It's one of those things that is poorly defined and therefore conveniently perpetually a bit further out because we can keep on moving the goal posts. But we're

It's one of those things that is poorly defined and therefore conveniently perpetually a bit further out because we can keep on moving the goal posts. But we're at least being forced to do that now that we have chat gpt. If you look at it from a "this is way better than what science fiction predicted", we're getting there. Hal 9000 is basically science fact now. That arguably happened about five/six years ago when MS had to pull the plug on tay bot which (like hal 9000) still had a few unfortunate glitches but wasn't a bad effort. Arthur C Clarke was only off by about a quarter century. Arguably, chat gpt is a bit better conversationally than hal 9000. I wouldn't necessarily let it control my space ship yet. But then, those don't require a lot of AI and are mostly flown by computers at this point. We've explored most of the solar system without a single human going much further than the moon. The main issue is perhaps that it is confidently wrong some of the time. Which is true of humans as well, of course. Chat GPT wired together with state of the art speech recognition and generation would be pretty doable at this point. It would be expensive to run though. But it would sound better and be able to have some coherent conversations on all sorts of topics. I find it interesting that openai is choosing to not demonstrate that. It must be very tempting to do that since they also have decent solutions for speech synthesis and recognition. That would instantly make things like Siri, Alexa, and "Hey Google" historical curiosities.