Why AI’s Progress Might Destroy Its Own Future

Author: Matt Ferrell

You’ve seen the headlines. The viral demos. AI writing essays, generating photorealistic images, even creating entire videos. It feels like we’re living in a sci-fi future … or dystopian one. But while we’ve been mesmerized by chatbots and deepfakes, something much more important is happening behind the scenes. AI is solving problems we thought were decades away, from curing diseases to inventing new materials that could reshape our world.

And yet… there’s a catch. A paradox.

Because this same AI revolution that’s supposedly going to create incredible wealth for companies, might be laying the groundwork for an economic collapse that threatens the very progress it’s helping to create. Let’s dive into the most overlooked part of the AI revolution and the economic flaw that could unravel it.

The Hidden AI Revolution

This video is a little bit of a different one for me. Usually, I focus on the brighter side of tech advances that are impacting our lives. But over the past few weeks I’ve seen several YouTube videos that took the spark of an idea that was quietly flickering in the back of my mind and doused gasoline on it.  Many of these videos kept bringing up the incredible financial potential for companies that are building the hardware that’s powering the AI revolution, like NVIDIA with its GPUs or Tesla with its Optimus robots.  It always came down to something like company X is valued at a $Y market cap because of the incredible future potential of what they’re doing with AI or robotics.

But let’s start with what most people are missing in that conversation. When you hear “AI,” what comes to mind? Maybe the latest news about ChatGPT 4.1? Maybe MidJourney or Sora? I’ve been using tools like ChatGPT and Perplexity more and more to help with early research and parsing the news for YouTube videos, or for looking through large documents and research papers to find the exact sections I’m interested in.  I use it kind of like Google search on steroids.  Those tools are amazing (and sometimes horrifying), but they’re just the tip of the iceberg.

Large language models (LLMs) may be getting all the buzz, but AI tools go far beyond that. Take materials science, for example. Scientists used to spend years, sometimes decades, trying to find the right compounds for better batteries, solar panels, or superconductors.

But now? AI is accelerating that process exponentially. DeepMind recently predicted the properties of over 2.2 million new inorganic materials compared to the roughly 50,000 inorganic materials previously cataloged by the Materials Project.^{1 2} Microsoft’s MatterGen goes even further.³ It’s a generative AI trained specifically to discover new materials. What used to take entire research teams years now takes weeks or even days.³ The implications are staggering: better batteries, faster electronics, cleaner energy.

A recent collaboration between AI researchers and Microsoft have newly discovered materials to prototype batteries with up to 70% less lithium.⁴ It’s already influencing next-gen consumer tech.¹³

Then there’s medicine. AI is helping detect cancer earlier, design drugs faster, and even assist in gene editing. Companies like NuMedii are using AI to find new treatments for diseases like cystic fibrosis and sickle cell anemia.⁵ A recent study found AI could improve early-stage cancer diagnosis⁶, which could literally save millions of lives.⁷ AI is also being used to simulate complex drug interactions, saving billions in research and helping to get treatments to patients faster.⁵⁷

And it doesn’t stop there. From product development to climate modeling, AI is driving innovation at a pace we’ve never seen before. For example, PepsiCo (of all companies) used AI to reduce product development cycles by 40%, leading to lower emissions and better aligned consumer preferences.⁸ And BMW now uses AI not just in design, but to optimize entire supply chains, shaving off inefficiencies that once cost them millions.⁸

So if all of this sounds like great news… it is. Or at least, it should be. But here’s where the story starts to twist.

The Automation Avalanche

The same AI that’s helping us cure disease and build better tech… is also replacing human workers at an unprecedented rate. Goldman Sachs estimates that 300 million full-time jobs could be replaced by AI in the coming years.⁹¹⁰ That’s not a sci-fi scenario. That’s from one of the largest investment firms in the world. Customer service, banking, logistics, finance … entire sectors are being automated. A 2023 survey by ResumeBuilder found that 37% of companies using AI had already laid off workers, and another 44% expected to in 2024.¹¹ Even creative industries aren’t immune. AI tools are already writing marketing copy, generating visuals for ad campaigns, and even scripting rough drafts of TV shows.

It should be obvious why this concerns me personally.  AI is coming for my job. In the earliest days of my career I did UI and graphic design work and got paid good money to work in Photoshop. Now, there are tools I use today that can isolate the subject of a photo, drop it into a completely different space, blend the lighting conditions, and give me a fantastic looking YouTube thumbnail … all from a text prompt and attaching some source photos.  No Photoshop required anymore.  I’m still part of that process as the human and creator, so in theory, these tools boost productivity. That means fewer people are needed to produce a given thing.  You can do more with less.

And that creates a chilling question: If more people are out of work, who’s left to buy the products and services these companies offer? That’s the paradox. Because in economics, there’s an assumption: that productivity leads to prosperity. But if AI automates people out of an income, the demand side of the equation starts to collapse.⁹ This is what most people, and most businesses, aren’t talking about.

The AI Paradox

Let’s break it down. Companies like Tesla aren’t just building cars anymore: they’re building robots. CEO Elon Musk has said Tesla’s value will eventually rest more on Optimus, their AI-powered humanoid robot, than on their cars.¹²

Tesla is betting big on AI not just for self-driving vehicles, but also for fully autonomous manufacturing and service robots.¹² Their stock price reflects this belief, with analysts pricing in trillions in future revenue from products that don’t yet exist.¹² But who will buy those cars (or anything else) if automation leads to mass unemployment? It’s a feedback loop with dangerous implications.¹⁰

We saw this kind of paradox before, in a very different form. It’s called Jevons Paradox: when efficiency gains lead to more consumption, not less.¹³ In the 19th century, as steam engines became more efficient at using coal, coal use didn’t fall … it skyrocketed. People found even more ways to burn it, driving up overall demand. Efficiency didn’t solve the problem; it made it worse. But with AI, we might be seeing the reverse: greater AI efficiency leads to more automation, which leads to job loss, which leads to less consumer spending … even as companies ramp up production.¹³

Let’s call it the AI Economic Paradox.

