3 ms·
They have a philosophy which may be wrong but is not entirely incoherent (expose risk as capacity becomes available, so we aren't caught off-guard by high-capac
by justrealist 3y ago
They have a philosophy which may be wrong but is not entirely incoherent (expose risk as capacity becomes available, so we aren't caught off-guard by high-capacity risks we can't recover from).
And let's be honest, GPT-4 is vastly better aligned than any of the other LLMs. Not a super high bar, but it's a bar.
- JohnFen 3y ago> GPT-4 is vastly better aligned than any of the other LLMs. If I'm being completely honest, I can't judge that at all, because it's far from clear to me what, specifically, people mean by "alignment" when they use that term. And I'm pretty sure that different people are meaning different things.