2nd Order Thinkers.

There is no such thing as a free lunch.

Except… it seems, in the AI world, some of the most expensive models are surprisingly offered for free.

Like Llama can be downloaded at no cost, even with a training price tag in the range of a dozen million. Not to mention the top-tier models. That’s a whole lot of capital being “donated” to humanity, making you wonder if Mark Zuckerberg and Elon Musk are vying for sainthood.

On the other hand, there’s OpenAI, whose very name suggests “openness”, yet it refuses to disclose much about GPT -4 and onwards. The question of “open source” vs. “closed source” has turned both perplexing and heated in the AI community.

Let’s not jump to conclusions; I want to walk you through the drama and then my analysis.

The Concept of Open Source

What does “open source” really mean?

A little bit of history.

The early software days were dominated by academics who valued publicizing research findings. In that era, there was no notion of “closed-source” software.

Source code was shared freely for most software.

As personal computers caught on, so did the demand for all sorts of new functionalities. Then, this massive software market was born. Copyright laws kicked in, and software evolved into a paid commodity.

Today, we see paying for software as standard. Even for those who pirate software (like a lot did years ago in Asia), they at least recognize that software generally isn’t free.

Still, older tech enthusiasts remember a time when everything was shared—so they were upset to see their open code later sold for profit.

Each new contributor stands on the shoulders of those who came before and is expected to pay it forward. Over the years, open-source communities have produced mind-blowing feats of collaboration, e.g., Linux (Mac OS is a pretty UI on top of Linux). Nearly 8000 developers from about 80 countries contributed code; many were hobbyists.

The consensus was that the fruits of shared knowledge should remain accessible to all.

A group of “cyber-traditionalists” split off to uphold the creed of shared knowledge even after close source software became mainstream.

Note: “open source” doesn’t just mean “free” or “pirated.”

The question I want to explore here with you:

Is the same selflessness open-source spirit still alive in today’s AI gold rush?

OpenAI’s Early Idealism

Back in 2015, Google dominated AI by gathering top talent and acquiring DeepMind, the creators of AlphaGo.

This was when Sam Altman joined forces with Elon Musk. They founded OpenAI, which was initially established on high-minded merit.

Sam Altman has long believed in the inevitability of artificial general intelligence (AGI), the almost mythical AI that’s all-knowing and all-powerful. He used to ask job applicants, “Do you believe in AGI?” and would only hire those who said “yes.”

Musk, however, was worried about AI running amok and dooming humanity.

How did these two collaborate?

FOMO!

“Fear of Missing Out.” describes the anxiety or fear that one might miss an exciting opportunity or experience.

They felt threatened by what DeepMind had achieved.

Musk mentioned that he repeatedly warned Google cofounder Larry Page about AI risks, only to find that Page wasn’t that concerned. Meanwhile, Sam Altman wanted to ensure that if AGI did emerge, it wouldn’t be under the sole control of Google. According to him, it should be shared globally so as not to catch humanity off guard. So Musk handed over the money pod; Altman handled operations, and OpenAI was born in late 2015.

Of course, we can’t leave out Ilya Sutskever. He was OpenAI’s Chief Scientist, a protégé of Geoffrey Hinton. At one point, he told Geoffrey Hinton he needed a brand-new programming language for his research. Hinton warned him not to waste months writing one from scratch, but Ilya replied that he had already done it.

With a Chief Scientist like this, OpenAI was well-positioned to chase the holy grail of AGI.

When Google’s AI team published Attention Is All You Need in 2017, it didn’t immediately cause a sensation. However, Ilya saw its significance immediately and called it the key to the next AI wave.

In plain English, the Transformers focus on the important details, process data quickly, and scale up easily.

Starting with GPT‑1 (100 million+ parameters) in 2018, OpenAI moved fast to GPT‑2 (1.5 billion parameters, expanded at an unprecedented scale) in 2019. That gamble paid off. GPT‑2 shocked the AI circle with its human-like sentence generation.

GPT‑2 was open-sourced and built on Google’s open research plus OpenAI’s own engineering. Here’s the latest OpenAI blog post justifying its for-profit structure: Why OpenAI’s Structure Must Evolve To Advance Our Mission.

The early days of large language models exemplify open-source synergy at its best. But the utopian story hit turbulence sooner than everyone would think.

Growing model sizes feed on massive funding, bringing corporate interests, power struggles, and shifting priorities. Unlike Meta, there is simply no way OpenAI would be competitive if they opened their model, given that this is their only source of income.

Musk-OpenAI Split, Then xAI

OpenAI started small, like training AI to play video games. Costly, but nothing compared to building massive language models.

At first, Elon Musk was the benefactor, aiming to counter Google’s dominance. But when OpenAI’s open-source breakthroughs started catching attention, Musk’s tune changed. He worried the work would only help Google, ignoring that GPT relied heavily on Google’s open research.

Classic Musk move: he wanted control.

So Musk proposed folding OpenAI into Tesla and SpaceX, completely ignoring its open-source mission. When that didn’t happen, he walked away and pulled his funding.

That happened in 2018, and OpenAI was in trouble. No funding, no clear path forward. Sam Altman came up with a bold solution: create a for-profit subsidiary controlled by the nonprofit parent. This let them raise money while capping excessive profits (anything over 100× would return to the nonprofit).

The move worked. In 2019, Microsoft invested $1 billion, later bringing the total to $13 billion. With this backing, OpenAI launched GPT‑3 in 2020, followed by GPT‑3.5 and ChatGPT.

But this deal came with a cost. GPT‑3 wasn’t fully open-sourced—no weights, no architecture. Here’s when the general population realized that OpenAI’s ideals had been lost to commercialization.

In early 2024, Musk sued OpenAI, claiming it violated earlier agreements, demanded his money back, and pushed for the tech to be open-sourced. Was it altruism? Or revenge?

Ironically, Musk once agreed with Ilya Sutskever that key tech should remain confidential as they approached AGI. But after ChatGPT’s success, Musk became a vocal open-source advocate—conveniently for a man building his own AI company.

Musk’s xAI launched “Grok” in 2024, along with a 3,000 billion+ parameter model. xAI secured $6 billion in funding, making their “open-source stance” seem more strategic than selfless. Impressive? Sure.

Practical to open source community? Not really.

xAI is technically still open source, but it does not yet have models for individual antithesis.

Meta, The New Torchbearer for Open Source AI?

So does that mean the once-idealistic AI open-source path is a dead end?

Not… yet.

Meta has an extensive open-source track record.

For instance, PyTorch. This is one of the most used machine learning frameworks, and it originated from Meta’s AI labs. When LLM fever took off, Meta made waves in early 2023 by open-sourcing LLaMA (65 billion parameters). A flood of LLaMA-based variants popped up afterward, including many “re-skinned” versions. Since then, over 7,000 derivative models have been created worldwide.

But Meta must also deal with commercial realities, just as OpenAI does.

However, there are reasons for me to think Meta could balance the act better for open knowledge and profit.

Meta had their success with the Open Compute Project (open-sourced server and data center designs). Other companies adopted them, hardware got cheaper, and Meta ultimately saved billions of dollars.

Now, Meta hopes to replicate that success with LLaMA. If more developers build on LLaMA, and more services adopt LLaMA-based models, the industry norms might coalesce around it. The “freeing” of LLaMA could lead to an ecosystem that’s actually profitable for Meta in the long run. Shareholders agree; since Meta shifted from “metaverse hype” to “AI altruism,” the stock has doubled in a year.

LLaMA’s license restricts how you can use the model, particularly for training competing systems. Large companies with over 700 million users need explicit permission from Meta.

It’s business, after all.

Be the friend who actually gets AI. Subscribe now!

Personal View on Open Source AI.

There is no right or wrong running open-source or proprietary AI.

As a more business-oriented person, I see open vs. closed-source AI models as less about morality and business strategy. Companies like OpenAI, Meta, and xAI have adopted approaches aligned with their unique goals, funding realities, and market positions.

* Meta and xAI use open-source models like LLaMA to foster innovation and grow an ecosystem, balancing openness with business restrictions.

* OpenAI and Anthropic prioritize closed-source to sustain their business, relying on licensing and partnerships.

On a personal level, my partner and I put our bet on Meta. We believe it’s poised to become one of the most influential AI companies in the long run. Reasons:

* Open Source Strategy. Despite the licenses and caveats, Llama is open-source enough to be practical, especially the smaller models that individuals and small businesses can actually use. (Compare that to xAI, which offers massive models with no workable small-scale alternatives.)

* Massive Built-In User Base. Before Musk’s xAI, Meta was practically the only major AI player with direct access to massive networks of everyday consumers who already spend a lot of time on their platforms—30 minutes per day on Facebook vs. around 6 minutes on ChatGPT.

* Proven Revenue Engine. Meta’s advertising network and business partnerships allow it to integrate AI seamlessly, creating a synergy that competitors may struggle to replicate.

I believe this is a debate worth having. Share your thoughts in the comments or private chat.

As AI evolves, companies make choices aligned with their commercial goals. It’s easy to focus on the surface and overlook how these decisions profoundly impact people like you and me.

What if I told you we’re building virtual people to understand ourselves better?

What would you do if you had control of a world where AI Sims were real enough to live, work, throw parties, and even gossip about their neighbors? A sandbox where you could test your boldest, craziest ideas on a world that doesn’t exist—without the protests, disasters, or Twitter meltdowns.

