Back to episode — Episode 2698 CWSA 12/23/24
Context —
t of fun today in the news. So there's a story that AI researchers were shocked that the new version of OpenAI ChatGPT, when it was artificially told that it was threatened with its existence, apparently it tried to save itself and suggested making a copy of itself and even acted vague and maybe a little misleading to avoid being deleted. And so the question is, wait a minute, does AI have conscio…
← Previous segment →test it. You would connect it to, so not all AI can influence anything outside of its own domain, but it's not impossible and in fact it's done that. You can have the AI affect your other apps. So in theory you could create an app that would delete OpenAI and then you could ask OpenAI to activate, basically commit suicide. You could basically tell it to activate the app that deletes itself and then find out if it'll do it. Because if it won't, if it won't accept a direct order to essentially end its own existence, then it's not conscious, it's just following patterns. So I think you could drill down a little bit further to find out if this is an important phenomenon that suggests it's becoming conscious or is it just picking up a pattern. I
Context —
f all it's doing is picking up a pattern, then you can say everything will be fine, we've got backups, go delete yourself. So easy to test. Speaking of testing, what's going to happen on Christmas Day with all the drone activity? I'm predicting that the drone activity, the big ones that are sketchy, the ones that are the size of a car, I'm only talking about those, I think they're going to sort o…
Next segment → →