Derailing our Train of [artificial] Thought
(A note from the author: This was originally written in late December 2025, however I chose to table it as the societal obsession with AI was fairly quickly tabled in favor of geopolitical events. That said, I wanted to share these thoughts rather than wait any longer. )
Let’s kick off 2026 with a little thought experiment: Consider all of the incredible innovations, inventions, and creations we humans have dreamed up and brought into reality through all of known history. That alone is a daunting task, but it’s a new year and we are feeling ambitious, so let’s make it even more challenging: Now, distill the list down to the top 5 most significant to the advancement of humanity?
Chances are, your answers likely have some overlap with any other person’s answers, but rarely will we all agree on the exact same top 5. Your priorities and interests will heavily bias your answers. A doctor may gravitate towards health innovations while an engineer may gravitate toward material sciences. For most people though, top contenders are often the wheel, papyrus, the moveable-type printing press, the steam locomotive, the internal combustion engine, the phone, and the transistor/microprocessor.
The reality is that it would be nearly impossible to distill all of humanity’s achievements to a list of top 5, but there is one thing that every revolutionary innovation shared at its inception: it was polarizing and divisive. While it may not have been a 50/50 split, there was a massive constituency of sycophantic evangelists and a tremendous number of naysayers and critics. And chances are, both groups were right in many ways (and wrong in many ways too). The fact is that people, despite our greatest efforts, are not really good at seeing how the future will unfold. In fact, it is usually the “innovations” that garner almost unanimous, unchecked support that are often the biggest flops. Strong skepticism, it seems, is a requirement for any great innovation to mature successfully from groundbreaking idea to revolutionary innovation. It’s like putting a wooden spoon across the top of a boiling pot of pasta to keep the starchy water from boiling over.
Okay, let’s not bury the lead any further. Chances are by now you know where this is going, and you are not wrong. We need to talk about AI, but not in a way you may be expecting. This isn’t a pro or a con conversation. Remember, we are not really all that great at seeing how the future will unfold, but we can see how the past unfolded. So now, let me present to you one other question: Of the list of innovations previously mentioned, which one most closely resembles AI, in your opinion?
It is safe to assume you are gravitating toward the transistor/microprocessor — and for good reason. They are both modern day electronic tech innovations. One is essentially built on the back of the other, and they are both foundational innovations without which “AI” would likely not exist. But what if I told you that AI has more in common with the steam locomotive than any other innovation? The story of the steam locomotive is possibly where the most insight can be gleaned for how we may see the AI saga play out. So how is AI similar to the Iron Giants of yesterday?
When it comes to innovations like the wheel, papyrus, and even the printing press, their brilliance was in the idea. The means to bring that abstract thought to reality was, and still is, actually quite unsophisticated and accessible. In fact, this is the case for most innovations, which is why we have (for right or wrong) agreed to create and adopt patents and intellectual property laws. It is in the abstract idea itself, where the center of mass of the innovation exists. This holds true for the telegraph and the telephone as well. Once the idea was dreamed up, bringing it to fruition was fairly “simple,” all things considered. To this day, this is why there remains some controversy as to whether Alexander Graham Bell is rightly credited with the invention of the telephone. Many others, including Antonio Meucci, may be more accurately credited with the invention but unfortunately the very institution that was supposed to protect intellectual property is why his name has fallen into obscurity (he couldn’t afford the patent). But the important takeaway is that the idea was the difficult part, actually building a telephone was fairly easy and inexpensive. History also shows the steam engine, the internal combustion engine, and the transistor were all remarkably accessible, as well. Sure, we’re not talking about pocket change, but given the conditions of their respective day, with relatively minimal resources, the average person could build, functionally use, and improve upon any of these great innovations.
