If you train it on the internet, it will be almost as reliable as the internet. (Almost. Not quite.) And if you train it to always give a responsive answer (and not say "I don't know") then it's going to generate hallucinations even when the internet doesn't lead it astray.

OK. Basically an AI can be trained to do anything that a person can do **IF THERE IS A WAY TO CHECK THE ANSWER**. If the AI can't get good feedback, it can't learn to handle the problem correctly.

This is why I think that AGI is going to require embodiment, and LOTS of sensors. And it still often won't reach the same conclusions people reach. (Consider the arguments that even men and women have, and their goals/sensoria/brains are extremely similar.)

The thing is, LLM is only one type of AI. All it does is try to handle language. Train it differently and it's not a LLM, but it can do quite well at, say, protein folding. (Note it's guessing how the protein would fold, but that used to be the domain of experts. After you get the guess, then you go and check, and the AI is doing as well, or possibly better, than the experts were doing.)

I normally find most of the check out lines unmanned. Perhaps it depends on the time of day.

But this was going to be LOTS of stories tall, so you'd still get quadratic scaling.

Do you mean the big island, or the state?

It's not really clear, because there are LOTS of "very special and limited applications". Consider what the automated check out stations at supermarkets did to the checkout clerks.

OTOH, it's also probably largely lies. But you can't scrap the "probably largely", and those are pretty vague.

It *could* come out of thin air, using solar cells or wind power. And it's inefficient enough (i.e. 1 gallon/day) that you might as well. It might be useful in a remote cabin with a motor/generator set that you rarely needed.

Actually, gasoline *is* just the short chain hydrocarbons. But you need the other stuff to make it a fuel for modern cars. (The "s" is on "hydrocarbons" because it's a mix of different molecules.)

At 1 gallon/day it doesn't seem like it would justify itself much of anywhere. Perhaps the same principles could be applied to a different design, This looks like a lab curiosity.

No. Apple was less open with it's documentation that the preceding personal computers. (The S100 bus computers.) I'll admit that I believe the Tarbell interface was portable (I never tried), but that wasn't invented by Apple. The floppy disk format was proprietary.

That's lousy reasoning. Being wealthy is not, in and of itself, sufficient grounds to justify an attack. Come up with a better reason. (There are a few that I could think of, and probably several that I haven't.)

OTOH, extortion is really difficult to justify. If they want to be justified, they should just release the data publicly, possibly after sanitizing it a bit.

If this is the case I remember from a few months ago, then no, it didn't do what he told it to do. One of the basic instructions was "don't change anything in my folders". (That's an extreme paraphrase, as I read about it several months ago. I think on Slashdot. But it's the essence of the original instructions.)

OTOH, I read about it several months ago, and it had happened several months previous to that. So it was one of the really early LLM chatBots. And notice that he wasn't a computer technologist, so one shouldn't have really expected him to have made his own proper backups. (But perhaps his department should have.)

Not this year. He's nearly 90, and short term memory declines as you approach that old.

That's what they say publically, but just how true is it? I don't think anyone knows. (And "true" doesn't imply that the AI could actually replace the folk, merely that management really thought that it could.)