Thoughts on AI
Thoughts on AI
5 min read
·
Just now
\--
Share
Humans have a tendency to overindex on the sequence and importance of events while they are happening. This is the phenomenon we are experiencing as we go through the wave of artificial intelligence that has almost taken over the world, at least in the developed parts.
No matter how much media pundits out there proclaim that this is an unparalleled transformation in the history of mankind, I believe there have been equally, if not more, transformative events that have shaped humanity into what it is today. There are numerous examples to that effect , more than you can count on two hands. Think of the agricultural revolution 12,000 years ago, which transformed humans by taking them out of the cave and putting them in a settlement that survived through organized labor, which we now call the society.
Think of the invention of the ship as early as 3,500 BC in ancient Egypt. In medieval times, the ship was the means to conquer and inhabit remote lands, with disastrous consequences in some cases. Think also of the more modern, systematic supply of clean water and electricity, which revolutionized everything from textile manufacturing to eliminating the need to haul water from the nearest river or water hole. In fact, the invention of water supply to individual homes is arguably the biggest transformation humanity has ever seen. We just don’t see its effects because, in today’s day and age, clean water supply is taken for granted. If all the AI models disappeared tomorrow, your life would change a little bit or a lot, depending on your profession. If water supply were yanked from your residence however, your life would no longer be the same. The list of inventions that have had an equally large transformative effect on humans is long and beyond the scope of this writing.
What I do find interesting, though, is going over some of the key impacts of AI, as well as where I think we are headed. Firstly, much like any other technological invention before it, I think artificial intelligence will follow an S curve. We are already in the midst of rapid adoption of AI, aka the initial parts of the S-curve. There will come a time, however, when AI models will be pervasive across industries, devices, and even sectors that are slow to adopt AI, such as manufacturing. These sectors will have these new technologies deeply entrenched, both in software and hardware. We can easily take a cue from past inventions. Think of the automobile. Gas-powered vehicles were rarely seen in the early 20th century, but by the middle of the century, they were everywhere, not just in the developed world. It took less than 50 years for the automobile S-curve to hit a plateau.
The same applies to packaged food and drinks that you find in a typical grocery aisle today. None of that existed, or most of it did not, around World War II, but by the end of the century, packaged food and drinks were pretty much everywhere, irrespective of country and continent. To be clear, for a technology to hit a plateau is not necessarily bad. It just means the technology is being widely used, and thus every segment of society is reaping its benefits.
Secondly, it is very clear to me that there is going to be a major wave of consolidation in the AI space. Hundreds, if not thousands, of AI businesses are being launched every month across the globe. Very few of them will survive. Most will simply shut down because they will not be able to generate enough revenue or profit to be sustainable. Some will be bought out by the big players.
In the end, we will be left with the top 3 to 5 major players in the space, who will not just be busy creating frontier models but also will be focused on their application. Some of the current big tech companies certainly stand a chance of being among the survivors. Most industries have the space for a handful of leading companies in a given sector, and the AI sector will be no different. Most of the companies in the AI space will be focusing their efforts on applied AI because the models themselves will become commodities and will be available on most devices around us.
An interesting outcome of this wave of consolidation will be the urgent need for CEOs to deliver real, demonstrable profits from the use of AI. Venture capitalists’ patience would run out over time, and the generous use of chatbots like ChatGPT would be severely curtailed. This is also the time when regular businesses that use large language models to produce software code, do accounting tasks, and so on will face a 3x to 4x bump in token costs from leading frontier models. It is an established business practice for a new kind of product or service offering to first capture the market and, once people get used to it, jack up the prices. Leading frontier models will be no different. As a result super cheap models will replace some of the frontier models for inference, which will be very specific to the subsector where they will be used.
Thirdly, the current GDP growth, especially in the U.S., which is being led by data center development, is temporary. Over the next 2 to 3 years, most of the data center development would be completed, and the tailwind the economy has been getting from this massive construction blitz will no longer be there. In fact, some of the data centers will become ghost towns, and there will be a glut of construction sitting idle. A few builders will go underwater, while some others will seize the opportunity to buy additional capacity.
There are a lot of fear and scare stories around the evolution of AI, but I am more optimistic. I believe that the fear that AI will eliminate humanity is overblown. Over the 2 million years of existence of our genus we have seen much change, and our survival instinct has been tested time and again. I cannot believe that software or even software-powered robots have the ability to rule us. The most that software-driven robots can do is perhaps drive us back to the savannah, from where we came, but even that sounds too far-fetched.
I’m also much more sanguine about the job situation for future generations than most fearmongering media outlets purport. There have been innumerable technological inventions that disrupted existing technologies. It is true that the generation that is in their prime working age bears the biggest and most immediate brunt of this change. However, future generations grow up native to those new technologies, and before long, those technologies are no longer new. This is true of the assembly line replacing the handmade goods industry, the automobile replacing the horse buggy , and the aircraft replacing the ship. The current generation is the one that is undergoing and will have to undergo the most pain and change. Future generations, however, will grow up to an entirely different set of jobs as new kinds of devices, product offerings, and services get rolled out powered by AI.
AI is certainly here to stay. As adaptive as the human species is, I am sure both current and future generations will adapt well to this new idea. While the journey will have its own pitfalls, the human race will emerge better on the other side.