AI Is Not Actually a Technology #
Historically, world-changing technologies didn’t devour industries with such speed.
Consider the automobile, invented in 1886. The number of horses didn’t truly begin to decline until 1920. By the same token, the Wright brothers took flight in 1903, but commercial aviation didn’t truly become widespread until the 1960s.
Even the internet was slow to spread. Early internet service providers like AOL became widely adopted around 1990, but by 2000, only about half of the North American population had internet access.
The slow adoption of these world-changing technologies was precisely because they afforded society ample time to react and adapt—a process that sometimes took a generation or even longer.
AI, however, is disrupting industries and taking over modern jobs at a pace orders of magnitude faster than previous disruptive technologies.
This has profound implications for how AI’s proliferation will affect society and how we, as humans, should respond.
ChatGPT Devours The World #
ChatGPT is arguably the first truly powerful AI technology widely accessible to the average person. This technology launched as recently as November 2022.
In just five days, ChatGPT garnered one million users. In a mere two months, its user base soared to 100 million users.
Today, a little over two years since its launch, the chatbot claims over 800 million users, roughly 10% of the world’s population! Even more astonishingly, at least 76% of companies regularly use ChatGPT — and that’s only counting companies that admit to it!
Competitors like Anthropic’s Claude have also amassed massive user bases. Google estimates that 1.5 billion people a month encounter their AI Overviews feature, which appears directly at the top of search results pages.
In just two years, AI has transformed from a technology primarily used by advanced software developers and academics into a tool used daily by billions, for virtually everything. What’s more alarming, this insane growth isn’t limited to the digital realm; it has spilled over into the physical world.
Waymo, an autonomous driving company that relies on AI, opened its ride-hailing service to the public in San Francisco in 2024. At the time, it offered limited services to its employees and journalists, with a negligible market share.
By August 2024, Waymo was providing 312,000 autonomous rides per month. By early 2025, Waymo had surpassed Lyft, becoming San Francisco’s second most popular ride-hailing company—directly displacing hundreds of human drivers and claiming roughly 30% of the market share. And it’s still growing.
Even high-end “knowledge work” is not immune. According to The Washington Post, by March 2025, over 25% of computer programming jobs had vanished. OpenAI’s Codex and similar products from Anthropic are reportedly rapidly replacing them.
We used to believe that acquiring high-level skills like programming would ensure a high-paying, irreplaceable job and a high quality of life for human workers.
Not anymore.
The Secret to Its Rapid Ascent: The Infrastructure Secret #
Why is AI devouring industries so quickly, while previous disruptive technologies took decades to have a similar impact? The secret lies in AI’s relatively low reliance on infrastructure and its ability to replace human labor very quickly and easily.
When past technologies were introduced, to realize their full potential, entire new industries often had to be created.
Take the automobile, for instance. Yes, the first internal combustion engine cars existed in the Victorian era. But for them to truly become practical, the entire world had to fundamentally change. The roads were unpaved and rutted—fine for horse-drawn carriages, but a disaster for the early automobiles’ tires and rudimentary suspension systems. Similarly, there was no infrastructure to extract, refine, and distribute petroleum fuels—early drivers bought canned gasoline from local general stores.
Cars were custom-made, built in local workshops, or assembled by hobbyists from parts. It wasn’t until people like Henry Ford began innovating—investing massive sums—that cars became affordable and standardized enough for the average person to buy. Keeping them running well and somewhere to park them also meant building out a network of repair shops and new ways of building homes and commercial properties.
In short, even when the innovation itself existed, it took decades to build out all the necessary infrastructure and systems for the technology to truly become functional.
AI is different.
Yes, today’s AI models have a massive amount of physical infrastructure supporting them. But most of it is hidden away in distant cloud computing centers—something the average user doesn’t need to worry about, or even know about. And because it’s so highly centralized, building AI infrastructure is comparatively much easier. While the demand for AI data centers grows by 33% annually, supply has largely kept up, apart from occasional service disruptions.
Humans, Step Aside! #
Beyond its minimal infrastructure requirements, AI differs from other technologies in another critical way: In many cases, AI can directly replace human labor on a one-to-one basis.
For instance, a traditional ride-hailing company like Uber, if it wants to increase capacity in a city, needs to recruit hundreds of people, onboard them into the Uber system, and provide basic driving training. This is a significant cost. In 2021 alone, Uber spent $250 million training and incentivizing drivers to return to work after the pandemic. Moreover, drivers who quit or simply choose to take breaks during holiday rushes and lunch peaks all limit their ability to grow.
Waymo operates differently. If it wants to increase capacity—or expand into new cities—it just needs to build a few more cars and deploy them. Reportedly, there are already 300 vehicles on the road in San Francisco.
