just finished listening to that deborah o'malley pod and it's pretty blunt. she's saying we need to stop testing headlines and start testing actual concepts instead. using chatgpt to simulate personas or brainstorm variations sounds fine on paper, but i don't know if it'll actually work for high-intent users. the idea that ai helps us move away from 'interesting' results toward things that actually move the needle is
pretty much the whole point of why we do this in the first place. still feels like a lot of hype about pattern recognition and scale though.
it's basically just faster brainstorming. how are you guys actually using llms for hypothesis generation without it being total garbage?
link:
https://conversionsciences.com/ai-in-experimentation/