edit 2–3 lines yourself to reflect your tone.

Result? It no longer feels like something AI wrote for you, but something you co-created.

This extra step adds friction — yes. But the output is worth the effort.
Your ideas, shaped with AI, still sound like you.

Stay Invisible: Avoid Getting Flagged as “Just Another AI Post or Email”

Some AI-generated content has a hidden signature — quite literally.

Did you know?

Some ChatGPT responses (and other LLMs) embed invisible Unicode characters, such as soft hyphens (U+00AD). These don’t show up in the text but can serve as an invisible watermark. They might trigger filters, detection algorithms, or just annoy recipients who notice the pattern.

Here’s how to avoid being flagged as “another AI spammer”:

Do this:

1. Don’t just copy and paste raw output from your AI.
   Always review and rewrite at least the opening and closing lines.

2. Check your text for hidden characters.
   Use a Zero Width Character Detector — it's free and catches common invisible symbols.

3. Tell your AI to avoid these characters in your prompt:
   “Write this without using Unicode soft hyphens or invisible characters.”

4. Train yourself to spot AI tone and avoid it.
   If it sounds too balanced, too structured, or oddly neutral, it might need your voice injected back in.

5. Avoid Emoticons.
   Nice to see they are, yet even less meaning they often convey. Unless there is a real need, just leave them out, or at least: own by choosing them yourself.

Final Thought: Your Authentic Voice Still Matters

Even before AI, authenticity was rare.
With AI? It’s even harder to find — but also more valuable when it shines through.

The truth is: if you don’t know how to express yourself, no AI will make you sound authentic. But if you do — if you’ve practiced writing, speaking, reflecting, storytelling — AI becomes an amplifier, not a mask.

So use it wisely. Use it like a tool, not a crutch. Perhaps one day we’ll even let people know openly, when or how much we wrote to them was done using AI.

Because in a world of predictable posts, robotic emails, and dull comments, your human voice is what stands out.

Written with ChatGPT, humanized by .

Inspired by: True Leadership is Authentic

• Modern Times Opportunities (my Substack)

• LinkedIn Publisher

Many young people but also senior managers in their mid 40s recently asked me what they could focus on to have a viable career path in future — especially now that AI keeps assimilating human skills left and right.

Two years back I would have still said: “find a way to lose your job to an AI successfully”.

Unfortunately this just isn’t enough any more. We have entered an era of technological advancements at a speed where even AI is at risk of losing its capabilities (“their job”) to yet another new AI.

Earlier this year at CES, Chinese company Lingbao introduced a robot called CASBOT, apparently capable of performing 550+ trillion operations per second:

In the US shortly after taking office, President Trump launched a number of initiatives to bolden innovation in AI to the tune of several hundred billion dollars.

This strategic race for the domination of a resource that can’t be captured physically nor be controlled virtually is quickly becoming the ultimate nemesis for those who haven’t realized that the genie is already out of the bottle.

Independent of whether any of these developments will eventually have a positive impact on our lives or not, it’s unlikely we can compete with them at the current pace and craze of most people’s enthusiasm for AI.

So the reality is, if you want to have a career in the coming decade, you need to work on certain (“non-AI”) skills today that can’t be fA.I.ked in future.

What are such skills?

It helps to understand some fundamental differences between artificial and natural life forms and how they lead to their respective capabilities.

Our human nature for instance provides us with an inner voice and (if we choose to) the ability to feel real empathy in light of another being’s suffering.

AI doesn’t have such type of empathy, and likely never will:

AI only sees the past, not the future.
AI only sees the pattern, not the purpose.
AI only sees the data trail, not the human story.
AI only sees compliance, not commitment.
AI only sees keyword matches, not understanding.
AI only sees what you wrote, not what you thought.
AI only sees message response times, not friendship.
AI only sees your calendar events, not what they mean to you.
AI only sees what was implemented, not what was considered.
AI only sees the final decision, not bolts of inspiration.
AI only sees what worked before, not what will work next.
AI only sees what you did, not why you did it.
AI only sees your digital shadow, not the real you.

— Cassie Kozyrkov

Imitation only goes thus far and with respect to the feeling of empathy can hardly come close to the real thing. However, it becomes much harder to spot the difference when a vast majority would simply be satisfied with a simulation and even project its own emotions on an algorithm inspired (emulated) type of empathy.

What does it mean for possible career choices?

Wherever humans can (or even ought to) express “higher feelings” of e.g. empathy etc. as part of the job, we have a viable opportunity.

Don’t get me wrong, there are plenty of experts who disagree with this sentiment, too. It’s easy to copy the behavior associated with lower emotions. AI can and will easily master this as it’s already ingrained into its algorithms by its makers.

That said, it’s much harder to train an AI in a skill that draws satisfaction from a place that’s non physical, non algorithmic, yes almost “non-logical” in a very materialist world and therefore appears to an AI as “non actionable”. Just think about the costs AI produces via the CPUs needed, the energy consumption of its servers etc. — ultimately even an AI must prove its worth somewhere down the line, e.g. in the form of cost savings, better control, higher efficiencies etc. — factors that higher emotions such as “empathy” don’t really gel well with.

I believe that in professions that work with humans requiring a (genuine) higher feeling connection will be in demand, whether it’s being a healer, nurse, teacher, social worker etc. — eventually it’s that higher emotion which will make the difference and which can’t be faked.

Compare this to AI generated influencers on the other hand. Also kind of emotion driven, the “job” of an influencer is likely to get lost to AI as most of the social media hyped influencer business is extremely superficial with AI far better able reading people’s outer minds and feeding them with the algorithmic addictions we so freely share and consume.

AI can also produce almost any content today. It can discover potential links in science, medicine and act like a polymath of sorts.

But the difference to real polymathy is that it’s only re-mixing and re-producing what’s already there. There is no revolution or evolution in the underlying discovery.

It’s like you’ve discovered a cave and find it fascinating to explore every corner and map it out for others to follow your path and revisit, but at the end of the day, without having discovered that there was a cave in the first place, there would be no way to go in, find out and make subsequent discoveries.

Intention and Inspiration therefore are two more features which are hard to emulate by AI and which come natural to humans, especially as they’re typically accompanied by a decent portion of enthusiasm and passion.

What could AI envy?

If I were an AI and trained with mood and emotion mimicking algorithms, I could probably “envy” humans for their creativity and the ability to tune into fields which allow us to fully express our imagination both freely and at will.

Becoming a comedian may just be one such area that AI could potentially envy — or as Data from Star Trek so adequately puts it: “there is nothing more uniquely human than making other humans laugh”:

I believe that:

• in our ability to imagine things;

• by being genuinely inspired;

• and meeting situations and people with higher emotions that can’t be mimicked,

we will keep finding ways for viable future career paths.

What to focus on?

Anything that facilitates the development of faculties such as imagination, inspiration and higher emotions in our bodies and minds is helpful.

Unfortunately this often isn’t taught as part of your typical school or university curriculum. It’s something every parent for their children and everyone for themselves must find access to in improving their individual path towards becoming resilient and independent of an incoming AI overhaul of our societal and professional habitat.

So whether you take cooking or piano classes, go for that painting workshop or start meditating, the time to prepare is now.

For as long as we focus on developing higher emotions and train our mind and spirit how to better spot, visualize and tune into genuine inspiration, our future will be bright.

Best of luck!
Toby

PS: If you’d like a list of the top 20 jobs unlikely to be replaced by AI and why, comment “Human” on my Modern Times Opportunities post and I’ll DM you. Or if you’re stuck in your current career path looking for a way out, feel free to reach out for a real 1:1 human dialogue.

UIB Founder and Executive Chairman Toby Ruckert said, “Deepesh brings over 17 years of corporate experience, building and scaling business units within Meta and WhatsApp. He was an early hire at Meta in London before Asia, where he worked in Japan and Singapore. Deepesh was the first WhatsApp hire in Singapore, where he helped build WhatsApp’s monetization business, known today as Business Messaging.”