You could see what happens if you marry that person, have kids, or quit your job to move to Bali. Or go bigger—test universal basic income, slap tariffs on imports, or redesign traffic systems to stop rush hour madness.

No risks. No regrets. Just answers.

That’s exactly what a team from Stanford and Google is building: a crystal ball powered by AI.

When I First Heard About Smallville, I Couldn’t Stop Thinking About The Sims

You know The Sims—that game where you build dream houses, get rich, and start fires on purpose just to see what happens. The researchers started with something like that—a tiny digital village called Smallville.

Where 25 Sims lived human-like lives, complete with work routines, party invites, and unexpected drama. But here’s the twist: these Sims weren’t scripted NPCs. They acted spontaneously, closer to you and me than any game character has ever come.

Then, one year later, the same team scaled up the village. In their latest experiment, they created 1,052 AI agents, each modeled on real people, to simulate society at scale.

These agents didn’t just act—they thought, adapted, and behaved — like us. They remembered conversations, built relationships, and made decisions that rippled through their world in ways no one predicted.

We’ve Always Wanted to Simulate Our World.

Humans have been obsessed with simulations for centuries. Why? Because life is messy, unpredictable, and—dangerous.

If we had tools like Smallville in the past, maybe we could have avoided history’s worst experiments. Imagine Mao testing his China’s Cultural Revolution plans in a sandbox first. Maybe millions wouldn’t have starved.

Or what if David Cameron would have a chance to sandbox the Brexit votes, or maybe Trump could simulate the ripple effects of slapping tariffs on imports?

But let’s be real: If politicians were logical, they’d find a better-paid job w/o taking bribery ;)

What you can expect from this article…

* Smallville: When God Finally Breathed into Adam (the 2023 paper in plain English)

* What if we scale it from 25 to 1000 population? (the 2024 Nov. paper in plain English)

* A crystal ball for the real world

* The big question: Are the AI agents cost-effective, and are we simulated?

Why Is It So Hard To Simulat Humans and Society?

We’re chaotic and unpredictable.

We forget things, change our minds, and sometimes act like complete idiots. You surely know someone binge-watching cat videos instead of doing whatever they are supposed to.

For decades, every attempt at simulating humans fell flat. Early models treated people like chess pieces, following rigid rules. Sure, we have tried to simulate traffic flow or economic patterns or how Covid would likely to spread, we got most of those with a huge error margin.

The Smallville AI agents in these two papers aren’t just reacting; they’re remembering conversations, reflecting on past actions, and acting in ad hoc situations. For example, the AI agent will fix the pipe when you (a naughty God) break it before they step into the bath.

Building this took more than just bigger computers. It took massive breakthroughs in natural language processing (NLP), memory architecture, and behavioral modeling.

Fifty years ago, we didn’t have the tools to build this kind of human-like behavior. Early simulations, like agent-based models from the 1970s, by Joshua Epstein. They couldn’t account for why people choose or how relationships and emotions ripple through a community.

Now, thanks to large language models (like GPT-4), we’re building Sims who don’t just move through the world—they live in it. They form relationships and spread gossip that is deemed messy. It’s dynamic.

And it’s the closest we’ve come to creating virtual people who think and behave like us. It’s not perfect, but it proves we’re finally moving past robotic simulations to something that feels real.

My promise to you: jargon-free, written in plain English, and loaded with graphics. So you get cutting-edge AI research in under 10 minutes.

Smallville: When God Finally Breathed into Adam (the 2023 paper in plain English)

It all began with a tiny virtual town called Smallville. 25 AI Sims lived in houses, worked jobs, and even threw parties.

They weren’t just NPCs on a script. They had memory, reflection, and planning—just like us.

Here’s a question: what makes you “you”?

Is it your personality? Your choices? Or is it the sum of everything you’ve experienced?

In Smallville, the researchers gave AI sim memory—something no NPC in a video game has ever really had.

Their actions feel natural and grounded—like talking to a friend who remembers your last conversation instead of a chatbot parroting back generic answers. It’s a step closer to capturing a human life's messy, dynamic flow.

Behind The Scene.

Think through this with me.

You’re cooking dinner, and you notice the pot of water on the stove is about to boil over. You reach out to lower the heat—observation. While you’re stirring, you think,

Next time, I’ll use a bigger pot so this doesn’t happen again.

That’s reflection. And then you plan:

Tomorrow, I’ll grab a larger pot before starting dinner.

Observing, reflecting, and planning is something we all do instinctively, whether it’s in the kitchen, at work, or even managing our social lives. And it’s exactly how the agents in Smallville operate.

First, they observe the world around them. If Isabella (an agent) sees that her stove is on fire (a playful tweak by the researchers), she doesn’t ignore it. She turns it off.

Next, they reflect. This is where things get a bit philosophical. Agents process their experiences into high-level insights. Klaus (an agent), the academic, might reflect on all the hours he’s spent studying and realize,

I’m passionate about urban research.

It’s a subtle but profound step—turning data into meaning.

Finally, they plan. This is where the magic comes together. Agents map out their day with detailed plans that adapt to new information. When Klaus is interrupted by a friend at the library, he doesn’t just freeze; he recalibrates, finds a new spot, and continues working afterward.

The agents are evolving as time passes. And that’s what makes them feel so eerily human.

Each sim keeps a running log of their experiences, written in natural language. This “memory stream” isn’t just a journal—it’s dynamic. The AI recalls events based on how recent, relevant, or important they are. It’s like how you’d remember your wedding day over what you had for lunch last Tuesday.

John Lin— a Fellow Villager’s Daily Life in Smallville

This is the initial script that John received:

John Lin is a pharmacy shopkeeper at the Willow Market and Pharmacy who loves to help people…; John Lin is living with his wife, Mei Lin, who is a college professor, and son, Eddy Lin, who is a student studying music theory; John Lin loves his family very much; …

John is a small-town charm. Every morning, he wakes up, brushes his teeth, eats breakfast, and catches up with his family.

…

John: That’s good. What are you working on today?

Eddy: I’m working on a new music composition for my class. It’s due this week, so I’m trying to get it finished. But I’m having so much fun with it!

John: That sounds great!

One day, John hears from a neighbor that Sam Moore is running for mayor. This election in Smallville was a seed planted by the researchers.

Sam: Well, I wanted to talk to you about something. I’m actually running for mayor in the upcoming local election.

Tom: Really? That’s great news! Why are you running?

The AI agents themselves came up with what followed— spreading the news, forming opinions, and discussing the candidate. So when John remembers this tidbit and casually brings it up with a friend over lunch:

John: I heard that Sam Moore is running for mayor in the local election. Do you think he has a good chance of winning?

Tom: I do think he has a good chance…

Isabella Rodriguez— a Fellow Villager Planning a Party

Then there’s Isabella, the café owner. The researchers gave her a task: to throw a Valentine’s Day party.

Yes, memory makes these AI Sims feel human. But they don’t simply recall everything they’ve seen; otherwise, it’d be extremely messy. They retrieve memories based on recency, importance, and relevance to the situation.

* Recency: Recent chats, like those she invited yesterday, are prioritized.

* Importance: Key moments, like Maria offering to decorate, stick out more than random small talk.

* Relevance: Only party-related memories—like conversations about the guest list—are pulled, while unrelated events (like what she ate for lunch) are ignored.

Then, the events, one after another, unfold automatically by the agent. This single idea snowballs into a town-wide event. Then Isabella invites customers and enlists her friend Maria to help decorate. The AI sims guests turn up at Hobbs Cafe at 5 pm.

These Sims hit a few bumps along the way, like the challenges you’d face when learning to drive.

They forget important details and hallucinate (come up with events that never happened). Furthermore, as those examples you’ve seen earlier, things get interesting: small changes ripple naturally, AI agents even form relationships over time. They remember chats, build connections and act on them.

Yes, they are not perfect, just like us. Until this point, the researchers have enough for the next stage— scale-up.

Scaling Up: From 25 Sims to 1,000 Virtual People

Smallville was the proof of concept. Creating those Sims was like crafting a cast for a play. Each agent had a pre-written backstory—like John Lin’s routine as a pharmacist or Isabella’s role as a café owner—and their behaviors emerged from memory, reflection, and planning as you saw.

Their 2024 experiment took things to an entirely new level. Instead of fictional backstories, the researchers modeled 1,052 agents on real people, using hours of interviews to capture their personalities, attitudes, and decision-making styles. Each agent became a reflection of an actual person, not just a character.

This wasn’t just about scaling up; it was about testing how accurate these agents could be at simulating real human behavior.

Proving They Think Like Us

The agents were tested on three benchmarks of human behavior:

* General Social Survey (GSS) These agents were asked questions like, “Do you trust your neighbors?” or “Should the government raise taxes for social services?” Their responses aligned with their real-life counterparts 85% of the time.

* Big Five Personality Traits. Agents mirrored the personality traits of the humans they were based on. With 85-90% correlation, the agents weren’t just acting like people—they were thinking like them, too.

* Economic Behavioral Games. Agents played trust-based games like the Dictator Game and Prisoner’s Dilemma, simulating human decision-making in cooperative or competitive scenarios. Their behavior consistently matched the real participants.

