Dario Amodei: belief

Transcript context

…So this actually doesn’t apply. This isn’t just about Claude. I believe I’ve seen these complaints for every foundation model produced by a major company. People said this about GPT-4, they said it about GPT-4 Turbo. So a couple things. One, the actual weights of the model, the actual brain of the model, that does not change unless we introduce a new model. There are just a number of reasons why it would not make sense practically to be randomly substituting in new versions of the model. It’s difficult from an inference perspective and it’s actually hard to control all the consequences of changing the weights of the model. Let’s say you wanted to fine-tune the model, I don’t know, to say “certainly” less, which an old version of Sonnet used to do. You actually end up changing 100 things as well. So we have a whole process for it and we have a whole process for modifying the model. We do a bunch of testing on it. We do a bunch of user testing in early customers. So we both have never changed the weights of the model without telling anyone. And certainly, in the current setup, it would not make sense to do that. Now, there are a couple things that we do occasionally do. One is sometimes we run A/B tests, but those are typically very close to when a model is being released and for a very small fraction of time. So the day before the new Sonnet 3.5, I agree we should have had a better name. It’s clunky to refer to it. There were some comments from people that it’s gotten a lot better and that’s because a fraction we’re exposed to an A/B test for those one or two days. The other is that occasionally the system prompt will change. The system prompt can have some effects, although it’s unlikely to dumb down models, it’s unlikely to make them dumber. And we’ve seen that while these two things, which I’m listing to be very complete, happened quite infrequently, the complaints for us and for other model companies about the model change, the model isn’t good at this, the model got more censored, the model was dumbed down. Those complaints are constant and so I don’t want to say people are imagining it or anything, but the models are, for the most part, not changing. If I were to offer a theory, I think it actually relates to one of the things I said before, which is that models are very complex and have many aspects to them. And so often, if I ask the model a question, if I’m like, “Do task X” versus, “Can you do task X?” the model might respond in different ways. And so there are all kinds of subtle things that you can change about the way you interact with the model that can give you very different results. task X?” the model might respond in different ways. And so there are all kinds of subtle things that you can change about the way you interact with the model that can give you very different results. To be clear, this itself is like a failing by us and by the other model providers that the models are just often sensitive to small changes in wording. It’s yet another way in which the science of how these models work is very poorly developed. And so if I go to sleep one night and I was talking to the model in a certain way and I slightly changed the phrasing of how I talk to the model, I could get different results. So that’s one possible way. The other thing is, man, it’s just hard to quantify this stuff. It’s hard to quantify this stuff. I think people are very excited by new models when they come out and then as time goes on, they become very aware of their limitations. So that may be another effect, but that’s all a very long-winded way of saying for the most part, with some fairly narrow exceptions, the models are not changing. I think there is a psychological effect. You just start getting used to it, the baseline raises. When people who have first gotten Wi-Fi on airplanes, it’s amazing, magic.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.