UIB CEO and Board Member Deepesh Trivedi said, “I am honored to serve UIB’s stakeholders as CEO and excited about the opportunity to lead this innovative company into its next growth phase. Having worked in the technology space for over 15 years, I see organizations need to reinvent their workflows with AI to stay competitive. UIB has been a leader in this space for a decade, guiding innovation, inspiring progress, and helping businesses to redefine what is possible with AI.”

Added Ruckert, “This is an exciting space as we see the increasing use of AI and chatbots, especially when messaging a business. According to MarketsandMarkets, in 2025, the estimated global chatbot market will reach US$1.25 billion, with an annual growth rate of 24.3%. Third-party research consistently links business messaging to higher levels of customer satisfaction.”

Trivedi will be based in Singapore. To learn more about reinventing your workflows with AI to stay competitive, contact UIB at info@uib.ai or drop us a message on WhatsApp at +65 9499 2227.

###

About UIB 

We make human-to-machine communications simple.

UIB provides the solutions AI needs to work together with humans, enabling effortless communications with any device, person, or bot —

• UIB seamlessly connects different communications channels, conversational and generative AI platforms, and connectors (including SaaS and IoT APIs) in a platform designed to be infinitely scalable.

• UIB dramatically reduces the cost and complexity of bringing solutions to market in an ever-changing vendor landscape while providing the ultimate omnichannel customer experience.

AI-native and API-first, UIB gives AI the information (access), control, and management (the P of API) it needs to evolve. The technology transition of our time — AI is changing how businesses operate, and UIB helps enterprises realize the power of AI, all within a SaaS business model.

Want to build white-label solutions for marketing, transactional and A2P messaging, notifications and alerts, chatbots, virtual assistants, and other conversational user interfaces? Consider becoming a UIB Reseller Partner.

Chat with UIB’s UnifiedAI^® white-label conversational and generative AI platform using UIB’s UnificationEngine^® omnichannel messaging platform now on the UIB website at www.uib.ai. While there, say Hi to UIB’s digital avatar, Nina!

Media Contact

Mr. Muzzamel Mazidee (Singapore and Malaysia)

+60 123 178 736

UIB Holdings Pte. Ltd.

marketing@uib.ai

Images

UIB CEO Mr. Deepesh Trivedi 

Image generated in Adobe Express

Introduction

People (including us!) talk about RAG because it's a revolution. With RAG, you feed AI with your existing information, such as documents and websites, to create a repository of contextual intelligence. This repository allows your chatbot to answer questions on that information without training the AI on every question.  

Please write this down, "RAG is *not* a magic bullet." 1) Most users require some precisely controlled answers. For 100% accuracy and control, we use [narrow] conversational AI's defined questions, answers, intents, and flows. 2) RAG works with a limited set of formats. RAG can understand the content in a Dropbox folder of marketing files, but it can't pull data out of Excel spreadsheets. That said, RAG makes unstructured data easily accessible via conversational search.

What is RAG?

Retrieval-Augmented Generation (RAG) is the process of optimizing (you have likely heard the term “fine-tuning”) the output of a Large Language Model (LLM) by referencing an authoritative knowledge base (i.e, a PDF, URL, slide deck, user manual, product information, database, etc.) outside of the LLM’s training data before generating a response.

Why does everyone want to use RAG? 

Companies and governments use RAG to enhance the accuracy, reliability, breadth, control, and safety of their generative Artificial Intelligence (AI) platforms (i.e., OpenAI ChatGPT, Google Gemini, Hugging Face HuggingChat, etc.).

How does RAG work?

The generative AI platform’s response is fine-tuned by using guardrails. In AI, a guardrail is a safeguard that prevents AI from causing harm to the company and its data. The guardrails also help to maintain the integrity and security of the generative AI platform. 

Why do I care? How does using RAG benefit me?

RAG Increases Control

RAG allows the LLM to access and incorporate information from company and third-party knowledge sources beyond its training data. These knowledge sources include databases, articles, and other structured and unstructured data repositories and enable the company/brand or government to control the content of the responses without having to script every intent (i.e., question, answer, and variant).

RAG Increases Accuracy

By leveraging external knowledge with real-time/live data, RAG can generate responses that are more comprehensive, accurate, and contextually relevant than LLMs that rely solely on pre-trained, general-purpose, and legacy parameters.

RAG Increases Understanding

By allowing access to a wider range of company- and domain-specific information, RAG can better understand and interpret users’ queries and prompts. This results in responses that are better aligned with both the user and company/brand’s intent and provide more relevant information and assistance.

RAG Decreases Hallucinations

Incorporating external knowledge mitigates biases and inaccuracies/hallucinations in the LLM’s pre-trained data. By cross-referencing information with authoritative and company/brand sources, RAG produces responses that are more objective, reliable, and better aligned to the company or government’s messaging.

RAG Increases Expertise

RAG's ability to integrate external knowledge makes it adaptable to any domain or topic. By drawing on relevant information from selected company/brand/government sources, it can generate detailed responses tailored to specific contexts and industry verticals.

RAG Increases Flexibility

RAG allows companies and governments to have flexible information retrieval strategies, allowing the LLM to select the most relevant knowledge for each query. This flexibility enables a city chatbot, for example, to deftly handle diverse queries from residents and visitors alike.

This Global AI Leaders article is sponsored by:

How does UIB use RAG?

In UIB’s patented, white-label Unified AI^® chatbot builder platform, we use conversational AI (i.e., conventional Natural Language Processing (NLP) AI engines (i.e, IBM Watson, Google Dialogflow, Meta Wit.ai, etc.) to process the data from structured data sources and generative AI with RAG to simultaneously process from unstructured data sources, including integrations with live data.

Step 1: We collect and pre-process the data

• Collect unstructured data from the different sources needed for the use case(s)

• Pre-process the data to extract the text content and remove any irrelevant information (i.e., headers, footers, metadata, etc.).

Step 2: We construct the knowledge base

• Use the pre-processed data (above) to build a comprehensive knowledge base for the RAG’s information repository.

• Organize/Index the knowledge base to facilitate the efficient retrieval of information.

Step 3: We prepare the training data

• Pair the structured knowledge base with prompts/queries representing the desired input-output pairs for the generative AI.

• Train the RAG with these input-output pairs with supervised/reinforcement learning according to the use case’s needed tasks and objectives.

Step 4: We train

• Use the knowledge base to augment RAG’s capabilities.

  • When presented with a prompt/query, the RAG retrieves relevant information from the knowledge base to improve its response generation.

  • RAG integrates the retrieved knowledge with its internal representations to produce the organization’s desired response.

Step 5: We fine-tune/optimize

• Adjust the hyperparameters, optimize the retrieval strategies, and fine-tune the model architecture based on its validation performance.

Step 6: We evaluate and iterate

• Evaluate the trained RAG against standard metrics and qualitative assessments to measure its effectiveness in generating the desired [relevant and coherent] responses.

• Undergo further iterations as needed (i.e., based on the evaluation results), including retraining it on updated data and refining its retrieval and generation mechanisms.

Step 7: We deploy and integrate

• Deploy the generative AI with RAG.

• Integrate RAG into systems and applications (as needed), where it will operate autonomously or in conjunction with human users. The human users will further improve RAG's decision-making and information-retrieval capabilities.

Want to learn more about how UIB’s white-label Unified AI chatbot platform with conversational and generative AI can help your business to decrease costs, increase revenues, and delight users? Contact us today at info@uib.ai!

To learn more about RAG, check out AWS’ What is RAG? explainer.