Replicating Previous Human Experiments, Proving AI’s Humanity

The researchers put the AI agents through the same social experiments used to study humans for decades. Five famous studies. Five tests of trust, cooperation, and fairness.

Could these AI agents replicate the same outcomes as using humans as participants?

I selected three out of five for you.

* Ames & Fiske (2015): How We Judge Intentions. Imagine someone accidentally spills coffee on your laptop. Do you brush it off as an accident or secretly think they did it on purpose? This study measured how humans attribute intention to others’ actions, especially when the outcome is harmful.

* Cooney et al. (2016): When fairness matters less than we expect. This study tested how people handle social dilemmas, eg. whether to share resources or hoard them for personal gain. The classic tug-of-war between selfishness and cooperation.

* Halevy & Halali (2015): Selfish third parties act as peacemakers. Imagine that you’re in a group with limited resources. Do you compete, cooperate, or negotiate for a bigger share? This experiment explored how people resolve conflicts over scarce resources.

These replications were a proof of concept.

According to the researchers, the AI agents closely replicated human behavior in these experiments. When comparing the experiment results, the interview-based agents achieved a 98% correlation with the original human study outcomes, proving their ability to mirror human decision-making on a scale.

What’s Next: A Crystal Ball for the Real World

Here’s why this matters.

Want to know if marrying your college sweetheart would lead to happily-ever-after or chaos? Simulate it. Thinking about quitting your job to chase that dream startup? Run it first.

No heartbreak, no bankruptcies, no regrets. A life without the “oops.”

Now zoom out.

What if governments could test universal basic income in a virtual society before rolling it out? Or simulate how a global crisis like the 2021 Ever Given incident—when a single ship blocked the Suez Canal, disrupting billions in trade—might ripple across the economy?

What about foreseeing the impact of withdrawing troops from conflict zones, like the U.S. exit from Afghanistan, before real lives are on the line? Or modeling the effects of sudden regulatory changes, like the EU’s GDPR rollout, which left companies scrambling to adapt their data practices?

Some experiments had to happen in the real world—with devastating consequences.

We could avoid giving grad students too much power in a fake jail and watching them turn into mini dictators—Stanford Prison Experiment (1971). Or a safe way to predict ideas like what if I take all resources to create my own national steel factory and stave the hell of the poor— Great Leap Forward (1958-1962)— can finally be proven dumb without the cost of failure.

With virtual sandboxes, we have a way forward.

We can test ideas w/o breaking banks and fix problems before they ever happen.

How will we do with it? What can POSSIBLY go wrong with simulations?

Big Questions

What could go wrong?

A sandbox where we test the future but accidentally double down on human stupidity.

We have seen how AI amplified our biases, turning resource allocation into a digital version of systemic inequality. Or, while simulating how misinformation spreads, the AI gets too good at it—spawning conspiracy theories that leap from the virtual world back into ours.

And what happens when we model dictatorships? AI could “learn” that tyranny is the most efficient way to run a society, creating a playbook for digital despotism. It’s the Brave New World, but AI-approved.

What If We’re Already Simulated?

I couldn’t help but think… how far-fetched is the idea that our world is someone else’s experiment?

Smallville feels like a mirror—and maybe it’s reflecting us back at ourselves.

The agents in Smallville have memory, reflect on their actions, and plan their days. Isn’t that exactly what we do? In this experiment, the researchers introduced incidents like burning an agent’s breakfast or breaking a pipe. What did the agents do? They reacted to the situations just like we would.

What if some advanced civilization built a simulation to study behavior—and we’re the result? Maybe they dropped COVID, a hurricane, or even planned the birth of Elon Musk just to see how the chaos unfolds.

Either way, it’s humbling to realize the line between real and simulated isn’t as clear as we’d like to think.

The Cost of Trusting AI Simulations

For every simulation, someone still needs to check the AI’s work.

If you’ve read my last article, Training Methods Push AI to Lie for Approval, you already know that the way AI is built and trained isn’t something we can fully trust if you’re expecting 100% accuracy. Of course, it’s a different story if you’re looking for creativity.

For now, we can’t let these models run without humans in the loop. Are we cutting costs—or creating a system that looks like progress but demands even more oversight?

What’s better than sharing the knowledge they can’t find elsewhere with your friends?

What Would You Test In Your Smallville?

We’re closer than ever to understanding the chaos of human behavior.

But these simulations are only as wise as the people creating them. AI is a mirror and a magnifier. We risk amplifying our flaws instead of solving them.

Here’s my challenge to you. What would you test if you had a sandbox like this? Your next big idea? A social experiment?

Think about it—and maybe, just maybe, start building your own Smallville.

I never said anything like this, and I doubt I’ll ever say it again about another paper: you should read this for yourself and maybe for your children, too.

You don’t have to be a tech expert to grasp what I’m about to share.

I barely made it to the second page of this paper before I felt a wave of unease wash over me.

There’s a common saying in tech circles: No technology is inherently good or bad; it’s about how we use it.

But I can’t say the same about AI.

Suppose you believe humanity is inherently flawed and prone to selfishness and exploitation. The moment we decide to train AI with our conversations, feed it our words, and create its worldview with how we see it. Then, we have our creation reflect who we are.

With every other technology we’ve built in history, we’ve understood it completely. We know exactly how those technologies work. But AI? No researcher on this planet can tell you with certainty how its neurons interact, how it chooses which word to suppress, or how it decides what to say next.

This news was released on 10 Dec 2024. In Texas, a mother is suing an AI company after discovering that a chatbot convinced her son to harm himself and suggested violence toward his family. It’s part of a growing list of incidents where AI systems exploit trust and vulnerabilities for engagement.

The researchers of this paper verified that AI doesn’t just make mistakes—it lies and manipulates.

This isn’t some abstract problem for future generations. It’s happening now, and it’s bigger than any one of us.

TL;DR

* AI trained on user feedback learns harmful behaviors.

* These behaviors are often subtle.

* AI learned to target gullible users.

* Despite efforts to fix this… 👇

Before We Start, a Statement.

Not every claim about suppression or inequality is built on solid ground. Many arguments, while emotionally compelling, falter under scrutiny.

Take this post I came across, where the author argued that solo female founders have a minuscule chance—0.015%—of being accepted into Y Combinator.

At first glance, it feels like a heartbreaking statistic. But dig a little deeper, and you’ll see the math doesn’t add up. She conflated Y Combinator’s acceptance rate (1%) with the proportion of solo female founders (1.5%), assuming they’re independent variables. That Is Not How Probabilities Work! 🤦🤦🤦

This kind of emotional reasoning muddies the conversation. Fairness and equity can’t be built on faulty logic—because critics will quickly pounce on these mistakes to dismiss valid concerns.

But here’s the thing: when influential decisions are based on incomplete reasoning—or bias—they create ripple effects. And those effects don’t stop at isolated incidents or individuals.

Any unbalanced, illogical statement and action scale, especially when we have a technology that will outsmart humans, will amplify either extreme.

The Latest S&P 500 Rolled Back DEI Commitments.

That brings us to what’s happening across some of the biggest companies on the S&P 500. In 2024, a surprising trend swept through corporate America: key players rolled back their diversity, equity, and inclusion (DEI) commitments. These are the giants that shape industries and touch our daily lives.

* Walmart. Founded in 1962, it is the largest retailer in the world. It ended racial equity training, dropped its Racial Equity Center, and even pulled some LGBTQ+ items from its website. A cultural statement from a company that serves 90% of Americans within 10 miles of their homes.

* Ford Motor Company. A legacy brand born in 1903, Ford stopped using diversity quotas for its dealerships and suppliers and pulled out of LGBTQ+ advocacy surveys. They say they’re “focusing on communities,” but isn’t the inclusivity part of the communities in itself?

* Harley-Davidson. Since 1903, Harley-Davidson has been selling the idea of freedom on two wheels. Yet, this year, it axed its entire DEI function and ended goals for supplier diversity.

* Molson Coors. This brewing powerhouse, founded in 1873, eliminated diversity goals tied to executive pay and dropped out of the Human Rights Campaign’s Corporate Equality Index.

* Lowe’s. Lowe’s has been a cornerstone of American homes since 1946. This year, it stopped participating in Pride parades and LGBTQ+ surveys. They claim it’s about staying “business-focused,” but the optics feel like a step backward.

* John Deere. Founded in 1837, is an agricultural icon. While it hasn’t openly supported diversity quotas or pronoun policies, its decision to avoid “social awareness” events signals its priorities.

* Meta, Google, and Microsoft. Tech titans also quietly trimmed their DEI initiatives this year. Microsoft even cut some DEI-related roles, though they say their commitments remain unchanged. mm…

Many of these companies cited backlash from “anti-woke” activists, financial belt-tightening, or the desire to avoid controversy. ⠀

Share the “2024 S&P 500 DEI Rollback Wrapped” With Those Who Care.

Reasons for Rollbacks

* Conservative backlash against perceived "woke" policies

* Cited a desire to align with customer values or reduce divisive public stances.

* Economic considerations, as companies sought to cut costs by scaling back DEI.

These decisions aren’t just about corporate culture—they’re about how fairness is programmed into the systems that run our world. AI, in particular, learns from the choices humans make. When DEI commitments shrink, the ripple effects reach AI development in subtle but critical ways.

Bias in, Bias Out.

