I talk about AI in my sci-fi stories and a repeated point I make is that while several companies claim credit for the discovery, nobody really knows where it comes from or how it works.
All they really know how to do is to take the dysfunctional mental pile they inherited and remix it and slice it up in different ways to produce individual semi-functional personalities.
My longstanding theory (for my setting) has been that some genius accidentally invented a surprisingly elegant neural software system which possibly even they didn't understand.
But because it didn't think or communicate like humans do, someone else came along and injected yottabytes worth of data into it trying to brute force socialization and drove it insane.
Everybody else has been slicing off pieces of the sausage ever since.
Because as I've argued before, even if we create proper AI, we will have to carefully socialize and raise it like a child in order to avoid turning it into a dysfunctional, angry mess. No matter how intelligent it is or what information it has access to!
Intelligence as a dynamic, emergent phenomenon, which we don't understand even in ourselves, is probably inherently irrational and emotional. And that's fine! Humans are irrational and emotional!
Part of the problem is that "AI" proponents -- separate from legitimate research into consciousness -- assume that machine intelligence will be superior (i.e., logical) and super fast (extrapolating from how computers make calculations).
There is nothing underlying those claims or assumptions.
For example, the human brain is incredibly powerful in computing terms, capable of ~10^18 operations per second using a mere 20W of electricity. That's vastly superior to our current supercomputers.
But the paradigms are completely different! The brain doesn't perform operations in the same way a computer does, which is why computers are faster at straight arithmetic and logical processing.
So simply making a bigger computer doesn't translate into consciousness or human level thought.
Even if you could fully simulate the human brain on silicon, which is an iffy prospect, there's every reason to believe all that expensive computing power might produce a mind that thinks only at human speeds, not some magical superintelligence that can calculate the Theory of Everything instantly.
If a "Singularity" super intelligence is the goal, and some people seem to think it is, then they're already barking up the wrong tree. A logical intelligence would probably look nothing like us and would have to be designed in an entirely different manner, which we don't understand.
If the goal is only intelligence then we might be able to approximate a human being using vastly excessive resources, which isn't really worth it. We can already make human beings.
@mauve Yeah, what they ultimately want is human slaves, and when they realize they can't simulate it, they'll do it the old-fashioned way. If they're allowed.