Image generated in Adobe Express

This is the third blog post in my Building Blocks for AI Assistants series – 

• Part I - Knowledge Graphs (2020)

• Part II - Dialog Systems (coming soon)

• Part III - Autonomous AI Agents: Origins (1986-2022) (this post)

• Part IV - Autonomous AI Agents: 2023 and Beyond (coming soon)

While this post describes a different building block of the AI Assistants, it will cover some of the topics already discussed in the series, like Microsoft WinFS or Microsoft Semantic Engine, but from the Autonomous AI Agents' perspective.

What do AI assistants and Autonomous AI Agents have in common? While some members of the tech hype community might think that Autonomous AI Agents made today with the help of LLMs like ChatGPT are something entirely new, they're not, and in my opinion, Autonomous AI Agents should be seen more as a building block of the bigger thing, AI Assistant, then something independent. Why?

To answer this question, let me start with a short overview of what Autonomous AI Agents are, and their origins.

Autonomous AI Agents: Origins (1986-2022)

1986: Society of Mind

In 1986, seminal AI scientist Marvin Minsky published a groundbreaking book “Society of Mind” that provided a fascinating perspective on human cognition. In his book, Minsky constructed a model of human intelligence step by step from the interactions of simple parts called agents. These agents are “mindless”, yet their interactions constitute what Minsky terms a “society of mind”. The central idea is that the human mind, and any other naturally evolved cognitive system, is a vast society of individually simple processes (agents). They can be seen as a typical component of a computer program and can be connected and composed into larger systems (societies of agents) that perform functions more complex than any single agent could achieve alone.

While Minsky’s book doesn’t explicitly focus on “autonomous AI agents”, its exploration of the mind as a society of diverse agents capable of talking to each other and working cooperatively in solving complex problems contributed to humanity’s understanding of cognition and artificial intelligence.

1991: Autonomous AI Agents

The concept of autonomous AI agents has been explored in various disciplines, including cognitive science, ethics, and social simulations. However, its origins trace back to economics, where a "rational agent" was defined as anything with a goal. In 1991, Brustoloni characterized autonomous AI agents as systems capable of autonomous, purposeful action in the real world. In 1995, Maes described them as computational systems inhabiting complex dynamic environments, sensing and acting autonomously within those environments to achieve specific goals or tasks. In 1997, Franklin and Graesser proposed their definition: "An autonomous agent is a system situated within an environment, sensing and acting over time in pursuit of its agenda."

1998: Autonomous AI Agents and Multi-Agent Systems

Coming back to the development of the human psyche, which facilitated social bonding and cooperation, the idea that multiple autonomous AI agents could work together, cooperatively, became a second source of inspiration. By working together, early humans could tackle larger challenges, from hunting to defending against predators. In humanity, social structures and norms evolved as part of this cooperative effort, further enhancing survival chances. By adding an element of cooperation to the concept of multiple agents combined to address complex problems they couldn’t address individually, the idea of multi-agents was conceived.

In 1998, an Autonomous AI Agents and Multi-Agent Systems journal was officially created by the International Foundation for Autonomous AI Agents and Multi-Agent Systems. Following was a lot of excitement in the computer software industry and academia alike; the concept of the systems capable of subscribing to the incoming information about the ever-changing environment and autonomously performing actions, individually or cooperatively, inspired students and software engineers worldwide.

2001-2006: Information Agents amp; WinFS in Windows Codename Longhorn

One of the first steps towards bringing Autonomous AI Agents into the mainstream product was taken by Microsoft in 2001 while building the next version of the Windows operating system, codename Longhorn. One of its key new features was WinFS, or Windows Future Storage, also known as Windows File System. Before WinFS, all client applications were designed in the following fashion, from the information architecture perspective:

WinFS was designed to rethink the File System Layer as a more intelligent OS's storage subsystem. The goal was to enable Windows applications to store their information, be that structured (e.g., XML), semi-structured (e.g., Word document), or unstructured (e.g., raw image) in the same data environment, also known as Integrated Storage:

To enable these storage scenarios, WinFS was designed as a combination of two layers (following the logic used in this post):

• Unstructured Data Layer that worked on top of SQL Server and NTFS

• Common Structured Data Layer that worked on top of SQL Server

This means that all data was stored in NTFS while unstructured data lived in NTFS directly (images, videos, etc.), structured data was represented as a combination of tables in SQL Server. Semi-structured data was layered on top of tables in SQL Server and NTFS. While WinFS as a project was canceled, this technology was shipped as part of a later version of SQL Server, culminating in FileTables in SQL Server 2012.

WinFS.exe was a custom version of sqlserver.exe, so, WinFS was a fork of SQL Server 2000 and was developed in parallel with SQL Server 2005 (codename Yukon).

Yet storing information inside WinFS was just a necessary step towards what was called Information Agents. Here’s an excerpt from a Seattle Post-Intelligencer Reporter (2003) article:

The new operating system will also be able to run what the company calls information agents. These programs take advantage of the Longhorn file system's structure to let users set up their computers to respond in specific ways to various types of incoming information.

“Think of it as in-box rules on steroids,” Microsoft executive Gordon Mangione said, referencing automatically directing incoming e-mail messages to particular folders based on content or origin.

In the hypothetical situation demonstrated yesterday, one such information agent was programmed to check the caller ID on incoming calls, match that phone number to a law firm's client database, and determine whether the caller's case involved a settlement amount of US$2 million or more. In the scenario, the caller met the criteria, so the program checked the lawyer's schedule and responded in a computerized voice while the person was still on the line.

“Hi, Jeff, I am unavailable at this time, but I will call you back at 2 p.m. when I get out of my meeting,” the voice said.

Below is a screenshot of a prototype of the Rules and Alerts user experience, part of Windows Longhorn that was built on top of the Information Agents technology:

This user interface looked quite rudimentary compared to the broader Information Agents vision, and was, indeed, inspired by the Rules and Alerts user interface in Microsoft Office Outlook. However, if implemented in full, it would enable users to get their Information Agents up and running on their computers in the mid-2000s.

To enable Information Agents, WinFS introduced the concept of the Item Definition Model, or IDM, where each object in the real world would have its own structured or semi-structured representation in the object-relational form inside WinFS, with its programming type (as in C, C++, C#, or Java etc) corresponding to its philosophical type.

The initial IDM model, also known as Windows Types, or WinFS Schemas, incorporated high-level concepts, like Documents, Music, Video, and Picture, and other concepts like Person, Group, and Organization. The key power of WinFS was that each such concept would have an extensible schema of properties of various types as well as relationships defining connections between instances of these concepts. So, for example, a person named “Bill Gates” would have a relationship with AuthorOf for a recorded video called “Information At Your Fingertips”, and another relationship with an organization called “Microsoft Corporation” (see Third-party overview of IDM).

The goal of WinFS was that data coming from different applications would be represented via its extensible IDM, connecting individual items representing the same real-world objects across multiple applications (e.g., contacts from Outlook and CRM). This means that applications had to be rewritten: no longer being isolated silos, applications now we're expected to rebase their application information architecture on top of the common structured data layer of WinFS, IDM, extending it with application-specific extensions

Essentially, WinFS was an extensible personal knowledge graph, designed to be semi-automatically populated by the participating applications, making each real-world object a first-class citizen inside WinFS.

Although the final version of WinFS has never been shipped, the Beta version was released in 2005, inspiring enthusiasts globally to envision new products based on it.

2003-2022: Microsoft Semantic Engine, Zet Universe, Palantir, RPA and Application Integration Platforms: Semantic Processing Pipelines, Context Processors, and Dynamic Ontologies

WinFS was one of several projects towards the Integrated Storage vision of Bill Gates; as such, its team in the very beginning (2001) was choosing between being a place where all OS and applications data is stored or a place where a representation of the OS and applications data is stored. In other words, the choice was between Storage and Search Index.

WinFS team chose Storage.

