Artificial Intelligence: The Most Misunderstood Technology of Our Lifetime
Every generation has new technology that people believe will change the world. The steam engine. Electricity. The automobile. The internet. And many others worth debating. The fact is, Artificial Intelligence (AI) probably deserves a place on that list. Unfortunately, it may also be one of the most misunderstood and overhyped technologies we've ever seen. Depending on who you ask, AI is either about to cure every disease known to man or become self-aware and enslave humanity sometime around Christmas. Reality, as it normally is, is probably somewhere in between those extremes.
Part of what makes AI different from previous technological breakthroughs is that it remains something of a black box, even to many of the most knowledgeable people in the technology industry. The field has become so specialized that experts in one area of AI will readily admit they do not fully understand every other area. That complexity has created plenty of room for speculation, hype, and misconceptions.
To understand AI, you have to have a basic understanding of the concept of computers. Every computer on Earth - from the phone in your pocket to the most powerful AI supercomputer ever built - is ultimately doing the same basic thing. An oversimplification is that everything boils down to tiny electrical switches called transistors. A transistor is nothing more than an incredibly small on/off switch. Think 1s and 0s. Modern chips contain billions of these microscopic switches, each flipping billions of times every second. The sheer speed is almost impossible to comprehend when you sit back and think about it. But this technology is the basis of how you can pull out your phone, speak into it, and someone on the other side of the planet can hear your voice in an instant. This is because when you speak into your device, your voice is sampled thousands of times every second, converted into digital information, compressed, transmitted across the globe, and reconstructed almost instantly.
The real breakthrough today isn't that computers suddenly became smarter. It's that those switches became unimaginably small, unbelievably fast, and incredibly inexpensive. Exponentially smaller and faster than before. So the unimaginable becomes even more unimaginable. Every generation packs more computing power into a smaller package than the generation before it. When you compound that progress over decades, the results begin to feel almost magical. That's how we went from room-sized computers with less computing power than a modern calculator to speaking naturally into our phones and receiving an intelligent response almost instantly.
When people talk to ChatGPT or another large language model (LLM), it often feels like there's some form of consciousness on the other side. However, LLMs aren't sitting there contemplating life and forming opinions based on emotion. Instead, they've processed enormous, incomprehensible amounts of human writing and learned incredibly complex patterns in human language between words, phrases, concepts, and ideas. When you ask a question, the model predicts what sequence of words is most likely to come next by building complex internal representations of concepts, relationships, and reasoning patterns. It’s hard to imagine a process like this could be so successful, but at an incomprehensibly large scale of computing, results feel conversational because the AI has become extraordinarily good at recognizing patterns and can pore over an indescribable amount of data in a near-instant. This technology and the concept of AI aren't brand-new ideas. Researchers have been working on neural networks for decades. What's changed is that today's computing power, data availability, and specialized AI chips have finally made these systems practical at an enormous scale.
Think about teaching someone to play a complicated game like Chess. Their first game is usually terrible. By the tenth game, they're noticeably better. After a hundred games, they’re significantly better. Now imagine playing millions and billions of games instead of hundreds - but not over several years, in an instant. That's machine learning - or reinforcement learning, to be more precise. The computer isn't "thinking" about chess the way Bobby Fischer or Magnus Carlsen does. It's repeatedly adjusting based on previous results until it recognizes incredibly subtle patterns that humans would never notice. Google's AlphaZero famously became one of the strongest chess players ever by playing millions of games against itself and eventually developing strategies that caught even the most experienced grandmasters by surprise. The same idea applies to image recognition, voice recognition, medical diagnosis, and self-driving technology. If you feed enough examples into an incredibly powerful computer and allow it to adjust itself billions of times, remarkable abilities begin to emerge that are not humanly possible. Human beings often mistake fluent language for understanding. If someone speaks confidently, answers quickly, and remembers enormous amounts of information, our brains naturally assume there's a conscious mind behind the words. AI inadvertently exploits that tendency. It can sound remarkably human without actually experiencing emotions, desires, or self-awareness.
This is why AI deserves the excitement it has received. It will make doctors more accurate and engineers more productive. Scientists and businesses will be more efficient and operate faster. How will that translate to future jobs? Of course, there are jobs that will be eliminated. We are already seeing that. Quite possibly, AI will eliminate an economically unhealthy number of jobs in due time. If I said I knew for sure, I would be lying. No one knows for sure just how far AI will take us.
Much of the debate centers around the idea of Artificial General Intelligence (AGI) - the hypothetical point at which AI can perform virtually any intellectual task as well as, or better than, a human. On the surface, that sounds both exciting and intimidating. With that said, it is important to make a distinction between perceived intelligence and actual consciousness. It’s also important to remember that humanity has imagined many groundbreaking concepts that have never become reality and may never do so. We've theorized about teleportation, faster-than-light travel, and traveling through time. These ideas have fascinated scientists and the public alike for generations, yet they remain theoretical. Simply because we can imagine something—or even describe how remarkable it would be—doesn't mean it will ever exist. AGI may one day become reality. It may not. At this point, anyone claiming certainty is simply expressing an opinion, not a fact. And hypotheticals are powerful drivers of investment hype.
None of this means every AI investment and company will succeed or fail. Not long ago, the largest and most influential companies in America were primarily energy companies that emerged during the petroleum era. Now, the tech companies have taken hold. When did this transition occur? Not overnight, but it did happen over time. Today, we can look back and say that while Exxon Mobil is still very relevant and influential, they are clearly looking up at companies like Amazon and Apple.
History has a funny way of confusing transformational technology with profitable investing. The internet changed civilization forever. Hundreds of internet companies still went bankrupt during the dot-com crash. Railroads changed America, but many railroad investors lost fortunes. Electricity transformed the world, but not every electric company became a great investment. The biggest investment mistakes are often made by confusing a revolutionary technology with a revolutionary investment. They are not the same thing.
Today we're seeing trillions of dollars flowing into AI infrastructure in the form of Chip manufacturers, memory producers, equipment, data centers, power generation, software, etc. Every company wants a piece of the boom. While some of this spending will prove worthwhile, a lot of it ultimately won’t. Eventually, investors will ask how much profit is actually being generated relative to their investment in AI. Capital expenditures will slow, and history suggests it often slows much faster than investors expect. Because of the circular nature of AI-related companies investing in one another, a domino-like effect will ensue. When AI companies stop buying chips, chip manufacturers stop ordering memory. Memory manufacturers slow production. Equipment orders decline. Data center construction slows. Eventually the effects ripple through the entire supply chain. Once it happens, it will look like it happened all at once without warning, but the signs will be there if you pay attention. We’ve already seen investors begin to question capital expenditures during earnings reports. While some big-name stocks have sold off, now is probably not the beginning of the end of the AI bubble. But this is the chorus that we want to watch grow louder as the risks become more pronounced. And when the AI bubble ultimately bursts and things get nasty, it doesn’t mean AI failed - it just means expectations got ahead of the economics, just like they did at the end of the 90s.
In summary, AI is almost certainly going to improve our lives in countless ways over the coming decades by automating work, accelerating R&D, and assisting businesses in being more productive. But despite what Hollywood would have us believe, today's AI isn't plotting world domination and developing thoughts about enslaving human beings. It's doing exactly what we designed it to do - recognizing patterns across staggering amounts of information. This part alone is revolutionary, and whether it eventually becomes something even more remarkable is a question only time can answer.
