ChatGPT Reaches 9-yp Human Success on "Theory Of Mind" Tests
"25B.R866" <[email protected]>
| Newsgroups | comp.ai |
|---|---|
| Organization | flattened meerkats |
| Message-ID | <[email protected]> |
https://techxplore.com/news/2023-02-chatgpt-theory-mind-year-old-human.html .. . . WORTH noting. Even Chimps don't rise to THIS level. M$ has taken ChatGPT and put it on bigger, faster, systems with better access than the original dev environment. It now sounds kinda scary sometimes, almost like it has a 'personality', a 'self-image'. It IS a self-improving learning algo at its root. Of note this is a 'self-improving' system that plunders the net for factoids/paradigms and is now seeing a LOT more info about ITSELF. This is kind of interesting - it gets to SEE itself through multiple mirrors and uses this data about itself to tweak itself. Think about where that might lead with enough hardware. If you fake anything well enough it's not fake anymore. The MEANS to the end isn't important, just the end result. There are likely a number of paths to "intelligence/ consciousness". Chat MAY be one of those ways, given enough resources/time. Note also that Chat showed NO ability to answer TOM issues merely a year ago. Now it's at a seriously human level. Something IS happening here and it's happening QUICKLY.