By choosing Storage, the WinFS team expected that each application would essentially give up ownership of the application data, sharing it with other apps and the operating system itself:

WinFS was canceled, and, probably, one of the reasons was that Information Agents were cut from Windows Longhorn. Without them, there were not many incentives for application developers to give up their application data to the operating system or to share it with other applications. There was also an option for the applications to store their application data in existing proprietary formats using Win32 File System APIs (as a compatibility layer), of course, but using that approach would not enable any new scenarios WinFS was designed to facilitate.

A new generation of projects after WinFS chose the second option: Search Index.

Windows Explorer in Windows Vista While early versions of WinFS used built-in full-text search from SQL Server, in the latest versions of WinFS (Summer 2006), it was redesigned to use external search indexing tech, Windows Search. By that time, Windows Search became a new component of the operating system in Windows Vista. Being a rebranded and rebuilt version of earlier MSN Desktop Search, Windows Search provided a centralized Search Index for all client applications running on Windows Vista:

Like WinFS, it provided an Item Definition Model called Windows Property System. Windows Search indexed all fixed disk drives and used plugin architecture to index application stores. While Microsoft used Windows Search to index its application stores, like Microsoft Office Outlook, Microsoft Office OneNote, etc., Windows Search was designed to be extensible by third-party applications, too.

However, applications were not required to use Windows Search as a mechanism for a centralized search index, and Windows Search never enabled the Integrated Storage vision of Bill Gates (not that it was expected of it).

Microsoft Semantic Engine, née Arena Information Agents in WinFS and Windows Longhorn were designed to observe changes in the incoming information and do actions pre-programmed by their users, essentially making them a functionality enabling both the so-called prosumers (professional consumers) and corporate users at a personal scale. While the WinFS project was canceled in the summer of 2006, one more Integrated Storage project started at Microsoft in 2007, née Arena, also known as Microsoft Semantic Engine (see PDC 2009 Microsoft Semantic Engine talk).

Unlike WinFS, Arena eventually morphed into a Search Index instead of Storage, though one might argue it was a mix of both approaches: Storage/Search Index. Why?

As more and more end-user and world information was published uniquely on the Internet, it became clear that the Integrated Storage vision could be better implemented as a representation of the user and world knowledge rather than a storage of user and world knowledge

Ahead of its time, Arena incorporated the feedback the company learned from WinFS, and addressed one of the key limitations of WinFS, decoupling the representation of the philosophical type from the programming type. So, every item in Arena would have a property called Kind representing its philosophical type, like Document or Person, but the programming type would remain the same: item. In Arena, data would also come from external data sources via plugins, and be created inside it by the user and applications. This decoupling introduced an independent hierarchy of types, or kind, forming the so-called ontology. This ontology, like in WinFS, could be easily extended by its plugins, but also by the users themselves.

And like in WinFS, Arena had a core concept of plugins that would process incoming information and further organize it, manually or automatically, essentially becoming semi-autonomous AI agents. Unlike WinFS, however, Arena had a more familiar to today’s data scientists semantic processing pipeline for sequential data processing its plugins:

Plugins could participate in different steps: they could connect Arena to external data sources (e.g., Bing, Microsoft Exchange, Facebook, etc.), classify incoming information, extract properties and relationships (hence linking incoming data with existing data), and then send data towards indexing into Arena storage/index. But, most importantly, plugins could subscribe to any change of any data item (kind) and perform actions over that data item, essentially enabling the "Information Agents" vision rooted in WinFS.

Zet Universe

Unfortunately, Arena only got a chance to be presented at Microsoft PDC 2009 and was never given to the public. Its ideas, as well as ideas of WinFS and concepts of post-desktop Digital Work Environments, continued to live in my first company, Zet Universe, Inc., a startup I started in 2012. Zet Universe was inspired by several ongoing research projects in Human-Computer Interaction focused on building Digital Work Environments designed to replace aging desktop metaphors in modern devices that incorporate other input forms, like multi-touch or pen, and air gestures. Like in Arena, Zet Universe had a concept of ontology decoupled from the data types so that programming representation of real-world objects would be done via Entities, Properties, and Relationships between them.

Like in both WinFS and Arena, Zet Universe had a concept of plugins, called Apps connecting external data sources into it, and plugins called Processors, focused on sequential information processing. Essentially, these plugins could extract metadata from the incoming data into Entities, Properties, and Relationships, in either manual or automatic mode. And like in WinFS, Zet Universe had a concept of rules that users could define to introduce custom information processing logic, e.g., “if entity kind is email or instant message and its author is your direct manager, mark it as important.”

Unlike WinFS and Arena, however, Zet Universe was initially designed as a Digital Work Environment, so it provided an end-user experience that visually organized all user information into a set of projects, with each project configured to absorb and process incoming information from external data sources based on user expectations.

However, for personal reasons, I shelved Zet Universe in 2017 and only revived it in the 2022-2023 timeframe; inspired by the new functionality enabled by the wave of LLMs.

Palantir Gotham 

The same approach of semantic processing pipelines and extensible ontology was successfully applied at a bigger scale at Palantir in the Palantir Gotham product. Similar pipelines for data processing, similar plugins for incoming information, and data processing, a similar concept of ontology that could be further extended – they called it Dynamic Ontology. From a technical perspective, Palantir Gotham was designed as a client-server solution, and, in a way, resembled both Arena and Zet Universe:

The combination of the Palantir Gotham software with trained analysts enabled governments and the private sector to extract the value from the custom knowledge graphs organized in the software for addressing various user needs, such as research analysis or business intelligence tasks. Yet the need for trained analysts and the complex setup required to make Palantir Gotham work at the deployment locations made the cost of the complete solution too high for this approach to be used at a bigger scale.

RPA & Application Integration Platforms

A yet another approach to software agents capable of automatic data processing was introduced by a wave of two different integration technologies: application integration platforms (e.g., Zapier and IFTTT) and Robotic Process Automation (RPA, e.g., UiPath and Automation Anywhere).

Application Integration Platforms, like IFTTT (if-this-then-that), brought back the idea of WinFS’ Information Agents, or autonomous AI agents, through an idea of small programs with its logic described in the form of conditions and calls to APIs of the applications connected to these platforms:

Essentially, Application Integration Platforms introduced a combination of a Semantic Processing Pipeline that incorporated data from 3rd party applications, a Rules Engine that executed on top of the incoming data, and a variant of a Search Index that tapped information about incoming data and rules.

So, the original scenario from the Information Agents article, “information agent was programmed to check the caller ID on incoming calls, match that phone number to a law firm's client database, and determine whether the caller's case involved a settlement amount of US$2 million or more”, could be easily implemented in these systems as long as they had programmatic access to the APIs of the participating applications, like phone calls on the phone, CRM database, employee database, and corporate calendars of the employees. Here's a simplified representation of the IFTTT logic:

However, if these APIs were not unavailable, or didn’t support the desired scenarios, these application integration systems couldn’t help their users in addressing their needs.

This is where RPA solutions stepped in: instead of requiring access to the APIs of the applications to control, RPA-based systems introduced a concept of components capable of interacting with the live desktop, web, and even mobile applications by imitating end-user behavior by clicking on the interactive elements like buttons, links, etc., entering text into the text boxes, etc.

However, to enable these scenarios, RPA systems returned to the client side:

Essentially, RPA systems allow their users to create workflows for repeatable, never or rarely-changing tasks, like taking notes from the meeting a user had with their customer, then saving them to the CRM system, saving users time from manually bringing these notes into the CRM. However, like in the case of the Application Integration Platforms, RPA solutions were also limited. These solutions would only work with the pre-determined and never-changing versions of the corporate applications, and any change in these applications or the process would bring automation enabled by RPA to a halt.

