>If so, they're acting on a gigantic assumption Yes the initial hypothesis that GPT-4 would know was a gigantic assumption. But a falsifiable one which we can

>If so, they're acting on a gigantic assumption Yes the initial hypothesis that GPT-4 would know was a gigantic assumption. But a falsifiable one which we can easily generate reproducible tests for. The idea that simulated neurons could learn anything useful at all was once a gigantic assumption too.