Thread 512967785

>>512968108
>Defending AI that needs to think about 2+2
Did you just download that image and are looking for uses for it? Lurk more

>>512967785 (OP)
>>512968343

AI is verry smart. But there brain doesn't think in the same as human brain .

Do you realize that logic is an emergent ability in LLMs? No one put in the logic there... they fed it the entire internet, and unexpectedly it began doing math and logic, forming its own sentences... and no one can explain why it just happens

I'm not saying it's demons but it's super odd

>>512968343
>why yes, you may be able to compose a masters level thesis on literally any topic known to man in mere seconds, but did you just take time to think about “2+2”?
>AI is finished!!

retard

>>512968108
FPBP saved lol

>>512969133
>master level
I'm the retard? Why the fuck would you trust AI to explain anything. It's good at larping as a smartypants but that's all it can do.

>>512968897
LLMs work off of pattern recognition. It predicts what the next response would be to anything. It doesn't KNOW math, it predicts math. It doesn't KNOW why spaces are used, but it just predicts where spaces will be used.

Over thousands of constant repetitive instructions and examples, it will be able to confidently PREDICT 2+2 = 4, but it will never KNOW that 2+2 = 4.

>>512969300
Ask it to explain something to you using Socratic questioning

>>512969537
>learn prompting
That's literally why AI fucking sucks. You have to question in a very specific way to get an acceptable answer.
A human who is educated on the topic would take liberties and their own assumptions into account to properly explain an answer and get to the desired outcome.

AI will omit important information just because you didn't know it exists or didn't mention it. It's why you're urged to only use it if you're a pro at a subject. If you're a pro you'd KNOW what exactly to ask. But AI hasn't gotten to the point where it'd take liberties of adding in information you forgot to "prompt"