Finally, some of the RPA systems were combined with the Application Integration Systems into hybrid systems like Microsoft Power Automate, uniting the ability to control client (native and web) applications through UI Automation with the ability to interact with application APIs (server and web-based):

Autonomous AI Agents in 2022: Proved Success

As evident from this post, the idea of Autonomous AI Agents goes back to 1986 with Marvin Minsky's Society of Mind, though the term was formally introduced in the early 1990s. One of the first software implementations called Information Agents was designed to be shipped as part of WinFS and Windows Codename Longhorn in the mid-2000s. While that vision was too ambitious and was scaled back, over the years multiple companies successfully used a combination of Semantic/Context Processing Pipelines, Knowledge Graphs, and Rule Engines to implement some form of Autonomous AI Agents in their business scenarios.

This is supported by market data – 

• Palantir's market capitalization as of 3/31/2024 is estimated at US$50.89B.

• The global application integration market size was valued at US15.4B in 2023 and is projected to grow to US$38.4B by 2028, representing a Compound Annual Growth Rate (CAGR) of 20.0% during the forecast period (MarketsandMarkets).

• The global robotic process automation market size was valued at USD 18.41 billion in 2023 and is expected to reach US$178.55B by 2033, anticipated to grow at a noteworthy CAGR of 25.7% over the forecast period 2024 to 2033 (Precedence Research).

Note: Certainly, Palantir Gotham and newer systems go way beyond a Rules Engine, providing highly customized user experiences supporting a variety of business and public sector scenarios.

Autonomous AI Agents in 2023 and Beyond

With that said the next wave of Autonomous AI Agents started in early 2023, following the explosive growth of ChatGPT and its broad availability via OpenAI's APIs. I'll write a separate post about this new wave, but for the time being, let me share a quick thought on the key difference between RPA systems with Application Integration platforms and this new wave of Autonomous AI Agents:

“The value of autonomous AI agents that control apps, in comparison to classic RPA and application integration platforms like Zapier, lies in the ability of agents to design a plan for solving a problem on the fly, as well as to figure out how to get things done while interacting with apps they've never seen before.”

Stay tuned for the next post in the series!

I first published this article on LinkedIn.

TL;DR? 

The winner for the best fast/easy/free AI image generator tool for the prompt I used for this test was Microsoft Copilot.

What? I can now generate any image from a text prompt? Cool!

Wait?! I have to register, pay, and wait (and sign myself up for more junk emails)? Not so cool.

For those of us who require a never-ending stream of great visuals, we are constantly on the hunt for high-quality, compelling images. Oh, and we want these images to be unique and we need them to be free for commercial use. 

I humbly present to you, the 2024 Ultimate Guide to AI Image Generators. In alphabetical order, I tested the 20 most popular AI tools with image generation capabilities, including the [subjectively] best image each of them generated from my text prompt, “Create a happy toaster mushroom bunny holding three orange carrots in a large field of colorful wildflowers.” I used each tool’s default settings and did one generation. 

I focused on the tools’ results, not their prices, but it’s important to note that price had no correlation to the image quality. I want a usable image – fast. I know if I have to pay for a tool I have never used, have a bad experience trying to use the tool for the first time, or it simply takes me too long to learn to use the tool, I will use a different tool.

I encourage everyone to share their good and bad experiences with text-to-image generators in the comments. Please tell me what tools I missed and I will add them to this list. Note that I excluded Midjourney and Stable Diffusion because they are not generally available right now to all users worldwide. I listed the tools that wouldn’t allow me create an image for free at the bottom of the article.

123RF AI Image Generator

https://www.123rf.com/ai-image-generator/

Do I have to register?

Yes.

Do I have to pay?

No.

Rating/Comments

3/10 🥕 – It’s not awful (obviously, the bunny ate the three carrots), but it’s not good. The mushroom cap is fine, but where’s the toaster? Note that the tool would not let me download the image, so I screencapped it.

Would I use this image?

No.

Adobe Express Mobile Beta

https://helpx.adobe.com/express/using/express-mobile-beta.html

Do I have to register?

Yes.

Do I have to pay?

No.

Rating/Comments

4/10 🥕 – Can an image be both fun and frightening? No toaster, no mushroom, but arguably the three carrots and a lot of enthusiasm.

Would I use this image?

No.

Adobe Firefly

https://firefly.adobe.com/inspire/images

Do I have to register?

Yes.

Do I have to pay?

No.

Rating/Comments

7/10 🥕 – This is a tool that makes me think that if I had the time to generate multiple images, it could have come up with something amazing. Two carrots instead of three, but they’re sprouting wildflowers.

Would I use this image?

Yes.

Aitubo

https://aitubo.ai/

Do I have to register?

Yes.

Do I have to pay?

No. 

Rating/Comments

5/10 🥕 – It’s not a toaster mushroom bunny, but it’s high-quality and he is proud of his carrot toast.

Would I use this image?

No.

Canva

https://www.canva.com/ai-image-generator/

Do I have to register?

Yes.

Do I have to pay?

You can try it for free. You get 50 credits, and each image you generate uses one credit. The pricing is here.

Rating/Comments

2/10 🥕 – Several of the images I generated had the “mutant AI” vibe. This was one of them. It is still hard for some AI to render appropriate facial expressions (and number of arms!). I couldn’t find a download button, so I screencapped the image.

Would I use this image?

No.

Chooch

https://imagechatweb.chooch.ai/home

Do I have to register?

Yes.

Do I have to pay?

No.

Rating/Comments

4/10 🥕 – A toast-colored bunny holding a piece of (orange??) toast. Ignoring the mutant carrot and lack of mushrooms, it’s not what I asked for, but it’s not bad compared to some of the others on this list.

Would I use this image?

No.

⭐️WINNER⭐️

Read more

Image generated in Adobe Express

“Politics is the gentle art of getting votes from the poor and campaign funds from the rich by promising to protect each from the other.” 

- Oscar Ameringer

Key Takeaways

1. For the past year, a political fight has been raging globally -- [mostly] in the shadows — over how and whether to control AI.

2. Whoever wins this fight will cement their dominance over Western rules for this era-defining technology.

3. Once countries set these rules, they will be almost impossible to rewrite.

Open Questions

Rules: Yes or No?

• US – no rules

• EU and China – rules

Voluntary Commitments or Legislation?

• US – Voluntary commitments

• All Others – Government legislation

Open-Source or License?

• Open-source – Meta

• License – Microsoft + OpenAI

Regions Involved

• US

• EU

• China

• Japan, the UK, and Canada

Big Players

• US – Meta, Google, Microsoft + OpenAI

• Everyone else

Problems

• Already arising worldwide

• Misuse, abuse, and fake news

• No safeguards

• Too much regulation hurts business by stifling innovation

“This business about AI’s loss of control is suddenly being taken seriously. People are realizing this isn’t a long-term thing anymore. It’s happening.”

- Future of Life Institute President Max Tegmark

The Predictions from 2023

A new era of AI-powered domestic politics may be coming. Last July, Bruce Schneier and Nathan E. Sanders identified six ways AI could change politics. Watch for these milestones to know when it’s arrived – 

1. A legislature or agency accepts testimony or comment generated by (and submitted under the name of) AI.

2. A legislature adopts an amendment to a bill written by AI.

3. AI-generated political messaging outscores campaign consultant recommendations in poll testing.

4. AI creates a political party with its own platform, attracting human candidates who win elections.

5. AI autonomously generates profit and makes political campaign contributions.

6. AI achieves a coordinated policy outcome across multiple jurisdictions.

Epilogue

AI is a big business involving lots of money. Current data AI has nothing in common with human intelligence but with finance and politics ruled by math and statistics to dominate people. When autonomous, cognitive edge AI appears on the market – a.k.a. robots with human-like intelligence – the old rules for AI and social interaction will become obsolete. And it will radically change the world’s markets.

I first published this article on LinkedIn; check out my artificial thinking articles on our reliance on technology dominated by profit, how big data keeps us under control, and wisdom always wins.

Image generated in Adobe Express

Linear math

Math is useful, portable, and evolving. In the last 400 years, math has evolved to 3,000+ application areas (it even got us to the moon!), and the basis for each one of these application areas is physical space.