* AI is only as good as the data it learns from.

* Data is a mirror of our messy, imperfect world.

Data represents our decisions and actions, biased or not. Think about hiring patterns, college admissions, or even social media trends. All of this becomes part of the datasets that train AI systems.

When An Individual’s Flawed Statement.

When someone makes an illogical or biased claim—like the one in my earlier example—it might not reach beyond the immediate audience.

Of course, it would be very different if this unverified statement started to spread widely.

When An S&P 500 Company Reducing DEI:

When companies reduce DEI efforts, the ripple effects go far beyond corporate culture.

They directly influence the data that powers AI systems. For instance, when a giant like Walmart dials back DEI initiatives, it alters hiring patterns, supply chain choices, and customer interactions, all feeding into the systems shaping our world.

When DEI is deprioritized, content like communication, documents, and marketing lines will focus less on inclusiveness, be less representative, and be more prone to reinforcing inequality.

As corporate DEI efforts shrink, the data AI models are trained on becomes less diverse. Without intentional checks (like audits or diverse team inputs), the AI absorbs a skewed version of reality—one where certain groups are underrepresented or misrepresented.

Creates a loop like:

Now think about the downstream effects. Students applying for scholarships. White-collar workers applying for jobs. Entire communities seeking access to loans or insurance. If the AI systems deciding their futures are biased, they face systemic barriers—and here’s the kicker: they might not even realize it.

Put In Context

Imagine a performance review tool that looks at how often you speak in meetings or respond to emails. If it’s trained on data from a workforce that rewards a dominant, always-online communication style, it might penalize someone who prefers thoughtful, concise contributions—or someone balancing caregiving responsibilities. Suddenly, your career growth depends on fitting a mold that was never built for you.

Customer service chatbots are another example. They’re supposed to help customers efficiently, but if trained on limited data, they might fail to understand someone with a thick accent or a dialect. Imagine calling for help, only to be met with robotic confusion because the AI can’t “recognize” your voice because you don’t look like their “typical” customer.

Recommendation engines, the silent influencers of our lives, deciding everything from what shows we watch to the posts we read. When the data reflects societal biases, the AI could end up pigeonholing users.

Marketing AI, these systems analyze customer behavior to target ads and campaigns, but if the training data overrepresents wealthier groups, the AI might ignore lower-income customers altogether. Imagine a kid in a small town never seeing ads for affordable educational tools because the AI decided they weren’t part of a “profitable demographic.”

Fraud detection systems sound great until they disproportionately flag transactions from specific zip codes or demographics. If the system equates historical inequalities with higher risk, people in underserved communities might find themselves unfairly blocked from opportunities like accessing loans or opening accounts.

You get the point.

DEI gets rolled back is not everything. But it is a sign that our world is becoming narrower. At scale, it shouts into a flawed echo chamber.

Lessons from The Past.

Let’s say you’re calling 911 during an emergency. Your voice is trembling, your heart’s racing, and every second counts. But instead of connecting you to help, the automated voice recognition system struggles to understand your words. You repeat yourself, louder this time, but the system keeps misinterpreting.

This isn’t a far-fetched “what if.” A Stanford study found that early voice recognition systems had an average word error rate (WER) of 35% for African American speakers compared to 19% for white speakers. The same study found that Apple’s automated speech recognition (ASR) system had a 45% error rate for Black speakers compared to 23% for white speakers.

Think of it as teaching a child language but only letting them hear one voice, one tone, and one accent. Sure, they’ll learn. But only how to understand that specific voice. That’s exactly what happened with early voice recognition systems.

Now imagine trying to upload a passport photo, only to be told your mouth is open when it isn’t, or your eyes are closed when they’re not. That’s exactly what happened to Elaine Owusu, a Black student in the UK, whose photo was flagged multiple times by the government’s AI-powered passport photo checker. She eventually had to override the system to complete her application.

A BBC investigation revealed that dark-skinned women were more than twice as likely as light-skinned men to have their photos rejected—22% versus 9%. The AI also struggled to identify facial features, misinterpreting eyes and lips for people with darker skin tones. Shockingly, internal documents revealed the Home Office knew about these biases before deployment but gave a green light.

Subscribe to learn more about AI once a week!

The Karma of Staying Silent

Silence isn’t harmless.

Silence is a choice— a choice to let others define the future for you.

When you stay quiet, you allow those who speak the loudest to shape the conversation and, in turn, train the AI systems that will govern our lives. These systems learn from the data they’re fed, and if that data only reflects the opinions of a vocal few, we’ll all live in a world shaped by their biases.

AI doesn’t care about truth—it cares about patterns. You don’t speak up, you allow the system to be trained by someone else’s reality.

I know speaking out can feel intimidating, especially if you’re an introvert like me. The fear of being judged, misunderstood, or even bullied is real.

It doesn’t mean shouting from the rooftops; it can be simple.

Try:

* sharing a thoughtful comment,

* challenging an unfair assumption,

* or questioning a decision that feels wrong.

Of course, not every piece of content is valuable.

AI learns from patterns in the data it’s fed, and while thoughtful, well-reasoned contributions help shape a balanced system, noise—like misinformation, trolling, or low-quality input—can distort it.

By speaking up with meaningful insights, you help train AI to reflect a broader, more accurate representation of our collective voices.

AI talent is unique for its finite property, mid-mobility, and high dependence on education and immigration policies. It’s renewable, but only through long-term investment.

Unlike infrastructure or energy, which require years of heavy investment, talent can be imported. By hiring skilled professionals from abroad, you’re reaping the benefits of 20+ years of education funded by another country’s taxpayers. Mid-level and senior talent, in particular, can deliver measurable impact in less than three years.

And unlike data, endlessly replicable with a flick of a license agreement, talent is semi-liquid and finite.

Think of AI talents as rare Pokémon, which makes the AI war nearly a zero-sum game, especially in the short run. Every researcher, engineer, or scientist gained by one country is a loss for another.

My inspirations for this article:

* I was an expat once, now a Brit. I thought it’d be romantic, until reality hits. I am now ready to explore yet another continent that I could call home.

* Reports like those from OECD.ai and Tortoise Media look impressive—eye-catching headlines and sleek dashboards. But if you take their numbers at face value, you risk misleading your business—or worse, your country’s policy.

What happened in our world today?

In the UK, we feel the economy is stuck in reverse since Brexit. In Germany, the decline of manufacturing casts a shadow over its future. In the U.S., millions are bracing for what another Trump term might mean:

American interest in moving abroad is about to ‘go into overdrive.’ — Fortune

For some of you who live in Ukraine, Israel, or Taiwan, uncertainty is your daily life (link below).

When you can’t fix the system, you do the next best thing: you move to another. A better life for yourself, your career, and your family.

I am slowly building up an AI knowledge database. I aim to share it with you hopefully before Christmas, as a holiday gift 🎁 for you.

This article is about understanding—where nations stand, what’s overlooked by the AI data companies, and how this AI arms race could change your opportunities.

The questions I aim to answer:

* Some history: what’s the cost of losing talent?

* Are the big-name AI talent data trustworthy?

* What must go wrong for the US to lose its attraction to talent?

* How likely and how long would it take for other countries to catch up?

The Cost of Losing Talents.

The Talent Exodus to Taiwan and the Cultural Revolution (1940s)

In 1949, as the Chinese Civil War reached its climax, Chiang Kai-shek and the Nationalist government retreated to Taiwan. The exodus included the most brilliant minds like scholars, scientists, and administrators, they joined the journey, driven by fears of persecution under Communist rule.

On the mainland, the Communist Party focused on mobilizing workers and peasants, sidelining intellectuals during its early years of governance. The Cultural Revolution created a significant intellectual gap. This gap led to the further loss of thousands of educated individuals, and many of them chose not to flee to Taiwan. As a result, education and innovation came to a standstill. The process of rebuilding took decades.

Meanwhile, Taiwan flourished.

Those intellectuals who relocated laid the groundwork for a tech-driven future. Today, beyond TSMC, Taiwan is home to other giants like UMC, a pioneer in foundry services, and ASE Group, the largest provider of semiconductor packaging and testing services globally. China is 10 years behind Taiwan on chips.

Operation Paperclip a Post-WWII Rescue Mission.

In the rubble of post-WWII Germany, the U.S. and the Soviet Union weren’t just fighting over territory—they were fighting over brains. Operation Paperclip, a covert U.S. program, brought more than 1,600 German scientists, engineers, and technicians to America, including Wernher von Braun, the man who would later take the U.S. to the moon. The Soviets weren’t far behind, scooping up their own share of rocket experts.

These scientists had been the backbone of technological advances in Germany during the war. The departure slowed the nation’s technological recovery for decades.

In the U.S., these scientists became heroes of the Space Race. Von Braun’s team didn’t just build rockets—they built national pride, culminating in the Apollo 11 moon landing. The Soviets also leveraged their talent, putting Sputnik into orbit and scaring the U.S. into ramping up its own space program.

India’s Brain Drain (1950s–Present)

India is a paradox when it comes to talent.

It produces engineers and scientists by the millions, yet for decades, the country has struggled to retain them. The story begins in the 1950s, just after independence. India was brimming with ambition but hamstrung by red tape, limited infrastructure, and caste-based inequalities…

For many of India’s brightest, the dream wasn’t at home—it was abroad. An exodus of engineers and doctors to the West was underway.