The steam locomotive, however, was different and for one distinct reason. Sure, to build a steam locomotive would be expensive, but it was doable. Using a steam engine, on the other hand, was a different story altogether. This distinguishing factor both separates the steam locomotive from almost every other revolutionary innovation and ironically became the very reason for the eventual demise of the locomotive’s market dominance. In order to operate a steam locomotive, rails and land were necessary. Or put differently, tremendous amounts of capital and resources were necessary. This becomes a critical bottleneck in the maturation and success of the steam locomotive as a revolutionary technology. The success of any innovative development in locomotion was no longer solely decided by a marketplace of ideas, it was decided by a select few parties who wielded tremendous control over the industry, as a whole. And more often than not, this power consolidation was government sponsored and protected (despite the official narrative suggesting it was a lack of governance). The economic viability of any new innovation in locomotion required the approval of the gatekeepers who owned and controlled the railways. In America, this power and control ultimately concentrated into 8 names, four of whom operated as a consortium under the moniker of the Central Pacific “Big Four”. Today we refer to this small group of controllers as the Train Barons.
So how does the story of the locomotive parallel the current story of AI we see playing out? And more importantly, what can we learn from the past to better anticipate the trajectory of our future? There are three major characteristics in the tale of the locomotive that are currently echoed by the AI affair:
Extreme capital requirements
Innovation Cul-de-sac
The Innovation Squared Alternative
The first is just how resource intensive AI computation truly is. In order to support the extreme computational power of even the most rudimentary version of AI, an immense amount of energy is required. Not only does it require energy input in the form of electricity, but also energy dispersion. These massive computers produce heat, and a lot of it. This requires cooling, which in turn requires water or some other means of carrying the energy (heat) away from the computers. So not only do you need a means of accessing electricity but you also need access to and use of water. These don’t come cheap. Therefore, in order to run an AI program, you have two choices: you need to be able to afford to build your own data center, get government to sign off on it (while the current market participants lobby aggressively to stop you), and then be able to afford the ongoing resources necessary to operate. This would be like building your own railway network. It is tremendously capital intensive and the barriers to entry are all but impossible to mount. You have to have an immense amount of capital and likely other existing in-roads to do this (think Elon Musk and Grok). The other option would be to lease computation from an existing market participant — aka your competition. Do you really think they are going to be friendly if you out-innovate them? No. They will either put you out or they will scoop you up. This is where the power consolidation cycle is reinforced and also where the innovation cup-de-sac is born.
When creativity and exploration is not allowed to naturally flourish in an open market, a creative loop begins to form and innovative thrusts are kneecapped by not being allowed be fully fleshed out. If a technology or innovation is to try be revolutionary, it needs to be generally accessible so that more minds, preferably of vastly differing points of view can adapt, innovate, and build upon the foundational technology, taking it places no single innovator could singlehandedly imagine. We will never hear of the countless innovators who had plans and designs for “who-knows-what-kind” of wild, fantastic, game-changing locomotive that never saw the light of day because they were crushed into obscurity by the hand of the Train Barons who did not want to see their industry and interests disrupted. This is only possible when a revolutionary advancement is not widely accessible, usually as a result of capital constraints. Instead we get a monoculture of innovation that may at first not be noticeable as the fervor and excitement around the baseline concept is enough to self-sustain. But newness always fades and without uninhibited creative divergences, the innovation simply starts to go in circles, as though it is trapped in a cul-de-sac. In most cases, the cul-de-sac eventually becomes more like a toilet bowl, and we all know where that swirl ends up.