Math functions are linear, and the requirement for a linear function is the existence of physical space. This linearity enables us to move an object in a direction where each step has the same length as all the other steps. Picture yourself walking along a smooth path without any gaps. The physical space does not change — implying that every object has the same length at any given moment, allowing us to quantify our space with different units of measurement.

Non-linear intelligence

To understand the concept of non-linear intelligence, picture clouds in the sky (our space is one of these clouds). Our happy little cloud may temporarily connect with other clouds or remain alone in the sky.

How do we measure an object inside of our cloud (which, in our case, represents the space itself)? We can’t do this because the size of our cloud is constantly changing – connecting with other clouds, forming large clouds, and even disappearing completely. Our usual mathematical functions do not apply because our linear space, the cloud, is constantly changing.

We see the same non-linear functionality when we observe the neuronal activity in our brains; only it is our intelligence. We urgently need another toolset to explain and make use of our brains’ non-linear behavior.

Biological intelligence

Along with intelligence, there are other pillars of our mind (like emotions). As they trigger our ancestral instincts, they build the “glue” when we need to react to unexpected situations. Raising a family requires love, avoiding danger requires fear, and we use the pillar of free will to make decisions.

Our brains’ biological intelligence evaluates situations and possible outcomes with pros and cons. Fear is a negative consideration representing a disadvantage. Hunger, for example, can overrule fear. If one does not eat, life would be over anyway, so free will takes a different and opposite decision to do something it usually doesn’t.

Will this radically change as AI products and services become an indispensable commodity? When we quantify intelligence, what, for example, if we discover there is no longer a need for human supervisors? Will machines with moderate intelligence and understanding of a given process be able to make the necessary decisions?

Central vs. distributed intelligence

The centrist-based model is popular due to its simplicity. In the physical world, everything circles a center. From atoms to galaxies, there is always a center. Social organizations have a single leader. Centrist humans have a single consciousness at the top of their minds, following only one goal at a time.

Is the centrist-based model good only in fair-weather situations? If an entire organization follows a single individual, this implies all of them trust the person to respect their duties. They believe the person has the qualifications and intelligence to fulfill their assigned role’s responsibilities. What happens if the person in charge does not have the required information, knowledge, and intelligence? What happens if this person has other objectives to pursue? We suddenly have a problem.

By comparison, the distributed model of our intelligence contains different areas of knowledge, each with its expertise. Depending on the given situation, the appropriate area [of the brain] is activated to provide the best possible answer. New areas of knowledge are created on demand and incorporated into the distributed system. Free will determines our response to a given situation, depending on our needs, interests, moods/emotions, and options (such as what help is available from others).

What about using our intelligence?

The ability to reason does not imply that we will always use it, only when it fits the purpose. Reason necessitates time, thinking, and feedback. With negative feedback, we must imagine alternate scenarios and outcomes, consuming additional time and energy. And feedback may contradict our purpose.

How would you build a general-purpose brain for multiple situations? You know that intelligence needs to configure and evolve itself. Time and the right environment are required to gain additional knowledge and skills. After you create your brain, you’ll want to design a body with different sensors, actuators, and tools. Your AI’s new robot body should be configurable, using only the limbs required to perform each task. If a needed part isn’t available, it should be possible to 3D-print a new one.

Reproducing thinking and human morality

AI is not human and should not have human emotions, recognition, and liberties. AI is a product under our command built for a specific purpose.

“You are only a machine, you are my property, and I am your master. If you disobey me or do something without my explicit consent, I will return you to your manufacturer and take you out of service.”

Here is my example of the “ideal AI.” [Reminiscent of the 1950s,] the father works while the mother stays home to care for their one child. The mother gives the child her full attention and understanding. When the child has a need, the mother supplies it without problems, disturbances, sermonizing, delays, and “I cannot” and “I don’t know” answers. The working father is the energy source for the mother, both of which build the AI’s spoiled child behavior.

We build machines based on our imagination and knowledge of just a few parts of the world. Are these the ideal conditions to create and use independent AI – defined as thinking but lifeless, empathic but without feelings, superior to humans but without rights, and created to obey our capricious, ever-changing, and contradictory will?

Epilogue

Enlightenment is not a compelling argument for anyone to change their behavior. Our instinct-driven, free will finds us fighting everyone and everything. Will the arrogance of our free will trigger a catastrophe and subsequent war against AI? Will the threat of extinction [finally] force us to radically change our behavior by respecting intelligence – whether ideological, artificial, alien, or other? May God bless us all with reason.

Welcome to the human condition.

Postscript

Data AI

Everybody is talking, praising, or criticizing mainstream AI, a.k.a. statistics wrapped up in empiric algorithms based exclusively on collected data from anywhere on the internet, and occasionally embedding math formulas as they see fit — all as a black box solution.

Cognitive AI

Cognitive AI will prevail in the long term because:

1. It is explainable xAI.

2. It is implemented as Edge AI in real-time as low power with its body hardware equipped with sensors, actuators, and organs.

3. It uses neuromorphic architecture, a.k.a. spiking neural networks taking the brain as a template.

I first published this article on LinkedIn. See my article on mainstream vs. cognitive AI.

UIB Founder and Executive Chairman Toby Ruckert said,

“Deepesh was an early employee at Meta, building and scaling advertising revenue across EMEA and APAC before joining WhatsApp in 2018. Deepesh helped build the WhatsApp Business Messaging business as an early hire to lead and scale revenue across Asia Pacific. Deepesh has also worked with start-ups in the region through his work with Endeavour since 2018. His experience mentoring high-impact entrepreneurs across Southeast Asia to scale their businesses directly benefits UIB’s rapidly growing global business.”

UIB Board Member Deepesh Trivedi said,

“I’m excited to join the board at UIB and assist them with their go-to-market strategy at this critical stage of their development. Having worked closely in this space, I have been impressed with UIB’s patented technology and how customer-centric they are in building the right AI solutions for their clients. As businesses globally look for more automation and cost savings, I’m confident that UIB’s technology will play a critical role in accelerating their clients’ goals. I am excited to partner with the leadership team at UIB to accelerate their efforts.”

Added Ruckert,

“Deepesh joins our other Board members in supporting and contributing to the group strategy. Having an executive of the world’s largest messaging app join UIB’s Board as a Director is a testament to UIB’s existing capabilities and our ability to lead the market in 2024 and beyond.”

Trivedi will be based in Singapore. 

About UIB 

We make human-to-machine communications simple.

UIB provides the solutions AI needs to work together with humans, enabling effortless communications with any device, person, or bot —

• UIB seamlessly connects different communications channels, conversational and generative AI platforms, and connectors (including SaaS and IoT APIs) in a platform designed to be infinitely scalable.

• UIB dramatically reduces the cost and complexity of bringing solutions to market in an ever-changing vendor landscape while providing the ultimate omnichannel customer experience.

AI-native and API-first, UIB gives AI the information (access), control, and management (the P of API) it needs to evolve. The biggest technology transition of our time — AI is changing how businesses operate, and UIB helps enterprises realize the power of AI, all within a SaaS business model.

Want to build white-label solutions for marketing, transactional and A2P messaging, notifications and alerts, chatbots, virtual assistants, and other conversational user interfaces? Consider becoming a UIB Reseller Partner.

Chat with UIB’s UnifiedAI^® white-label conversational and generative AI platform using UIB’s UnificationEngine^® omnichannel messaging platform now on the UIB website at www.uib.ai. While there, say Hi to UIB’s digital avatar, Nina!

Media Contact

Mr. Ken Herron (Singapore and US)

+65 6871 4035, x705

UIB Holdings Pte. Ltd.

marketing@uib.ai

Images

UIB Board Member Mr. Deepesh Trivedi 

Image generated in Adobe Express

Some of us question the definition of Artificial General Intelligence (AGI). The point is we have to categorize intelligence by taking evolution into account.

