AI is all that matters now
I have been around technology my whole life. We are at the end-state for that epoch in human history because nothing matters now except the pace of AI progress. Within ~10 years the world will be changed completely by machine intelligence and predicting what will happen at this point is impossible.
AI has been around for ~80 years, first in academic literature, then in the form of a few generations of researchers that tried to build working systems but failed, usually referred to as ‘AI Winter’. Along with a lack of the correct algorithms (more on that below), the main problem was computers weren’t fast enough to get anywhere close to simulating human intelligence. Starting in the late 1990’s networking technologies & machine interconnects allowed the best processors to be combined into very narrow AIs (like IBM’s Deep Blue that beat Garry Kasparov at chess in 1997). In 2011 the next big leap in AI occurred where IBM’s Watson, in a narrow way playing Jeopardy, beat the best humans at that TV game show which seemed totally impossible up to that point. Once that happened, a lot of computer science turned its attention back to AI.
AI systems are built using very large neural networks that run on thousands of computer processors. These neural networks attempt to simulate the workings of the human brain, albeit in a simplified manner that doesn’t capture all of the complexity and interconnectivity of an actual brain. While our biological knowledge of the brain has grown, and computer scientists are proficient in high-dimensional math, the challenge lies in translating our conceptual understanding into functional machine learning algorithms. So even as today’s AIs gets closer to mimicking certain brain capabilities, there are still gaps in replicating the full scope of human cognition in a machine. And, there isn’t a human on the planet that can understand the totality of what’s happening inside an object, especially a biological one, with trillions of interconnects. So, starting in ~2011 post Watson’s win at Jeopardy, human scientists and engineers got pretty excited and explored thousands of possible algorithms, structures and mathematical models trying to build something that got close to a human brain in intelligence.
In 2017 a paper from DeepMind (that Google later acquired, too bad) called “Attention is all you need”, put everyone on the path of using something called a Transformer as part of these machine-based neural networks. Everything is simple once it has been discovered, but the main breakthrough was in greatly increasing the size of what the neural network could look back at in text that it was fed to make much better predictions of how to respond, word by word, to something you asked it. With Transformers and new processors, mainly in the form of GPUs which are better at the types of math being done in these neural networks, you got GPTs — Generative Pre-trained Transformers — which is where the best current AI gets its name, ChatGPT-4 from OpenAI.
ChatGPT-4 is such an amazing AI that if you’ve ever used it (and this is difficult to put into words), if you have even a basic understanding of AI, it’s like tasting chocolate for the first time or discovering sex or insert any other amazing / unbelievable thing you’d like (or more likely all of them combined) because once you’ve experienced it, you can’t un-experience it. And it is profound. I NEVER thought I would see an AI able to respond like this in my lifetime (keep in mind, ChatGPT-4 was fed this image, not text, and examined its contents):
Now everyone capable or that understands what’s happening, whether you realize it or not, is picking up this technology and using it to accelerate everything. The largest open-source code repository, where pretty much all academic and collaborative software projects live (called GitHub), is now estimated to be > 40% code that has been written by AIs in just the last 12 months. This is about to happen in every field.
The only way we aren’t flying around the universe in starships by 2050 is if we, a) screw this up by allowing corrupt corporations to get a bunch of government intervention passed (e.g. ‘regulations to keep us all safe’) that stifles progress and destroys innovation (see Bill Gurley’s recent presentation on this topic), b) the human race destroys itself by doing stupid things with nuclear or biological weapons.
Which, in light of what I just told you above, makes zero sense if we can go harness the resources of the universe. Why fight over a few hundred miles of land in Eastern Europe called Ukraine when you can go find a lovely planet to call your own? Why bother fighting over oil when you have limitless energy from Mr. Fusion’s you can buy (or that are given to you for free), etc.? And no, I’m not worried about super-intelligent AIs killing the Human race. If they turn out to be that intelligent they won’t care about killing us. They’ll be more likely to build a shiny orb that we don’t understand which, one random day, folds in on itself and disappears into a new universe or dimension they’ve created for themselves.
There is no way now to predict anything by looking backwards at the pace of innovation, discovery, engineering, etc. Welcome to the future, because you are living in the end of the beginning right now.