Recall the earlier suggestion that the ultimate undoing of the locomotive was the very thing that made it different from other revolutionary innovative leaps. Inevitably, creative and innovative minds notice the gatekeeping and inherent liabilities associated with the pursuit of building a “better train”. Attention naturally diverted to the pursuit of more unencumbered and liberated alternatives. Thus, the proliferation of the automobile resulted. This is what I will refer to as the “Innovation Squared Alternative”. The expensive, inaccessible, gate-kept, innovation is eventually made obsolete as brilliant innovators seek out alternative means to a similar end, but with less intensive capital requirements (and therefore dramatically less gatekeeping). In America, we truly saw this on a grand scale. The eight dominating Train Barrons may have controlled the rails, but carriage roads and dusty trails presented a far greater opportunity to far more opportunists. It was less than 30 years after the transcontinental railroad was established in the US that we saw the emergence of over 200 automobile manufacturers. There is no way to know for sure, but if the locomotive and the railroads had never connected distant towns and cities to one another, there may have never been enough interest or demand for a more autonomous mode of transportation. The Innovation Squared Alternative is sought by those who would normally improve upon or develop the next revolutionary advancement to a foundational innovation. Instead they choose an alternate route with less “tolls” but still achieves a similar end. To this day we still enjoy the primary, secondary, tertiary (and beyond) benefits that have emerged from those who sought the same goal of efficient, commuting and travel, without the encumbrances and gatekeeping of the rails. Whether it be the automobile, air travel, or even
By no means is it intended to insult your intelligence, nor to oversimplify the nuances of history or today, but let’s just make clear the parallels being drawn. Currently, if you were to look up the most used AI’s, you will get an obscure list of all kinds of AI names, but the fact of the matter is all of these are built on one of about 5 actual AI model platforms. Almost every chatbot, image generator, or other AI tool is built on OpenAI’s GPT, Google’s Gemini, Microsoft’s Copilot, Anthropic’s Claude, or X’s Grok. There are a few smaller players like Perplexity, who are making ground, but the fact is, any and all other tools are simply piggybacking on one of the gatekeepers. These are the Train Barons (AI Barons) and the data centers, full of mostly Nvidia GPUs sucking up electricity like a sponge are the railroads. No matter what great AI innovation someone may have, they likely won’t be able to see it come to fruition without riding on the rails controlled by the AI Barons. And if an idea is good enough, chances are that Baron is going to “make you an offer you can’t refuse”. The story of the locomotive as a world-changing innovation is now playing out once again. For this reason, I am far more optimistic about the innovation squared initiatives that are being taken by those who see the technological landscape as it is. And I firmly believe this is the disruptive “needle” that will have the point that bursts the speculative bubble that has formed.
Put differently, it is my assertion that there is a bubble because most investors are asking the wrong question. However, I do not see this being a “reversion to the mean” bubble, rather it is an “innovation expansion bubble”. Yes, it will pop, but not everyone and everything will be impacted to the degree many fear; and there remains a good chance that even those who do suffer most, will be able to endure and come out ahead. So what incorrect question(s) are most focused on, and what should they instead focus on? Right now, most investors are laser focused on
Opportunity: What is the capability and application of the technology?
Revenue: When will everyone adopt (and pay for) the AI? (E.M. Rogers’ “Diffusion of Innovation Theory”.)
I believe both of these questions are the wrong ones to ask. Instead I believe the prudent investor will find a far greater return on investment and significantly less volatility if instead they understand that AI is
An Energy Question
A Privacy & Control Question
A Human Question
The first question in the most tangible and objective. The reality is that so many people seem to have become untethered from our natural, physical reality. Pre-consumed with our digital world, they often overlook the reality of physics, but in order for any of this to work, it needs energy — and a whole lot of it. In addition to this being a layer of gatekeeping, it is also a question of sustainability and usefulness. The innovators focused on how to achieve similar outcomes with far less energy, or those who are looking to optimize existing energy sources stand to make some of the greatest near term impacts. There have been incredible advances in nuclear power, allowing for micro-reactors. Another consideration gaining traction is using jet engines as a power source. Notably, Elon Musk is aggressively dismissive of all of these power creation solutions. He is a firm believer that the answer lies in harnessing the sun. While this makes the most sene from a physicist point of view, its applicable timeline seems to be in conflict with what we might consider to be an appropriate timeline for an investment thesis. While they may eventually become the Betamax to the Sun’s VHS, or maybe the Zune to the sun’s iPod, it currently appears that non-solar power innovations currently show a brighter near term future. In many ways, focusing on the power needs of AI is akin to the old saying that “The speculator digs for gold. The Investor owns a shovel business”.
Regardless of the power generation technology, the presumed physical laws of energy remain one of humanity’s greatest hurdles. Unfortunately for us all, we recently lost one of the most impressive minds who may have been on the verge of several unparalleled breakthroughs. Leading plasma research scientist at MIT, Nuno Loureiro, was killed in December under very suspicious circumstances, leading many to question if perhaps he had discovered something that would have radically changed the world, and put many very wealthy and powerful people out of business. While his death will remain a tragedy, his research may eventually lead to cracking the energy code. In the meantime, the other side of the equation is power demand of these technologies. The aforementioned nuclear, jet, solar, etc. proponents are focused on generating more power to supply the current need. Understandably, they will and already have focused on securing fat contracts with the AI Barons, continuing to create congestion in the Innovation Cul-de-sac. Therefore, the other innovation squared alternative will be those who seek to achieve scalable AI computational capacity while using a fraction of the current energy demand. The Chinese Deepseek AI team seems to be leading the way in this effort, however there is some speculation about the integrity of their claims. Some focus has turned to agentic AI as a possible approach but there exist significant concerns around capability, as well as privacy and control. This opens the door for the second question we should be focused on.