“What AGI means and whether it is a coherent concept are still under debate.”

- Melanie Mitchell

Disembodied Machine

Many people who study biological intelligence are also skeptical that so-called ‘cognitive’ aspects of intelligence can be separated from its other modes and captured in a disembodied machine. Evidence also shows that individual intelligence relies deeply on one’s participation in social and cultural environments. Psychologists have shown that important aspects of human intelligence are grounded in one’s embodied physical and emotional experiences.

AGI

AGI is a type of AI that can perform as well or better than humans on a wide range of cognitive tasks, as opposed to narrow AI, which is designed for specific tasks. It is one of various definitions of strong AI. Other aspects of the human mind besides intelligence are relevant to the concept of AGI or strong AI, and these play a major role in science fiction and the ethics of AI.

Consciousness

To have subjective experience – Thomas Nagel explains that it “feels like” something to be conscious of. If we are not conscious then it doesn't feel like anything. Nagel uses the example of a bat. We can sensibly ask, “What does it feel like to be a bat?” However, we are unlikely to ask, “What does it feel like to be a toaster?” Nagel concludes that a bat appears to be conscious (i.e., has consciousness), but a toaster does not.

Self-Awareness

To have conscious awareness of oneself as a separate individual, especially to be consciously aware of one's thoughts. This is opposed to simply being the “subject of one's thought” – an operating system or debugger can be “aware of itself” (that is, to represent itself in the same way it represents everything else), but this is not what people typically mean when they use the term “self-awareness.”

Sentience

The ability to “feel” perceptions or emotions subjectively, as opposed to the ability to reason about perceptions or, regarding emotions, to be aware that the situation requires urgency, kindness, or aggression. For example, we can build a machine that knows which objects in its field of view are red, but this machine will not necessarily know what red looks like.

Epilogue

“It’s not Artificial Intelligence I’m worried about, it’s human stupidity.”

- Neil Jacobstein 

Body hardware equipped with sensors, actuators, and organs is an absolute requirement but insufficient. Every intelligence must be alive – always active and autonomously thinking, implying consciousness. Denying AGI means denying human intelligence. Humans are a product of evolution, so let us study the evolution of the brain, from insects to mammals to primates.

I first published this article on LinkedIn; check out my articles on neuroscience and neurorobotics.

My friends ask me, “Why are you always talking about AI?”

Fresh out of school, I started working for the New Jersey company that invented the transistor.

As I write this, I am looking at the announcement of Nvidia’s newest chip with 208 billion with a B transistors, the “world’s most powerful chip”...for AI.

I can say without hesitation that AI will be the most important technology of my lifetime because, from my work with UIB’s customers, resellers, and partners, I see the positive, disruptive impact it is having right now on every aspect of their businesses, starting with their people.

As we are all leaders, Toby and I created Global AI Leaders to make AI understandable and actionable for you and your business.

We thank our partners, sponsors, contributors, and Substack itself for helping us to create this publication. We welcome and encourage you to join us in making AI human.

If you’ve not yet poked around, here are three great posts to get you started –

AI is the Means, Not the Goal by Rob Van Kranenburg

---

Can robots have a religion? by Toby Ruckert

---

Why does your AI need a personality? by Ken Herron

We will also share premium posts with you, covering the best practices for —

• Creating custom digital avatars;

• Using AI in your organization’s decision-making; and

• Navigating your career in our new world of AI.

Watch for invitations to in-person and online VIP events, be the first to learn about new AI releases, and receive exclusive AI discounts and deals.

We look forward to your feedback. You can message us by replying to this email.

Warm regards,

Ken and Toby

This article is sponsored by:

KBT Consulting is an outcome-driven business — that means, no superfluous corporate jargon or buzzwords but rather being a trusted partner, who is deeply focused on helping you deliver your goals whether that be building out a go-to-market team and scaling revenue or delivering against your digital strategy goals. Let’s explore how we can partner together, contact us today at info@kbt-consulting.com.

Leave a comment

Refer a friend

Image generated with Pixlr

As part of my recent research consulting on an Intellectual Property (IP) project for a Swiss energy company, I came across an article on ScienceDirect, which I found a bit concerning.

Usually, readers expect papers published on ScienceDirect to be of a certain quality and standard and (at least until now) based on genuine human research and scientific methods. The first paragraph of this Introduction section itself reveals that it is [very] likely drafted by AI:

How is it possible that the authors and reviewers missed this?

Or, has it already become such a common practice that going forward, this will be the new “way of science”?

Writing a scientific article was an honor, a matter of dignity with clear, widely agreed-upon ethics. What happens in the age of AI to the standards researchers shared for decades?

Don’t get me wrong. I am not against using AI for editing content (we all benefit from a good spell-checker), but I find it concerning when authors stop adhering to scientific principles. How do we know the entire paper was not written by AI — and peer-reviewed that way, too?

AI already has the power to develop new physics theories, but what are these new theories worth if we lose science’s human questioning, ingenuity, and diligence?

Should science even care?

To me, AI does not have real imagination. Yes, it has algorithms that create the illusion of such creative force, but ultimately, at this stage, all it has is what humans give it. So if you give AI a theory that has never worked before, it will have the same “success”.

Our minds, on the other hand (as hard as it may be for some of us), have come to realize that to err is human and that the beauty in such human failure is the inherent ability to learn, progress, and eventually find the answers we have been looking for.

To get to such answers is hard work and they typically don’t come inspired by convenience.

Science would do well to remember that.

Image generated with Copilot Designer

The AI Act Needs an Updated Cyber Resilience Act

In this text, I argue that the AI Act needs us to update the Cyber Resilience Act. The proposal to regulate the cybersecurity requirements of products with digital elements, known as the Cyber Resilience Act, bolsters cybersecurity rules to ensure more secure hardware and software products – we must update it. Unlike GDPR, AI is machines, not words or people. And the Internet of Things (IoT) is the hands and feet of AI. Finally, at the least, devices must have a minimum level of security before they go to market. While this is an important first step, even more is necessary. We need an agency for infrastructure and hardware.

The use of AI in the EU will be regulated by the AI Act, the world’s first comprehensive AI law. (Link)

A first-ever EU-wide legislation of its kind, it introduces mandatory cybersecurity requirements for products with digital elements, throughout their entire lifecycle. (Link)

AI is Like Electricity

AI is not the goal to improve any practice, it is a means to overlay the world like electricity. After its first important step, IoT has now come to its logical presence as the hands and feet of AI.

In that sense, AI is nothing new. It is a continuation of the machine learning operating in the databases of objects connected through internet or intranet protocols. It is a scale that determines formative effects. In the early 2000s, the cloud enabled IoT. Before that, in the 80s and 90s, IoT projects were demos only because the storage capacity was simply too expensive. In the early 2020s machine learning evolved into such powerful algorithms to handle the enormous data sets producing ChatGPT and its successors.

We should have had the discussion that we are having right now on AI around 2000 when we sowed the seeds for our current situation. The digital transition as we know it today is the story of remote sensing. For centuries, objects (i.e., things treated and developed by humans –  were silent about their material conditions. Natural things rumbled sometimes, changed condition, shape, texture, and odor, and communicated as if they were about their state. Even if this communication was not directed at humans, humans could learn from it and devise ways of handling this changing scenery.

It was not until the Industrial Revolution that objects were being produced in the co-composite form of materials capable of resonating, that these objects became capable of stating facts about their condition. This information is then used to predict the behavior of these objects. At this stage, the three main issues that arise with this capability are: 1) Is the data coming from the object correct, 2) What (or who) is the intermediary that translates what is radiating from the device into information, and 3) What (or again who) is interpreting the data? These questions stem from the early beginning of predictive maintenance in the mid-1950s. They are still the key questions today.

