So is there going to be super AI or not? And by super AI I mean AI that is at least as smart or smarter than the smartest human. So an AI that could make scientific breakthroughs, and useful inventions, and most other things. I won't say that it would have to be the best at everything. Maybe a human would be better at detecting sarcasm. Just most thinga.
>>107996108>AI that could make scientific breakthroughs, and useful inventions, and most other things.Without being directed by human experts? Not anytime soon.
Retarded ai poster
>>107996108>super AI as in artificial super intelligence?the super monkey paw machine that will rape you to death in order to get closer to its terminal goal?the thing that would make me gladly collaborate with CIA and mossad just to destroy it even though i usually prefer to run them over with a car?that thing?
>>107996108Possibly, but it's not gonna happen through any sort of language model.
>>107996108Probably not. Its all empty hype that retards are pushing out. It may be possible but not something we are capable of right now. Though I could totally see humanity wiping itself out by giving an LLM access to something it shouldn't and we all get killed because of it.
>>107996108I believe they're going to make something retarded but knowledgeable, hail it as ASI then they entrust the entire world to it and let it fuck everything up immedieately in an idiocracy like fashion.
>>107996372>any sort of language modeI think it likely parts of LLM will be used in future more successful AGI attempts but I agree LLMs won't reach AGI alone.
Ai has already solved a math proof. Thats kind of a big deal.
>>107996414That's what they are already doing.Search is getting worse, junior jobs replaced with AI will make it so future competency isn't there so the future will be fucked.
>>107996108>will there be a glorified search engine that is more intelligent than the data it was trained on?I don't think you know what "AI" is
>>107996108Don't be ridiculous, predicting a likely next word in a sequence, or what the next pixels should be is already a pretty great breakthrough, but the technology is fundamentally limited, there is no general ai or super smart ai on the horizon, not even close
>>107996108No. I figured out a while ago that these models never have a 'genius' sudden insight into NP (or above) hard problems. It's something that happens occasionally, when a human somehow finds a better solution to say a specific type of hard problem (for example routing in circuits) that we didnt find with a massive cluster of computers algorithmically working on the problem for months. I have never seen that sort of 'deep thinking' from a transformer. It's somehow related to the neural arch of the brain
