5 ms·
This headline and some of the topics within will generate a lot of predictable, less-than-illuminating discourse... but there is something interesting amidst it
by civilized 3y ago
This headline and some of the topics within will generate a lot of predictable, less-than-illuminating discourse... but there is something interesting amidst it all. It is becoming widely known how readily LLMs spew nonsense. I hope this spurs research into AI models that can self-critique, self-correct, and acknowledge uncertainty in their assertions. I believe the biggest breakthroughs are yet to come.
- bboygravity 3y agoI find it fascinating that LLM's are the first "creature" (in the French sense of the word) that can be so incredibly knowledgeable and smart-seeming and completely dumb at the same time.
- invig 3y agoSecond "creature". I don't see it being different from humans in this regard.
- albert_e 3y agoEnsemble of LLMs, each with a different expertise and bias and temperature. GenAI by committee.
- civilized 3y agoIt won't work. GPT-4 already does this and the improvement is nice but limited. Feel free to mark this tweet and come laugh at me in 10 years.