cm0002Feb 21

AI advice

https://lemy.lol/post/61475626

Show threadHackworth
Feb 21 at 5:16pmWeb
Show threadmushroommunkFeb 21
The point isn’t that some models are better than others. The point is that yet again it’s an example that LLMs are not thinking machines and you can’t trust anything from them and people are burning the world to run a glorified auto complete.
Show threadHackworthFeb 21
My point was that some models are better than others.
Show threadmushroommunkFeb 21
Sure, fine, some get this right, and what else are they getting wrong? Something more serious and harder to spot?
Show threadHackworthFeb 21
I agree that we should never treat these things as oracles. But how often they’re right/wrong does matter.
Show threadGunriggerFeb 21

how often they’re right/wrong does matter.

That’s the wildest take I’ve heard on the question answering machine.

Show threadGrimyFeb 21
Most people get their info from forums and blog posts. Unless you limit yourself to nothing but peer reviewed papers, you probably do some kind of calculation on the legitimacy of whatever source you are perusing.
Show threadDemdaruFeb 21

Counterpoint: People are not thinking machines and you can’t trust anything from them and people are burning the world to run glorified slave labor.

Truly we are AI of natural world xD

Show threadkopasz7Feb 22

People can only make stupid mistakes so many times. Once exited the gene pool, that’s it. Meanwhile an AI can spew statistical nonsense 24/7 without repercussion.

I trust an intelligence way more that managed to keep iteself alive, than one that is optimized to generate signal shaped noise.

Show threadCethinFeb 22
People are thinking machines. The problem is, we aren’t a collective thinking machine. People thinking in their own self interest have caused most of the problems. It makes perfectly rational sense to burn the world if you only care about the quality of your own life.
Show threadAceFeb 22

For me, both Sonnet and Haiku gave the right answer:

i.ibb.co/m5mTQQyn/image.png

i.ibb.co/jvcZx7ZN/image.png

chatgpt fails though:

i.ibb.co/9kTVFSxp/image.png

gemini “fast” waffles a lot but does eventually give the right answer:

i.ibb.co/84KR0s94/image.png