It's amazing how supposedly smart people can drink their own Kool-Aid.
The sad fact is that we haven't made any substantial progress on strong AI - that is, an AI with which you can have a conversation where it can "reason" about what you said to the point it can answer questions about it - in fifty years.
Don't get me wrong - machine learning (ML) is a valuable technique that has done some amazing things! But it hasn't brought us one step closer to a machine that can reason about individual human utterances - something tiny children do with ease.
What we have instead are things like GPT-3 - a statistical model boiled down from an immense corpus of things that people have said, a model that can spit out a statistically likely response to texts - but there's no attempt to model any sort of meaning or symbolism behind what you say.
Consider this pair: "I have some eggs. Do I have any eggs?" and "I have no eggs. Do I have any eggs?" There isn't any AI that can even answer bone simple questions like this, let alone more complex questions like, "I hate eggs. Do I have any eggs?" (A good answer, "Probably not.")
ML can find patterns that a human intelligence couldn't, but it intrinsically has huge gaps. It has no "state" that changes when you interrogate it. It cannot explain why it came to any decision. It cannot learn new things on-the-fly and it cannot learn from a single example. If you learn that your friend Steve got married, you don't need to be told this more than once, but ML requires a huge corpus containing each fact repeated in many different ways to "learn it".
Today, nearly all the AI funding is going into machine learning. We're getting good results, but we're also reaching limits in this field quite fast (computing can't fix bad data quality, diminishing returns set in, etc) but more, there is no apparent path from machine learning to strong AI, for the reasons I mentioned above.
Am I claiming that "strong AI" like this is impossible? Heck, no. We know computers powerful enough to be able to process symbols well exist - look at our brains! It's certainly possible.
But I'm completely sure we won't see an AI a "billion times smarter than us" by 2045. I'm fairly sure we won't see an AI as smart as your average 7-year-old by then. It's also possible that we will never see a strong AI - that we'll run out of resources and have to retool our whole society before that happens.
(Another possibility is that no human or team of humans is smart enough to create strong AI. I tend to suspect that's not really the case, but I don't have a logical reason to rule it out.)