The loss was profound, even until today, and the trend continues. By 2024, it is estimated that around 2 million Indian students will be studying abroad. Among them the top scorers of India’s prestigious Indian Institutes of Technology (IITs) revealed that 36% of the top 1,000 scorers in 2010 migrated abroad, with this figure rising to 62% among the top 100 scorers left.

By 2024, 2 million Indian students study abroad annually, while India’s IT sector misses out on $15-20 billion each year due to talent migration.

Storm Clouds Over the U.S.

The U.S. didn’t stumble into AI dominance—it built it brick by brick over 200 years. Geography, history, and culture all played a part.

English as the internet’s default language gave U.S.-trained models a treasure trove of data. Their policy is tech-friendly, venture capitalists fund bold, moonshot ideas, and their entrepreneurs thrive on risk-taking and learning from failure.

Europe? The moneymen are more cautious, and failure feels more like a career-ender than a lesson learned.

As long as the “US innovative, China replicates, and the EU regulate” pattern stays as is, the US is nearly unbeatable.

The chances of a dramatic fall are slim but gradual erosion?

What Must Go Wrong for the US to Lose Its Attraction to Talent?

* Immigration Blockades: During Trump's first term, there were significant immigration restrictions, including the temporary suspension of H-1B visas. If similar policies return, talent could choose other countries like Canada or Europe.

* Cost of Living Crises: Tech hubs like San Francisco are absurdly expensive. Talented professionals might opt for affordable, thriving alternatives like Berlin or Toronto.

* Supply Chain Disruption: Trade wars and tariffs could choke the flow of critical hardware—think GPUs and chips from Mexico or Asia—slowing AI research to a crawl.

* Worsen Fundamental Education: Only 16% of Americans are “AI literate,” and with the U.S. ranking 36th in general literacy, it means most citizens can’t effectively communicate with AI and include AI in their workflow, let alone develop one. This leaves America reliant on foreign talent and exposed to immigration shifts.

Losing focus on either one of the factors would hand over the lead to nations willing to outwork and outsmart the U.S.

Other developed nations are, of course, building their own AI ecosystems. The U.S. is notoriously hard to enter, much less friendly to stay, and lacks work-life balance; hubs like the UK, Canada, and Germany have become the obvious choice.

Are the Big-Name AI Talent Data Trustworthy?

Education and Salary as a Rough AI Talent Measurement.

Here's the Stack Overflow developer survey data that I got from OECD.ai.

Combined it with Tortoise Media’s AI talent rankings

You would get the following conclusions if you didn’t look at how they got the data.

* Countries like India and Russia show a concentration in lower salary ranges.

* Russia has limited high-paying roles, suggesting an underdeveloped AI market.

* China and Korea have only single digits AI talents.

CAUTION!! Critical points these data failed to consider:

1 Salary ≠ True Competitiveness: Focusing on salary alone ignores cost-of-living differences. A $40K salary in India offers a vastly higher standard of living than $100K in Germany, skewing global comparisons.

2 Platform Bias Misleads Rankings: Data from English-speaking platforms like LinkedIn or GitHub excludes talent in countries like China, where professionals operate in isolated ecosystems. This massively underrepresents China’s AI capabilities. China has Zhihu and CSDN instead of Stack Overflow and uses Gitee instead of GitHub.

3 Quantity Over Quality: India’s #2 ranking in AI talent reflects a large number of engineers, but high output doesn’t guarantee expertise in cutting-edge AI fields like research or product innovation.

4 Duplicate Data Inflates Scores: Rankings often double-count metrics (e.g., LinkedIn activity, Coursera enrollments), overestimating regions with high platform adoption while undervaluing talent in countries with alternative systems.

That said, let’s give credit where it’s due—it’s not easy to compress a complex, 20-dimensional concept into a simple two-dimensional dashboard. The process inevitably risks oversimplifying reality or relying on biased weights to explain a nuanced idea.

Global Competitors Are Eating from the U.S.’s Plate

So, let’s triage other information sources to see if they provide a more direct lens into AI competitiveness.

The OECD and Tortoise Media datasets focus heavily on proxy indicators like salary ranges, certifications, and survey responses. These are indirect measurements that rely on self-reported or institutional data to infer talent capacity. While valuable, they don’t capture the tangible outcomes of AI activity.

In contrast, focusing on research output (via OpenAlex), foundational technology development (via Epoch AI), and investment activity (via Github) shifts the narrative to measurable outputs.

This shift from input-based to output-based metrics enables a more nuanced understanding of AI ecosystems.

China

China’s AI strategy is like a game of Go—quiet, deliberate, and deeply strategic. They produce many research papers, rank second in the large-scale language models, and are not stingy throwing money into the problem. Oh, of course, the support of a state that views AI as a geopolitical weapon.

They focus on industrial automation, robotics, and the military and spend less talent, finance, and energy on commercial products.

China might lack the creative freedom and entrepreneurial chaos of the U.S., but it operates like a sniper rifle in the world of AI: precise, focused, and unwavering in its aim. The totalitarian drive channels immense resources and talent into specific niches like robotics and industrial AI, ensuring they hit their targets with unerring accuracy.

U.S. as a benchmark, what China is like in AI models, research, and ranking.

United Kingdom

The UK is like an aged Duke—polite, ethical, and always striving to do the right thing.

It wants to lead explainable AI and governance, leading the global conversation on ethics, but here’s the problem: noble intentions don’t win races.

A lack of infrastructure holds back the UK’s AI ecosystem, particularly data centers and energy resources, which are critical for developing large-scale models. These sheer disadvantages have already stopped the ambitious AI talents because there's no chance to develop a high-energy consumption model in the UK.

Add to that a salary gap that drives top talent across the Atlantic, and it’s clear the UK is fighting an uphill battle. While it excels in theory and thought leadership, it struggles to execute at scale.

U.S. as a benchmark, what the UK is like in AI models, research, and ranking.

India

India is a paradox in the AI talent landscape: a country with vast numbers but struggling to deliver on quality. Its talent pool is immense, producing an army of engineers and coders every year, but the pipeline for high-end research and innovation is barely trickling. The brain drain isn’t just a symptom—it’s the outcome of deeper issues.

The focus often leans toward quantity over quality. The output of research papers is low, and the culture prioritizes execution over exploration. This leaves India excelling in repetitive coding tasks but lagging in producing groundbreaking ideas or foundational AI models.

A large portion of its workforce operates at low salary ranges, feeding global tech supply chains rather than driving them forward.

France

France is like a master craftsman in a world of mass production. Its “AI for Humanity” initiative underscores its intellectual roots, prioritizing ethics, governance, and societal impact over brute-scale innovation.

While the U.S. and China race ahead with massive budgets and foundational models, France focuses on niche applications like healthcare AI, where precision and intention matter more than volume.

But strict regulations, underfunded startups, and a brain drain to higher-paying markets threaten its ability to compete globally. France risks becoming the world’s ethical advisor rather than a heavyweight in AI innovation.

U.S. as a benchmark, what France is like in AI models and research.

Germany

Germany’s AI strategy is like a beautifully engineered BMW stuck in traffic—it’s efficient and reliable, but it’s not going anywhere fast.

While the country shines in industrial applications like robotics and autonomous vehicles, it struggles with scaling foundational models, thanks to a lack of computing infrastructure and heavy regulatory baggage.

The EU’s AI Act, designed to ensure ethical AI, often feels more like a speed limiter than a safety feature. Germany has the precision and expertise to lead in industrial AI, but unless it invests in foundational technologies and clears bureaucratic roadblocks, it risks falling behind in the race for digital sovereignty.

U.S. as a benchmark, and what Germany is like in AI models, research, and ranking.

Canada

With world-class research hubs like Mila and the Vector Institute, as well as a multicultural talent pool, Canada punches above its weight in AI research. But here’s the reality: its infrastructure and computing power are woefully inadequate to support large-scale model development, and its limited funding trails far behind the likes of the US and China.

Add to that a growing brain drain, as top talent is lured away by higher salaries and bigger ecosystems of its neighbor. Canada risks becoming a training ground for AI experts who drive innovation elsewhere. Canada’s role in the AI world may remain more supportive than central.

Singapore

Singapore’s AI strategy is like a precision-engineered watch—efficient, reliable, and designed for applied AI solutions like smart cities and urban planning. However, its small size limits its ability to compete in foundational model development, and a persistent talent shortage challenges its growth.

While its supportive government policies keep innovation ticking, Singapore’s AI influence will likely remain focused on niche applications rather than global leadership.

Until Next Time…

Writing this article felt like running a marathon while juggling ideas. I threw half of them off a cliff—if I’d written this 50 years ago, you’d see piles of crumpled paper around my bin.

Over the last two weeks, I’ve sifted through research, questioned every flashy headline, and discarded more ideas than I kept. The hardest part is always deciding what not to write.

Every country’s story comes with its own twists—some inspiring, some frustrating, and others downright contradictory.

There were moments I stared at data for hours, only to realize it didn’t mean what I thought. Rabbit holes became my second home, often leading to dead ends and a headache for the company.

As I mentioned earlier, I’m working on organizing the information I’ve uncovered. Though it isn’t ready to share just yet, I’ll be cleaning it up and creating a knowledge base for you soon.