They perform the same work as human drivers (arguably better), but require no wages, never get sick, and can work 24/7 without breaks or holidays.
The same applies to knowledge work. Ten years ago, if you wanted to build a web application, you might need a large team of programmers, specialized backend engineers to create database systems, frontend engineers to write Javascript, and designers to create CSS stylesheets and responsive designs. But now, the rise of “vibe coding” means that a single person, with minimal coding knowledge, can build equally powerful web applications in a matter of days. I recently built a personal AI tool that corrects Siri’s transcriptions, allowing me to dictate emails and articles directly into my phone and get a clean text. It took me 35 minutes and didn’t cost a dime.
Low infrastructure requirements, coupled with instant human labor replacement, combine to explain AI’s rapid ascent. Truly, these characteristics make AI less of a technology and more of a foundational discovery capable of social transformation on multiple levels.
Google CEO Sundar Pichai once likened AI to fire and electricity, stating its potential impact on the world would be just as profound. This might sound hyperbolic.
But at a fundamental level, he’s correct—like those powerful tools, AI can integrate into virtually every facet of human endeavor, making it more efficient and cost-effective, while also bringing about fundamental, unpredictable change.
What Should We Do? #
In the face of such rapid transformation, what exactly should we, as humans, do?
Many commentators treat AI like other disruptive digital technologies. The same voices that urged “Learn to code!” a decade ago are now shouting “Learn AI!” Major universities are rushing to offer courses on prompt engineering and other skills they deem crucial for success in the age of AI.
The problem with this approach is that it treats AI as an ordinary technological tool that merely needs to be understood, systematized, and then put to use. But again, AI is not like other tools.
Consider how early interaction techniques with AI have long since become irrelevant due to technological advancements. When I tested a predecessor to ChatGPT in 2021, it required detailed, specialized prompts to make the machine understand. It felt like programming. But now, Large Language Models (LLMs) are incredibly adept at understanding user intent. You can converse with them as you would a colleague, not a computer. If I had spent years learning the intricacies of prompt development for 2021 AI systems, by 2025, that effort would have been wasted.
Teaching AI as a traditional skill simply won’t work. The technology changes too rapidly to be taught as a conventional skill.
Similarly, some advocate a Luddite approach, completely resisting AI.
Yes, there will always be value in bespoke, non-AI spaces. This platform we’re on is one such example. Just as organic agriculture creates value by avoiding certain tools and technologies, non-AI spaces will continue to exist and thrive.
But they will be small in scale. In the United States, the organic food market generates $52 billion annually, a respectable market. But the entire agriculture and food market stands at $1.537 trillion; compared to which, organic food represents a very small, yet still significant, share.
The same logic will apply to AI. Some fields (like writing and art) will always benefit from non-AI-generated works—especially when the creator’s reputation or brand is an integral part of the work’s appeal.
But in fields like programming and driving—or even medicine or law—where people care more about the outcome than how the work is done, AI will dominate, and the Luddite approach will not work on a large scale.
Thus, the solution is not to avoid AI, nor to treat it like past teachable technologies. Instead, it’s to find a way to integrate AI into our lives in a way that enriches them without sacrificing our humanity.
I’ve found a perfect example in my own work and life: my AI transcription corrector.
I prefer to write my own articles rather than delegating them directly to ChatGPT. But I dislike the laborious process of typing thousands of words on a computer, nor do I enjoy feeling confined to a desk for hours at a time while writing.
With this AI-powered transcription corrector, I can dictate articles into my phone. (The corrector preserves my original words while AI corrects 99% of dictation errors.) I never have to type again.
I can also write anywhere, anytime. Instead of being cooped up indoors for hours, I’ve started dictating stories while walking around town, sipping Philz coffee. It’s an incredibly better experience.
The key is that this personal AI tool hasn’t replaced me or supplanted my human thought or ideas. It simply removed the tedious parts of my work, allowing me to work in different ways, in different spaces—and often at a much faster pace.
The same principle applies to programming. Vibe coding and new tools like Codex are excellent for rapidly developing simple applications. But at the level of large tech products and major corporations, the best programmers aren’t humans who completely eschew AI, nor are they AI without human oversight. Instead, they are skilled humans who understand how to leverage AI to accelerate their work.
This is the approach we need to take to safely and joyfully integrate AI into the human experience – and to harness its accelerating disruptive power.
We need to identify the elements of human creation that we don’t want (or can’t) replace, and then, with the exception of specific AI-free spaces, figure out how to apply AI everywhere else.
This approach fully leverages AI’s incredible speed and transformative power. But it doesn’t require us to bow down to it or treat it as a replacement for our own intelligence and humanity.
We can use AI without being devoured by it. We can improve our work without making ourselves obsolete. We can use fire without being burned.