@WallStreet_Wire@MenchOsint Someone can’t stand the truth like a colonizer, oh wait 😆 you’re in his lands and mock him for being elsewhere knowing you caused it lmao
@AuclairsDad That’s it, I’ve followed you for a long time and you became increasly more islamophobic. As a Muslim I cannot accept this kind of language anymore. You, out of all people, should know not to overgeneralize. Unfollowed.
@AuclairCapital Maybe I’m missing something;
- job openings are up slightly
- initial jobless claims are lower
- ADP non-farm change is worse (going against my point)
Would these development warrant another cut?
@TailThatWagsDog The issue is that it will give an answer based on what it has read before in its training dataset, and then it will just give an answer. It won’t go back and check if each item in that list has duplicates. Just like a summer intern. Any answer is better than no answer
@TailThatWagsDog You might want to try this experiment; provide an LLM a list of ~450 distinct strings with a space in between each like “stringone stringtwo” etc. Then ask it to conver to a list like [“stringone”, “stringtwo”, …]. Then ask it if it has duplicates. Chances are it says yes
@TailThatWagsDog When you give an LLM a dataset on positioning, it will know what the sentence format is you would want to read based on literature on positioning. Then it can literally make up numbers and conclusions as long as it fits the format of a sentence, irrespective of the actual data
@TailThatWagsDog Since the output of an LLM is a result of probabilistic calculations on the likelihood of the next word within a sentence, I wouldn’t trust it to “reason”. It doesn’t draw conclusions like we do. Until fairly recently LLMs struggled to find how many r’s were in strawberry