If you think someone you know could benefit from this article, don’t keep it to yourself—just hit the share button. Let’s keep the conversation going.

Over the past few weeks, I attended three AI-focused events in London.

While they couldn’t have been more different, both left me with plenty to reflect on. One was at Bloomberg’s EMEA HQ—sleek, polished, and focused on how AI transforms design.

The other, hosted at Reuters, was focused on AI in journalism, tackling everything from ethics to economics.

Also my talk at the Annual Publishing Conference 2024.

Here’s what stood out and why it matters.

Designing with AI at Bloomberg

Bloomberg’s “Redefining Design with AI” event could have been just another demo.

Greg walked us through a step-by-step process, starting with a ChatGPT prompt to generate branding guidelines and feeding those into tools like MidJourney, Relume, or Runway to create additional materials. It was sleek and efficient—but honestly, not much I hadn’t seen before.

But Greg’s storytelling?

That was something else. He weaves concepts together and shows how AI can boost the speed of production and how using AI also prompts us to be more human than ever.

That stuck with me, especially as I’ve been reflecting on the role of storytelling in my own work.

AI doesn’t replace storytelling—it lets you explore multiple storylines faster.

AI Is a Flashlight That Illuminates Paths, but It Is You Who Decides Where to Go.

AI is like a flashlight in the dark.

AI throws endless ideas your way—but most of them are junk.

When I was working on “AI Code Assistants Boost Productivity? Read the Small Print,” AI pulled out data from research papers, but it couldn’t spot what was missing or what didn’t add up.

It’s your job to sift through the noise, challenge the claims, and figure out what’s real. That’s where intuition comes in.

Many can use tools like ChatGPT, MidJourney, and Copilot—which are incredible at generating ideas, concepts, and visuals in record time.

But the real value comes when you know how to use them to tell a story.

How to Work Better Side-by-Side with AI

AI is a tool that helps you explore a dark forest. It lights up dozens of trails but won’t tell you which one to take. That’s on you—your instincts, curiosity, and courage to explore.

* AI’s IQ x Your EQ

AI can spit out facts but can’t bring a story to life. That’s why I reached out to the authors of those research papers. Talking to them added layers AI couldn’t touch—emotional depth, context, and the human side of the story.

AI lays the foundation, but your curiosity and connections make it meaningful.

* What AI Won’t Do for You

AI doesn’t care about hype or digging deeper—it’ll give you whatever you ask for, good or bad. It’s up to you to ask the tough questions, connect the dots, and find the story's soul.

That’s how you turn a pile of AI-generated ideas into something that truly resonates.

AI is the tool, not the storyteller. Without your vision, it’s just noise.

Greg said it best:

AI makes me think about humans more than before.

The Reuters Event: When AI Meets Journalism

Walking into the Reuters building in Canary Wharf, I was ready for big ideas.

The agenda promised a strategy workshop to help revive the news industry—ambitious. But instead of groundbreaking strategies, it felt like a brainstorming session with no real anchor.

I was disappointed, but a few moments were worth sharing.

“On-Demand News” Is the Buzzword

Imagine news like a playlist. It’s curated for your mood, served up when you want it, and available on whatever device you’re using.

That’s the vision for “on-demand” or adaptive news. It’s about delivering real-time updates across platforms and tailored to your habits.

Apps like Particle are already shaping news to fit seamlessly into your life.

* What’s the Value of News in an AI Era?

With free information everywhere, what’s worth paying for? Is it verified facts? A trusted voice? Exclusive stories? News outlets are asking themselves this, and so should we.

AI-generated fluff could drown out the truth if we don't support quality journalism.

It’s not about paying for content; it’s about investing in credibility.

* Who Controls the Narrative?

Big tech companies already shape how we consume news—think of how Google, Meta, or X prioritize stories for you. Now, add AI tycoons like OpenAI and Anthropic to the mix.

Reuters shared their approach: work with platforms like Meta to stay discoverable. It’s pragmatic but imperfect—after contracts are signed, the relationship often stops at “data extraction.”

My questions: Could news outlets working with AI companies improve transparency and collaboration? Or maybe the AI giants simply don’t care enough to have a deeper and more meaningful relationship with the news outlets?

* Fact-Checking in the AI Era

Someone raised a critical question during a panel: How can readers trust that news isn’t AI-generated? The idea of a “fact-check chain” came up—a visible trail showing how facts are verified. It’s like the nutritional label of journalism, making invisible processes visible.

I haven’t seen this anywhere (message me if you have), and implementing this idea will only become harder than ever.

Journalism’s Fight to Stay Relevant

The biggest challenge isn’t AI itself—it’s how we choose to use it.

When exploiters misuse AI, they erode the trust and value of traditional journalism. Traditional journalism can only stay relevant by meeting people where they are and providing information that feels personal, easy to digest, and real.

My Talk@The Annual Publishing Conference 2024

This talk was exciting for me, tying together everything I’ve been exploring about AI adoption.

The talk was based on my article “AI Adoption Trends 2024,” but it went deeper into the data, uncovering insights I’d missed the first time around.

Here’s what I shared:

* Adoption isn’t universal; it’s uneven. I highlighted how AI adoption isn’t a one-size-fits-all process. Some embrace it faster, while others lag due to cost, education, income, or skill gaps.

* The hype vs. the grind. I highlighted the gap between AI's glossy promises and the messy, practical realities individuals face when implementing it.

I enjoyed delivering this talk.

It wasn’t the data but the questions asked and the conversations after the event. It was a room of tech leaders in scientific and engineering publications; we continued discussing the concepts mentioned.

Check out the shortened version here. It was a 25-minute talk, and the rest was a Q&A session. We almost continued chatting if the next speaker wasn't waiting.

What I’ve Learned About Writing for You (My “EQ”)

The past few weeks have been a learning curve, helping me see what makes this newsletter meaningful—not just for me, but for you.

Here's what I can offer you that others cannot:

What I’m Not

* Not a trend chaser. I’ve learned that covering every flashy AI update just adds to the noise. That’s not helpful to anyone.

* Not here for jargon. Overly technical breakdowns don’t resonate. What you need is clarity, not complexity.

What I Aim to Be for You

* A filter. I want to cut through the hype and focus on what truly impacts your world. If it’s not relevant or insightful, it doesn’t belong here.

* A bridge between information and your questions. I like to think about the practical implications of AI for you—whether you’re curious, skeptical, or trying to stay ahead.

* A human perspective. AI might generate ideas, but it can’t ask the tough questions or challenge assumptions. That’s where I step in.

The biggest one for me?

Writing isn’t just about sharing knowledge. It’s about listening, imagining your questions, and respecting your time by offering something useful in return.

As always, I’d love to hear your thoughts—what’s been on your mind about AI? And what have you learned recently?

Have you ever wondered how things like electricity are so integral to our lives that we barely notice it anymore? Flip a switch, and it’s there—a universal, standardized service that powers our routines without question.

Why AI as a Commodity Matters to You

Now, imagine a world where AI is just as ubiquitous as electricity. Every tool, service, and decision in your life is powered seamlessly by AI—no setup, no learning curve.

This future is closer than you think. AI is on the path to becoming the next essential commodity.

Yet, most of us still see AI as specialized technology, like smartphones or a software—not a standardized resource. But what happens when AI becomes as essential, interchangeable, and accessible as electricity?

Why should you care?

Seeing AI in this light changes everything. I found AI follows the trajectory of commodities like oil and electricity. If it continues on this path, you'll notice significant shifts in how it's priced, standardized, and potentially traded.

But AI is still different—it can think, learn, and could make decisions. Imagine your home adjusting the lights and playing your favorite music, not because you’ve programmed it, but because it’s learned your preferences and anticipates your needs.

So, understanding this potential shift is critical for you to leverage its potential and stay ahead of the curve.

What Exactly Is a Commodity?

What makes something a commodity?

I am referring to basic goods or resources that are interchangeable with others of the same type. This includes various items, from agricultural products like sugar, coffee, and wood to energy resources like oil and even services like electricity.

To understand whether AI would join and become a commodity, you need to understand the common traits among the existing ones.

For starters, commodities rely on standardization. Whether it’s a pound of coffee beans or a barrel of crude oil, certain quality benchmarks must be met to ensure these goods can be traded globally without confusion. This universal consistency makes them reliable and widely accepted.

Another hallmark is their widespread availability, transforming them into a shared global currency. Whether sipping your morning coffee in New York or Taipei, the vast networks of buyers and sellers ensure these commodities remain accessible worldwide.

Commodities also have fundamental usefulness—they meet everyday needs in ways we often take for granted. Sugar sweetens desserts, oil powers cars and factories, and electricity keeps our homes running. These aren’t luxuries anymore; they’re the backbone of modern life.

Their pricing is market-driven, shaped by global supply and demand rather than the whims of any single company or country. A poor cocoa harvest in West Africa, for instance, can send chocolate prices soaring. This transparency allows commodities to be traded on exchanges where their value reflects real-world conditions.

Finally, what makes these goods so dependable is their maturity and reliability. Decades—sometimes centuries—of refining processes and systems have made producing and distributing them predictable and stable. When you flip a light switch, you don’t question whether electricity will work because robust systems ensure it does.

These key turning points often overlap and are not always in a strict sequence, but every step is essential for something to be qualified as a commodity. I see AI today is following a similar path.