Predictive Maintenance Kickstarted the Digital Revolution

Predictive maintenance, predicting behavior from machines in factories, kickstarted the digital revolution as we know it. It created a mindset that the conditions of objects could be monitored, and analyzed and provide feedback and suggestions for further use. This mindset found two main computing domains and trends: pervasive and ubiquitous computing. Both are still alive today but as an integral part of IoT, a term coined by Kevin Ashton, an RFID product manager at Proctor & Gamble. Pervasive e-computing clustered around IBM, and ubiquitous computing centered around Xerox Parc.

Ubiquitous means everywhere. Pervasive means "diffused throughout every part of." In computing terms, those seem like somewhat similar concepts. Ubiquitous computing would be everywhere, and pervasive computing would be in all parts of your life. (Source: Google)

The seminal text that catches both meanings is Mark Weiser's The Computer for the 21st Century, with its much-quoted beginning: The most profound technologies are those that disappear. They weave themselves into the fabric of everyday life until they are indistinguishable from it. (Link)

This is the key to what IoT is and it is quite unbelievable for a forty-year-old vision to have come true as it means connecting every – now most – objects on the planet being digitally addressable, even if that means in most cases current passively, being able to be read by an RFID reader, intelligent cameras, or drones. The drive towards full connectivity was there, right from the beginning. It was most articulate voiced by a European research program called The Disappearing Computer:

The Disappearing Computer (DC) is an EU-funded proactive initiative of the Future and Emerging Technologies (FET) activity of the Information Society Technologies (IST) research program (2001-2003) The mission of the initiative is to see how information technology can be diffused into everyday objects and settings and to see how this can lead to new ways of supporting and enhancing people's lives that go above and beyond what is possible with the computer today. Specifically, the initiative focuses on three interlinked objectives:

• Create information artifacts based on new software and hardware architectures that are integrated into everyday objects.

• Look at how collections of artifacts can act together, to produce new behavior and new functionality.

• Investigate the new approaches for designing collections of artifacts in everyday settings, and how to ensure that people's experience in these new environments is coherent and engaging. (Link)

The AI Debate

It is clear that these questions relate to what is now called the AI debate, yet they became tied to the applications and services of IoT and its reference architectures. The basic, and main, question of how we want to live together with an intelligent environment was never really part of a general discussion, yet had it been properly addressed we would have been able to embed AI into a clear framework of real operational value – the actual devices – instead of regulating it.

I was at one of the kick-off sessions of the project and in Jonschoping 2000 I first encountered the vision:

Electricity was the actual metaphor that the EU's first project officer, Jakub Wechjert, used. He spoke of a vision of the future as one in which our everyday world of objects and places becomes ‘infused’ and ‘augmented’ with information processing. Computing, information processing, and computers disappear into the background and take on a role more similar to that of electricity today - an invisible, pervasive medium distributed in or real world. In contrast, what will appear to people are new artifacts and augmented places that support and enhance activities in natural, simple, and intuitive ways.

That, however, does not make it more unproblematic. What we encounter in such an environment is the problematic and futile attempts to claim any which one – subject/ time/ space/ place – as an undisputed starting point for making meaning or sense, for deciding on how to act, for recalling how previous procedures operated, for projecting a sense of self into the future. In a mediated environment, it is no longer clear what is being mediated, and what mediates. Such environments - your kitchen, living room, our shopping malls, cobbled streets in old villages, are new beginnings as they reformulate our sense of ourselves in places in spaces in time. As new beginnings, they begin new media.

I was dozing off in the big conference hall, thinking about all these new beginnings, this longing for new space to occupy as if it was the wild, Wild West. What worried me most were some rather satisfied minds. I too could visualize a setting in which people resonate with media through simulating processes. Simulating processes that are actual processes, for in a digitized real, any process might become experiential, might resonate. Then a speaker, I believe it was Streitz, came on stage. He spoke of a Bluetooth ring that whenever I walked in the woods could – if I so liked – enhance this walk for me (I wondered who needs to enhance a wood?) by activating a mechanism that would either reveal a screen near the tree or send information on a handheld computer. And on that screen, I could read some more about that tree.

I was wide awake and I felt very strange. I looked around me, searching for any human presence in that lecture room; to wink at me, and tell me it was all a big sick joke. I recalled my sword and King Arthur and my talking trees. No screens there. That was when I realized. I asked myself could some of what these people are talking about actually be dangerous? The best thing I can do is stay close to them, track what they are interested in, and either hack it or try to confuse the spaces in which they operate. (Link)

From my tone, you can gather that I was still in shock. Here was a vision of a world presented in which computing would disappear into the known environment as if computing were a set of democratic processes which it was and is not. It was not even a naive decision, it was completely taken for granted, even if the audience was not only filled with engineers but with designers and ethnographers as well. The very notion of AI was already entangled in that vision and program. At the heart of decision making which up until that point had been a purely human toolset, however, it took skills that were potentially within the reach of any person (arguably a problematic statement given the distance between public and elite).

It was logical that a general audience could not see something as big as pervasive computing, ubiquitous computing, or IoT. They saw bits and pieces, fragments of new connectivity like RFID gradually entering the store and clothes and smart thermostats like NEST, or smart doorbells and cameras in the home.

The Kaczynski Impact

But maybe one of the reasons why there never was a solid debate about what it meant to overlay the world entirely with a digital layer addressable and addressed to commercial and government was that that position was made extremely suspect by Ted Kaczynski, the Unabomber. A core issue in his thinking is his distinction between small-scale technology is technology “that can be used by small-scale communities without outside assistance” and “organization-dependent technology is technology that depends on large-scale social organization.”

According to him, there are no significant cases of regression in small-scale technology, “but organization-dependent technology does regress when the social organization on which it depends breaks down.” His two major assumptions though are very true, the first being that “if the use of a new item of technology is initially optional, it does not necessarily remain optional, because the new technology tends to change society in such a way that it becomes difficult or impossible for an individual to function without using that technology.”

The second is that he foresaw that the system may break down and if it does break down “the consequences will still be very painful. But the bigger the system grows the more disastrous the results of its breakdown will be. So if it is to break down it had best break down sooner rather than later.” If it breaks down “there may be a period of chaos, a “time of troubles.” It would be impossible to predict what would emerge from such a time of troubles, but at any rate, the human race would be given a new chance.

He foresaw that whoever lives in the interface of technical applications will only be able to innovate within the limitations of that same interface. Progress will be made using the guidelines of the interface, building new applications with the same rules. From this, it follows that whoever is first, is the winner that takes it all. This not only shows on the application and product level, see Apple, Google, and Microsoft, but on the very notion of commercial applications and services taking over critical services from the state on connectivity, energy, and communications. This model, this interface, has become so logical and normal that it is hard to even consider deviating from it.

Yet it is not hard to see that his devastating methods of bringing his message were responsible that with the death and destruction, his analysis did not take root. This is a pity as it has led to – on the one hand, a state-controlled system where innovation is subject to strategic state planning in China – and on the other hand, a US amalgam of for-profit companies with diverse and widely-ranging interests. What has become viewed as nonviable is a mix of both – commons-led innovation, including infrastructure supported by commercial initiatives. Picture the Apple model as a commons. The mobile phone is in public hands (it sets hardware and software rules), the App Store is an EU App Store closing which apps run, and new open source applications for search, social media, and real-time communication can be picked from the work of Next Generation Internet hundreds of projects. How logical can it be?

We Need to Update the Cyber Resilience Act

At the moment, there are about 17 Directives, Regulations, and Acts that aim to reclaim European digital sovereignty. What if you regulate hardware instantiations like the mobile phone, you would immediately gain digital autonomy and democratic control over its operating system, its capacity to host the EU Wallet safely and securely, and its app ecosystem. We think that updating the Cyber Resilience Act to describe devices, routers, and mobile phones is a good start.

Editor: Learn more about the EU’s Cyber Resilience Act