Something that strikes me as really weird about the AI age that we're in is we all know we've entered it, but no one was aware when we entered it.
It was like a time jump.
Two years ago we were mocking AI videos for their crappiness, speculating that AI may or may not be all they're saying it's cracked up to be.
And now, we're in solid "AI is terrifying and I can't imagine the dystopia someone born today is going to grow up in regarding the all encompassing nature of AI"
And it's like having a black out when drunk. One minute you're in a bar. The next minute you're behind bars for public intoxication and you didn't remember anything in between.
That's how it feels with the AI thing.
Of course we always had trepidations from early on, but unlike certain ages, the age of woke, the age of covidians, the age of seeing the extent of the brainwashing, there were solid markers of "we were there, and now we're here and here's what led to that."
But with AI, I don't know if it's similar to you, but it was like a realtime mandella effect where suddenly we're in the AI age and it felt just like yesterday we were mocking the "Will Smith eats spaghetti" and there doesn't seem to be a clear moment where everyone took AI for the serious existential threat that it is....it just happened, like a black out from drinking. It's a very strange feeling.
No, I was paying attention to the AI space pretty closely since 2015 - 2016 ish. The progress is really straightforward and mostly predictable if you're familiar. Chat GPT caught a lot of people off guard, but GPT-2 had been creating surreal somewhat comprehensible stories for a while. The big breakthroughs of getting the AI to go the other way and generate generalized output had predictable consequences, and rapid development happened in the space that opened up without software to fill it. Things iterated rapidly, frameworks were created and discarded rapidly as we made quick efficiency improvements and picked off low hanging fruit.
'AI Winter' ends, AI Spring begins and rapid progress is made in the structural level. Now we begin to enter an AI Summer where we have to figure out how to convert all these gains into something productive.
You make some really good points.
IMHO, AI gains largely follow gains in computing capacity. The more computer power we get, the more powerful AI can be. If--or when--we hit the point that AI is able to start truly writing creative and novel algorithms to extend AI programming, that's when things get really interesting or scary.
Most of the AI techniques that are in common usage today have been basically been the result of decades of linear progress in AI sciences.
Perceptrons go back to the 1950s!!! The foundational ideas of neural network computing go back a decade or two before.
Capacity and Framework. They alternate. This time was a framework bottleneck that was broken by the development of transformers.
Now we find capacity lagging which is why Computer hardware has spiked.
"Novel" is a word with shifting meaning. AI is always the domain of the novel. In 1997 Deep Blue defeating Kasparov was cutting edge AI. Now it's just a computer chess algorithm.
The bar for "truely creative" is already rising, and is reaching the point where we will start to have to dismiss a large number of humans as incapable of creativity.
Not that what you're saying is incorrect, this is just a domain where definitions shift as State of the Art does, and you need to pay attention to how if you want an accurate perspective.
I know that, and you know that. But good luck getting the normies to admit that. It punches a lot of holes it what the common culture believes, or is at least taught, so admitting it is inconvenient enough to be painful on a psychic level for them.
I recently read about a billion-dollar Ai playing chess against a TRS-80 (or something from that era) chess program. Ai lost every match, it never learned anything.