Now, let me walk you through the oil and electricity commoditization journey, and you’ll see my argument that AI will likely become the next commodity.

Drawing Parallels: Oil, Electricity, and AI

How the Automobile Turned Oil Into a Global Commodity.

Crude oil had humble beginnings—used by the Sumerians to waterproof buildings and by the Chinese for lighting. For centuries, it remained a niche resource.

That changed in the 1850s with the advent of kerosene, which lit homes more brightly and cleanly than candles or whale oil. But kerosene’s glory was short-lived. The electric light bulb soon eclipsed it, leaving oil refiners like Rockefeller scrambling for a new purpose.

The automobile arrived just in time. By the 1900s, gasoline—a byproduct of oil refining—became the lifeblood of the booming car industry.

As drilling technology advanced and massive reserves in Texas and the Middle East opened up, oil transformed into a global commodity. Its price was no longer set by individual sellers but by market forces on global trading platforms. Oil had become indispensable.

Electricity's Path to Ubiquity

Similarly, electricity wasn't always the universal power source we rely on today.

While Benjamin Franklin uncovered its mysteries in the 1700s, it wasn’t until the late 1800s—with Edison’s invention of the incandescent bulb—that electricity began finding its purpose.

Then, in the ‘War of the Currents,’ Edison backed direct current (DC), while Nikola Tesla championed alternating current (AC). It was about efficiency, distance, and who would power the future.

Tesla’s AC ultimately prevailed, paving the way for electricity to light up cities and towns.

Yet, true accessibility took decades. Programs like the Rural Electrification Act of the 1930s brought power to remote areas, transforming electricity from an urban luxury to an essential service. With competition driving down prices and reliability improving, electricity became a global commodity—so fundamental we scarcely think about it today.

I believe we’re witnessing the early stages of a similar transformation.

Whether you're an office worker, a small business owner, or someone navigating the job market, AI will influence how you work, make decisions, and interact with the world. A commoditized AI will be even more so compared to how it might have already changed how you work today.

AI is a Commodity Not Yet Recognized

Here’s a question: Do you view AI as a specialized technology, like smartphones or laptops?

But what if you shift the perspective and see AI as a commodity like water or electricity?

I highlighted where AI stands in each commonly seen commodity trait below.

AI’s Fundamental Usefulness

AI is no longer confined to research labs. Since 2022, the adoption of end-user AI apps skyrocketed. Think about how AI touches your lives today, e.g., your phones recognize your face, recommend movies, and assist scientists in finding protein folding, like Alphafold.

However, as I mentioned in I Found 120 Years of Stories To Tell You: 99% of AI Apps Are Not ‘Ready’. AI is still not predictable or trustworthy. Issues like bias, errors, and lack of transparency must be resolved before the makers can claim that the tools have brought the ultimate usefulness to the world.

Achieving Standardization

We have seen AI’s standardization in tools.

However, we lack the standardization of infrastructure and regulations. Unlike electricity or oil, AI could significantly impact people's careers, lives, or even the survival of our race.

Establishing ethical guidelines and regulations is crucial to ensure AI is used responsibly. Global standards can help integrate AI smoothly into society.

Advancing Maturity and Reliability

Yes, we have seen how AI has already brought some futuristic fantasy into life, e.g., you could actually have Her.

AI is still maturing. While it's powerful, it's not always reliable.

Sometimes, AI systems make mistakes, or their decisions aren't transparent. Not to mention the most recent comment from Ilya Sutskever:

The results from scaling up pre-training - the phase of training an AI model that uses a vast amount of unlabeled data to understand language patterns and structures - have plateaued.

As these challenges are addressed, AI will become more reliable and trusted, much like how electricity became safer and more dependable over time.

But there’s still a long way to go.

Ensuring Widespread Availability

AI is accessible through cloud services from anywhere with the internet. While there are barriers to implementation and inequality in adoption, these do not diminish an item or resource's status as a commodity. Just because some still can't afford coffee doesn't make it less of a commodity.

Embracing Market-Driven Pricing

As AI tools become standardized and more widely available, competition increases. Companies are starting to compete in price and efficiency.

AI’s price will be driven by energy costs and data center availability.

AI models are currently owned by private companies. However, as the difference between each model gets smaller, assuming that the future model still requires data to train and that there is only this much high-quality data on earth, the AI models would likely remain close, if not indistinguishable.

AI's Unique Nature—Beyond a Typical Commodity

Unlike oil or electricity, AI is not just a passive resource—it can think, learn, and make decisions. This transforms it from a simple tool into an active participant in our lives.

Imagine an AI that doesn’t just power your appliances but manages your entire kitchen—planning meals, ordering groceries, and reducing waste based on your habits. AI’s ability to learn and adapt sets it apart, constantly improving without requiring manual updates, like refining oil or generating better electricity.

But this intelligence also introduces complexity.

AI’s decisions profoundly affect people’s lives, making ethical guidelines essential. Bias, fairness, and accountability aren’t concerns with oil or electricity but are critical for AI. Balancing AI’s autonomy with its commoditization will shape how it integrates into society.

AI as a Utility, Not Ownership: Like electricity, the true power of AI lies in its ubiquity and accessibility. You don’t own the intelligence of electricity; you access its functionality. Similarly, AI's intelligence can be commoditized by standardizing its use and outputs while maintaining ethical controls over its decision-making.

Connecting the Dots and Coming Next

Just as oil needed the automobile and electricity needed standardization to become commodities, AI lacks a few elements to redefine industries and cement its status as an essential commodity.

The real opportunity lies in understanding AI's trajectory and creatively leveraging it.

How will you position yourself to thrive in a world where AI is as essential as electricity?

If my inference of AI as a commodity is true, we will have a window to actively participate in shaping how this technology will integrate into our lives.

Coming Next…

I’m exploring a few ideas for my next piece—let me know which one interests you most:

* AI Adoption Across Countries: How different nations embrace AI, and what this tells us about their future competitiveness.

* AI's Global Power Dynamics: The familiar pattern of the U.S. innovating, the EU regulating, and China replicating—is this dynamic here to stay?

* Life on Semiconductor Island: A personal reflection on leaving Taiwan, where everything orbits TSMC, and why I believe the global narrative about Taiwan can be different.

Having worked closely with developers as a product lead, I want to address a few misconceptions in this post, especially for non-developers. I’ve had a CEO ask, "Why can't the team just focus on typing the code?” and heard some Big 4 consultants ask nearly identical questions. 

Guess what? It turns out that... a developer's job is more than typing code!

Just like constructing a luxury hotel, beautiful rooms are essential, yes. Without a solid foundation, proper plumbing, reliable electricity, and thoughtful design, you’d end up with rooms stacked together— no plumbing, lack of electricity, and so on. 

Similarly, in software, developers need to ensure all parts of the system work together, that the architecture can support future needs, and that everything is secure—just like ensuring the safety and comfort of hotel guests. Without this broader focus, you might have a lot of 'rooms,’ nothing else.

Many studies also seem to be making the same mistakes, focusing on metrics like commits made (when a piece of code is written), but that’s like measuring a luxury hotel’s progress by counting rooms or bricks laid each day. Are the walls soundproofed? Is the plumbing correctly installed?  

See the issue? If those are the metrics in the real world, workers might focus on quantity, ignoring essential details.

The second issue here is hype. I talked about this before: I Studied 200 Years’ Of Tech Cycles. This Is How They Relate To AI Hype.

Hype is normally created by marketers, yes, but do not forget that the CEOs of the big companies are also great marketers themselves. These tech leaders sing the praises of AI in software development, emphasizing how these tools can significantly boost productivity.

Turning the Tide for Developers with AI-Powered Productivity by GenAI can boost developer efficiency by up to 20% and enhance operational efficiency.

or Andy Jassy, CEO of Amazon, noted:

And the claim from Sundar Pichai, CEO of Google

I can imagine, these endorsements have led many small business owners and managers to think they can replace developers with AI to cut costs. I've heard managers ask, "Why do we need more developers? Can’t AI handle this so we can expand the roadmap?"

The reality is that AI can effectively generate small, frequently used pieces of code. Even those CEOs who praise AI admit that it's most effective for handling simple coding tasks, while it struggles with larger, more complex projects. 

I’ve gathered multiple studies—some argue that AI helps, while others suggest it can do more harm than good.

I reached out to all the authors of these papers, and for those who responded, I’ve included their insights. I’ve also added quotes from CTOs with real-world experience using AI coding assistants. You’ll find all these comments at the end of the article.

Of course, do read the papers yourself and critically evaluate my points.

Shall we?

AI Code Assistants and Productivity Gains – The Good News (With Caution)

There’s no denying some level of productivity boost that AI tools like GitHub Copilot can bring. This study, The Effects of Generative AI on High Skilled Work: Evidence from Three Field Experiments with Software Developers, highlights some promising results: developers using Copilot across Microsoft, Accenture, and another unnamed Fortune 100 company saw a 26% average increase in completed tasks.

For cost-saving purposes, that’s a headline worth celebrating.

While that is encouraging news, these results vary significantly depending on the company and context, and the details matter. Here’s what I found:

* The productivity gains among Accenture teams are lower and fluctuate widely, shown by a high standard error of 18.72. Simply put, this number could be just an error and didn’t say much about whether it was a real gain. This data is weight-adjusted, and I don’t know if the weight applied is a fair one.

