deleted by author
pull down to refresh
11 sats \ 6 replies \ @optimism 14h
Better in ~Politics_And_Law. lol
reply
100 sats \ 5 replies \ @Bell_curve OP 14h
You are right... I will post there...
reply
100 sats \ 4 replies \ @Bell_curve OP 14h
I got so excited that Claude agreed with me because I'm always arguing with ChatGPT and Perplexity and sometimes Co-Pilot
reply
11 sats \ 3 replies \ @optimism 14h
Do they not agree?
reply
100 sats \ 2 replies \ @Bell_curve OP 14h
Perplexity is blatantly left wing. It’s like the mainstream media or Reddit
reply
11 sats \ 1 reply \ @optimism 14h
All these things are trained (aligned) to a certain response due to "safety". The only reason why alignment is safety is because it's emulation of intelligence, not real intelligence. So it makes sense that models that aren't trained with GPT as a judge of what is socially aligned (Claude, Mistral) respond differently.
You may like Venice's uncensored finetune of dolphin-mistral.
view all 1 replies