Thought about this when I saw the thread below on the Tesla recall, because Tesla was an example in the interview. We’ve discussed here how AI is about to render so much human work obsolete. And that driverless vehicles are right on the horizon. Seeing this piece (I’m a fan of Ian Bremmer and the insights he and his guests frequently provide), I think we are not very close to either.

Computing power has improved massively. Storage has improved massively. So the ability to predict the next words in a string of words (like when you type a text and the program predicts the word you’re typing, but on steroids), can produce magical looking things. Also, lots of errors. Interestingly, nobody’s figured out how to get chatGPT to fact-check itself. To just…check against the internet. According to the interviewee, nobody will, anytime soon. We’re really no closer to that tech than we were 75 yrs ago. In fact, he thinks we’ll have to dust off the complex neural net technologies abandoned decades ago, because lo and behold, the answer likely lies there in making any AI truly “smart”. It’s basically illiterate.

The Tesla example - a car drove directly into a helicopter. Nobody thought to program a helicopter specifically into the self-driving app. And the car only “knows how” to avoid the things specifically shown to it. It can’t make inferences. And we’re nowhere near that level of machine smart. Possibly not even on the right technological pathway yet.

Maybe I’m just a boomer technophobe, but I actually found this somewhat comforting. Not, however, in its ability to make deep fakes and generally disseminate disinformation and divisiveness at never before seen rates. The interviewee definitely sees the tech well situated for that risk.




[Post edited by hoolstoptheheels at 12/13/2023 5:22PM]