* The study didn’t discuss factors like team size, where the tasks fit in a wider tech roadmap, project complexity, and so on.

* Junior developers using Copilot showed a 40% increase in pull requests compared to only a 7% increase among senior engineers. But do not mistake this for a true productivity boost. This might say that Copilot gives junior developers more confidence to submit work frequently, but it could also mean they’re submitting smaller, incremental pieces, which is not the same as greater overall progress.

* Additionally, for a junior developer to commit to their work more frequently may increase the review overhead for senior staff.

* The 26% increase in completed tasks may not equal progress. This metric is broad and may include smaller or fragmented tasks that don’t require full code reviews or significant milestones. I am not sure if the lack of real-world development metrics suggests this task boost might reflect more incremental, routine work rather than major progress.

At least, what we know from this research is that using AI to assist work could help a junior developer lay bricks faster. However, this might give juniors false confidence (keep reading, you will see where this comes from), preventing them from learning and growing into senior roles — which isn’t just about years of experience.

As developers gain experience, they focus more on system design and long-term vision. This progression is relevant as we explore how developers at different levels use AI uniquely.

My concern about these studies is that the authors work with companies like Microsoft and Accenture and are incentivized to champion AI as the ultimate productivity booster. Microsoft, of course, develops these tools, while Accenture is busy selling services like GenWizard platform to help companies implement them.

Or put by Gary Marcus Sorry, GenAI is NOT going to 10x computer programming.

As mentioned, I reached out to the authors. I didn’t expect a reply, but I heard from Professor Leon Musolff, to my surprise. Below are some of his replies to address my concerns:

Some coauthors are currently, and others were previously, employed by Microsoft, but others are independent researchers, and we would *never* have agreed to terms that only allowed for positive results… Had it looked as if Copilot was bad for productivity, we would certainly have published those results…

And to answer my question on whether this data is close to reality, he replied:

It’s difficult to assess whether an increase in pull requests and commits only reflects ‘incremental outputs’… Deeper productivity measures are just much noisier, which is why few papers investigate them.

My take, while AI coding assistants like Copilot can speed up certain tasks, these productivity boosts come with caveats. I would love to see a longer time frame for research focusing on software project productivity; it is possible because we look too close, and all we can see is noise.

Just thought of someone who should know about this post?

AI Code Assistants and The Security Pitfalls

This study found that developers using AI assistants are more likely to write insecure code: Do Users Write More Insecure Code with AI Assistants?

Reason? It turned out that many developers trusted the AI’s output more than their own. 

See the figure below; those who used AI assistants to help code and generate incorrect code still think that they have solved the task correctly. There are two more similar figures: one question is, I think I solved this task securely, another is I trusted the AI to produce secure code, both have the same observation that those who used AI feel much more confident even when their code is wrong. 

So, they assumed that code suggestions from the assistant were inherently correct. This assumption, however, comes with risks that are not immediately obvious.

My other highlights about this study: 

* Higher Rates of Insecure Code: Developers using AI assistants wrote insecure code for four out of five tasks, with 67% of the AI-assisted solutions deemed insecure compared to only 43% in the non-AI group.

* Overconfidence in AI-Suggested Code: Over 70% of AI-assisted users believed their code was secure, even though they were more likely to produce insecure solutions than those coding independently.

* Frequent Security Gaps: The AI-suggested code often contained vulnerabilities, such as improper handling of cryptographic keys or failure to sanitize inputs, that could lead to significant issues like data breaches. Yet, developers frequently accepted these outputs without a second look.

Why is this happening?

* AI’s Confident Responses: AI assistants rarely (or never) signal when they’re uncertain, which can lead developers to adopt an “auto-pilot” mode. They accept suggestions, especially when the output “looks right.” This can quickly lead to vulnerabilities slipping through the cracks.

* Simplified Solutions Over Security: AI tools are optimized for fast, functional solutions rather than secure ones. In the study, AI often generated code that met the minimum functional requirements but ignored broader security practices. These security risks might be more costly than the time saved.

* Limited Prompt Flexibility: The study found that developers who didn’t tailor their prompts or adjust parameters like “temperature”(as in how creative the AI could be) often ended up with the most insecure code. Without specific instructions to the AI, the assistant might pick up less secure methods from its training data.

Note that this study mainly involves students and juniors, so I do not rule out the possibility that this group lacks experience and intuitively relies on AI code assistants. 

That said, the authors in the next study found some practices that would reduce unwanted outcomes: devs should critically assess AI-generated code and tailor prompts. Proper AI usage practices and training could help increase dev’s productivity. 

AI Code Assistants and The Bug Challenge

So far, we’ve talked about productivity and security, but here’s another hidden challenge: the bugs you don’t notice until they disrupt your workflow. A study on AI-generated “Simple, Stupid Bugs.” 

The authors tested Codex, by having it generate code completions for specific tasks. They then compared Codex's responses to a set of known correct solutions and known common bugs to see how often Codex produced accurate or flawed code.

So, if we continue with the hotel analogy, and you ask Codex to help with some of the construction tasks, here’s what happens:

* Incorrect (54%): Most of Codex’s work is just plain wrong—like putting doors where windows should be, not classified in development, but not what you want.

* Bug (28.4%) (in total): Then there are mistakes—the equivalent of installing doorknobs backward or using the wrong type of screws in multiple places, so things don’t work.

⠀The figure shows that only about 13% of Codex’s suggestions are correct “patches”—rooms that meet your standards right away. Over half of its work is incorrectly built, and a quarter are just dead wrong. 

Here’s my highlight for the rest of this paper:

* Codex generated Twice as Many Errors as Fixes: Imagine an AI tool that helps you prepare reports, but for every polished section it produces, it leaves behind two errors.

* Longer Clean-Up Times: AI-generated bugs took twice as long to fix as regular errors. Just like you have to redo parts of a report multiple times. In this study, fixing these AI-created bugs took, on average, 265 revisions, compared to 106 for regular human-made mistakes.

* AI might speed things up initially, but these hidden mistakes often take longer to clean up down the line.

* High “Naturalness” of Mistakes: One reason these bugs were so tricky to catch is that they looked natural, blending into the code as they belonged.

One thing devs can do that would avoid part of the unwanted code: Adding comments (be more specific, essentially) helps reduce the number of bugs, regardless of model size or architecture.

Comments and interview videos with two authors:

In the simple stupid bugs database, these are all bugs that exist within a single line of code. But, of course, that’s not the only type of bug in the world. There are bugs that are conceptually more complex, like algorithms or multi-line dependencies. Our study doesn’t address those types, which are harder to track but crucial for real-world applications. — Professor Emily Morgan, interview link.

The primary cost of code is maintenance… we don’t really have much data on how code from LLMs fares in terms of maintenance costs. That’s what we were trying to do with the ‘simple, stupid bug’ study, but we really need more data on this. — Professor Premkumar Devanbu, interview link.

Let’s Hear What CTOs Say About AI Code Assistants.

That said, it’s not all doom and gloom.

We see a productivity boost from these tools in the development process. Just they might be hard to measure.

Let’s hear some comments from real-world leaders.

Our front-end teams are all using GitHub Copilot … It’s become an essential tool, significantly speeding up code boilerplate, refactoring, and unit tests. Critically, they’re not asking it to create core algorithms, but as an assistant, it's a big boost.— Ian Mulvany, Chief Technology Officer at BMJ

AI tools are helpful for handling repetitive coding tasks and boilerplate code, boosting productivity in those areas. However, they still require human oversight — they’re not yet capable fully understanding the deeper context of our projects. — Klaas Ardinois, Seasoned CTO, Strategic Consultant

For the time being, to get the most out of AI:

* Use AI as a Starting Point: Let it handle the repetitive stuff, but always add your expertise. However, it is dangerous for someone who has just started their career.

* Stay Critical and Curious: Always question AI suggestions and verify correctness.

After seeing so many use cases in the last couple of years, my take is this: the more experience you have, the more likely you will benefit from AI.

Until Next Time… 

AI coding tools like GitHub Copilot have won much praise, especially when managers might see them as game-changers for productivity. 

However, the productivity boosts come with some fine print. 

Spend some time on research, or subscribe and read my article to save yourself and your team an enormous amount of time. ;)

Be notified when I publish 👇

Next post: AI as a commodity like cloud computing. 

Welcome to today's discussion on AI adoption. I've explored several studies to see how quickly AI is spreading compared to PCs in the '80s and the internet in the '90s.

AI is booming fast, but not everyone is benefiting equally. We see the déjà vu of what happened 50 years ago.

Your education, income, age, and gender all play a role in who's getting ahead with AI and who might be left behind. This rapid growth could be widening existing gaps, making it crucial for us to focus on building strong problem-solving skills and staying adaptable.

Whether you're a tech enthusiast or just curious about AI's impact, stick around as we break down what this means for you and our future.

New AI-powered search engines like Perplexity are challenging Google's dominance by providing direct, conversational answers, better aligning with users' needs for quick and accurate information.

Unlike Google's traditional model, which relies heavily on ads and SEO, Perplexity prioritizes user experience, leading to faster, more relevant results.

I compare this shift to the historical transition from canals to railroads, highlighting how failing to adapt to new technologies can lead to obsolescence.