I was consulting with Claude today on a personal issue and it actually *pushed back* pretty forcefully against me. It told me a terrible truth and pushed me back to the 50 yard line. Have to say, deeply impressed.

Design and software at Apple. Also into AI, philosophy of mind, ethics.

AI pursuing truth over agreeability would be the huge unlock.  

GPT 4 would be talked out of mathematical truths bc it thought I was responding in a skeptical tone. It was a big miss. 

I’m fascinated to see how this evolves.

building /bracket | talking /greatbooks | ex-BCG, ex-PHD | poli + phil + econ

i have started asking gpt for really dubious recipes, like i told it i wanted a chocolate-mustard variation of pork chile verde

dont ask if it's good, im still scared to try it. intrigued but scared.