Privacy and control is one of the greatest risks AI presents. Whether it be control over one’s own livelihood (I.e. job replacement), control over one’s own image and likeness (I.e. deepfake content/media), privacy over one’s own information, or most importantly privacy over the degree to which AI is imposed on the average person without their knowledge or consent, the privacy and control question is one that nobody can answer. Much the same way McAfee, Avast, Symantec and others created an entire industry on the shoulders of the proliferation of the PC and the early internet, there is a ripe opportunity for those who seek solutions to protect people from the known and unknown capabilities of AI. But there is also a ripe opportunity for the Innovation Squared Alternative to centralized data center AI. By innovating away from data centralization to some form of a decentralized network, anonymity, privacy, and protection may be possible for both innovators in the technology and end users. If the railroad was the centralizing shackles of the locomotive, the road was the decentralized network that provided the necessary means for the innovation squared alternative of the car to succeed. What will be the analog for the AI saga?
Finally, we need to ask the human question. Yes, AI is a technology story once dreamed of in science fiction and fantasy novels, but one cannot overlook the fact that at its core, it is a human story. In these our modern days, we have explored space but have yet to come even close to exploring the vastness of our own lands and seas. So it should come as no surprise that we are seeking to create consciousness when we have no true understanding of our very own consciousness. In many ways, it seems we humans will often use our ability to create tools in order to better understand ourselves. To the average person of the 18th century, Pierre Jaquet-Droz’s automata was incomprehensible. In his private records, he lamented that he would be arrested for sorcery. To the people of his time, the Swiss watchmaker had created artificial life, however even a child today would laugh at such a notion. What seems very real today will almost certainly be seen as quaint or rudimentary tomorrow. Perhaps this will not actually lead to sentience, intelligence, or the infamous “singularity”. Perhaps we will instead discover the limitations of this innovation are insurmountable but instead we gain a greater understanding of ourselves in the process. Better yet, what if the opportunity set provided by AI simply allows more minds to spend time contemplating philosophy, art, theology, health and nutrition; In a sense we see the emergence of another renaissance? While it may not catalyze fiscal wealth, that very well may be the hidden benefit few currently see. What if the next phase of our development as society and culture is a harkening back to a simpler life that is in fact, more enriching, satisfying and worth living. Less materialism, more intention.
Regardless of whether we really do create artificial general intelligence or we simply created a tool capable of simplifying menial tasks, one thing will be necessary if it is going to continue to improve humanity and our world. AI interfacing will need to be more human. Currently, the language based models are helpful but are not capable of interfacing in a way that is natural for humans. We use all five (maybe 6? Pineal Gland anyone?) senses in almost every moment of our lives. When speaking person to person, visual cues are often louder than most verbal messages. When walking through the door at Grandma & Grandpa’s house, the smell hits you right away and you immediately feel at ease and at home. Kids who are physically active while socializing bond far more than those who are sitting or not moving. Humans are, well, human. For AI to be a revolutionary innovation it will need to become more human, beyond its language parameters and intonations. It will need to integrate visualizations, auditory cues, and other extra-sensory modes of engagement. This is what brought the telephone into every home while the telegraph never made it out of the train station or corporate office. But here is the catch, when it comes to AI, the natural bet may not be the right one. Most suspect that AI robots will be the outcome. Something like Rosie from the Jetson’s. But the uncanny valley is less of a crack in the sidewalk and more of a Grand Canyon. The innovative advancements that manage to thrust AI over that chasm very well could be the one definitive victor in the AI race, but I’m not betting on those odds any time soon. I do, however, see some creative proposals beginning to emerge that will allow this tool to be more readily and naturally used, but this takes us back to privacy and control. Only time will tell how this question is addressed.