This isn’t just theory. We’re already seeing early signs. Companies save money by automating, but consumer demand isn’t rising fast enough to match. Imagine this at scale: AI enables a biotech company to develop revolutionary new drugs, but because millions are out of work, no one can afford them. Or AI invents a super battery, but EV demand shrinks because people are struggling to make rent. Or retail companies use AI to optimize supply chains and logistics, reducing their need for warehouse staff and drivers, only to see their revenue decline because those same workers can no longer afford to shop.

We’re not just talking about lost jobs. We’re talking about an entire economic structure potentially undermining itself. The system is designed with a key assumption: that automation leads to lower prices, and lower prices lead to more buying.¹³ But that’s only true if people still have income. If AI replaces humans in a way that severs that income, and no replacement system is created, then the whole thing starts to fall apart.

Are New Jobs the Answer?

To be fair, there’s a counterargument: that AI won’t just take jobs. It’ll create new ones, too. And that’s true… to a point.

The World Economic Forum estimates AI will eliminate 85 million jobs by 2025, but create 97 million new ones in fields like data science, AI safety, and robotics.¹⁴ Remember that scary sounding Goldman Sachs report earlier about 300 million lost jobs in the coming years? Well, there’s a more recent analysis from the company looking at the longer-term possibilites. In that report they project a 7% increase in global GDP and a 1.5 percetage point boost in productivity over a 10-year period.^{15 16 17}

Sounds like a good trade, right? But here’s the rub with those assumptions: many of these new jobs that people will transition into require highly specialized skills. And historically, when industries transform, workers don’t always transition easily.

Take this example: Between 2000 and 2010, the U.S. lost 5.6 million manufacturing jobs. By 2010, only 39% of displaced manufacturing workers were reemployed,^{18 19} and most who found new jobs earned less than before.^{20 21 22}

The AI job boom is concentrated in metro areas with strong tech sectors, like San Francisco, Boston, and Shenzhen. That leaves behind rural and industrial communities without access to retraining or relocation support. Plus, even the new jobs being created can often involve supervising or refining AI systems that are doing the actual labor. A single AI manager might oversee processes that used to take 50 people. So the net employment effect is still uncertain.

And here’s something that rarely gets mentioned: Not everyone wants, or is suited for, retraining into tech.²³ We can’t expect a 58-year-old factory worker in Ohio to seamlessly become a machine learning engineer. So while new jobs may come, the distribution of those jobs, and the time it takes to retrain, may not match the pace of disruption.

This isn’t just a labor market issue. It’s a social and political one. If we don’t address the mismatch, we risk pushing more people into poverty, increasing inequality, and fueling social unrest.¹¹

The Human Cost of Ignoring the Paradox

We’ve seen what happens when economic shocks aren’t handled well. The 2008 financial crisis led to years of stagnation and a massive erosion of public trust. The COVID-19 pandemic accelerated automation even further, particularly in logistics, food service, and retail, permanently eliminating millions of jobs.

And now, AI is supercharging that trend.

A 2022 study published in the journal Demography found that an increase in automation between 1993-2007 led to increases in drug overdose deaths, suicide, homicide, and cardiovascular mortality.^{24 25} These aren’t abstract numbers. They’re indicators of human suffering.

And most worryingly, the companies at the center of this transformation are not incentivized to slow down.²⁶ In fact, markets reward them for doing the opposite.²³²⁶ So unless governments, institutions, and society as a whole intervene, we may be heading for a future where innovation explodes… and prosperity implodes.

Possible Solutions

So what do we do? There are ideas out there. Some are highly controversial. Others are still in pilot phases.

One very controversial proposal is Universal Basic Income, a flat monthly payment to all citizens that ensures a minimum standard of living.¹¹²⁷ This would decouple survival from employment and give people time to retrain, start businesses, or just live with dignity. In Finland, a basic income trial showed increased well-being, health, and small business formation, even though it didn’t significantly increase employment.²³

Another idea is an AI dividend, a tax or licensing fee on companies that automate jobs, which gets redistributed to displaced workers or invested in job retraining.²³ There’s also re-skilling programs. But they need to be dramatically scaled.

And finally, there’s the idea of decentralized AI ownership: models where AI tools are open-source and co-owned by cooperatives or communities, so that the economic value created doesn’t concentrate in just a few hands.²⁶

These aren’t magic bullets. But they represent a shift in thinking from maximizing efficiency at all costs to designing for inclusion and resilience.²⁸ We’re at a pivotal moment. The decisions we make now will determine whether AI becomes a force for broad human flourishing… or just another engine of inequality.

The AI revolution isn’t coming. It’s already here. But the most important breakthroughs, like the ones in medicine, energy, and science, are being overshadowed by flashy demos and short-term profit goals. We need to look deeper. Because the real risk isn’t AI turning evil or taking over the world.

It’s all of us using it so recklessly… that we destroy the very systems that let innovation thrive.

AI could solve some of the biggest challenges humanity has ever faced. But only if we solve the AI paradox first.