Using History To Understand Artificial Intelligence
Artificial intelligence is the only thing that matters in the world right now, or so it seems. AI is literally the beating heartbeat of the economy, and Wall Street wouldn’t have it any other way. But let’s talk about this trend without the creepy Silicon Valley doubletalk, without the uninformed Washington D.C. babble, and without the dinner table platitudes. Let’s talk about AI like a nerdy historian.
In this newsletter we’re going to talk about the nitty gritty history of some AI comparable innovations, look at what that means for our future, and learn some obscure facts that I guarantee neither you nor your boss has ever heard of.
If you like reading interesting, unique things from a youthful angle, we highly recommend you listen to our podcast below & share this with friends! This takes a lot of work & sure keeps us busy.
Ok. So let’s talk about GPTs. Haha, nope! Not that one. The GPT in “Chat-GPT” stands for Generative Pre-trained transformer. Let’s talk about the other GPTs – general purpose technologies. These GPTs are practically a once in a century innovation, and often change the course of history. So what counts as a GPT?
A GPT must:
Continuously improve (and a compound growth chart)
Applies across many sectors
Requires complementary innovation and advancement to unlock full value.
AI meets these rules, without a doubt. What else has? Steam engines did in the 1700s. Electricity did in the late 1800s. Internal combustion engines did. Also the Internet. We will dive deeper into each of these innovations in this article.
The general idea is that the biggest innovation comes decades after the first firing. Electricity was invented in the 1870s but didn’t become the norm until nearly forty years later. There are a few questions that investors, observers, and basically everyone else must ask in period of GPT innovation:
What complementary systems (such as enhanced electricity grids, digital wearables, and LIDAR tech)need to change?
Where is it still bolt-on and not native?
Who benefits first v. last?
These questions are of monumental importance, and they will dictate the winners, losers, laggards, and stars of the next thirty years. We’re not going to act as if we have the perfect answers – we just have some thoughts informed by the historical record that provide us with accurate, time-tested, and logical principles.
Before we dive closer into the various GPTs that came before the “Chat” - GPT, I want to draw your attention to the productivity paradox. Far too often in periods of great innovation, people expect 10x productivity tomorrow and 20x productivity in five years. The reality is closer to 1.01x productivity tomorrow and 100x productivity in five years. Time is your friend. Technology must be learned by the users. Workflows must not just be automated–they must be redesigned. The phrase often attributed to monetary policy, “a long and variable lag,” indeed applies also to the context of GPT-level innovation.
Let’s jump into our first historical parallel: the steam engine. Now, some say the steam engine was described by Hero of Alexandra in the first century AD. But, alas, we will assume the water pump of Thomas Savery was the first in 1698. The pump was near useless and was prone to boiler explosion. When Savery told the world he’d finally cracked the code and invented the “steam engine,” they didn’t reply with “Hoorah!” they replied with “What the hell is a steam engine?” Thomas Newcomen then came along and invented the atmospheric engine in 1712. This engine allowed for continuous transference of power to a machine. In 1720, Jacob Leopuld invented the two-cylinder high-pressure steam engine. Two heavy pistons would undulate and the motion would power a water pump. Finally, in the 1770s, James Watt made the largest leap yet. At last, the productivity gains began to accrue. Factories were reorganized, and labor changed forever. Mind you that we don’t grieve for the jobs taken by the steam engine – maybe we shouldn’t do so for the jobs taken by AI? Just a thought.
As a reminder:
Thomas Savery in 1698.
Thomas Newcomen in 1712.
Jacob Leupold in 1720.
James Watts in the 1770s.
That’s a 75+ year journey. So is ChatGPT our Savery moment? Or our Newcomen? That’s the important question to answer. Is Sam Altman our James Watt? We doubt it.
Next up, I’ll talk briefly about electricity. People very often misunderstand this parallel. The big theme here is a system renovation, not just dropping a technology into the status quo and expecting it to make some massive change. Edison invented his lightbulb in 1879 and factories began electrifying almost immediately. The problem? These were still factories designed for steam!
Productivity began to really surge in the 1920s. Implementation of decentralized motors, assembly line redesign, and flexible factory layouts added fuel to the productivity gains. A huge lag – and a major shift in system design – was needed. Electricity was not “better steam” and it was useless to think of it like that. This isn’t so different from using AI to write emails faster and thinking it’s the final form of the technology. It can do more, we just don’t know it yet. We likely aren’t even capable of fathoming the potential for AI. Only time can tell.
Third, the internal combustion engine. These engines allowed for cars. Cars changed the world. But initially, although the technology was nearly perfected, cars were toys for the ultrarich. Horses still dominated transportation for decades following the invention of cars in the late 1890s. We needed to innovate things like roads, gas stations, traffic laws, suburban zoning, etc…Regardless, supply chains accelerated and personal mobility changed forever. Regulatory advancement and consumer education were essential. Consumer tech needs to have massive and coordinated education of the users. Whoever can do that with AI the best, wins. Intuitive integration and a strong product interface are vital.
The internet provides the best blueprint for market dynamics surrounding AI. In 1969, there was ARPANET. Then came the huge and frothy internet bubble. And then the crash. Real value creation only came later, between 2005 and 2015. Markets are, after all, human. They get far too excited and they get far too angry. Expect the exact same to occur with AI.
Ok. So answer these for yourself, and no cheating.
What point in AI innovation are we at?
What system changes are needed for AI?
What consumer education is necessary for AI?
What does the AI regulatory outlook look like?
When will markets actually get AI pricing right?
See ya, folks. Stay curious.
–J&E