Sandhini Agarwal: We’ve plenty of subsequent steps. I positively suppose how viral ChatGPT has gotten has made plenty of points that we knew existed actually bubble up and turn into crucial—issues we need to resolve as quickly as attainable. Like, we all know the mannequin remains to be very biased. And sure, ChatGPT is excellent at refusing dangerous requests, however it’s additionally fairly straightforward to put in writing prompts that make it not refuse what we needed it to refuse.
Liam Fedus: It’s been thrilling to look at the various and artistic purposes from customers, however we’re at all times centered on areas to enhance upon. We expect that by way of an iterative course of the place we deploy, get suggestions, and refine, we will produce essentially the most aligned and succesful expertise. As our expertise evolves, new points inevitably emerge.
Sandhini Agarwal: Within the weeks after launch, we checked out among the most horrible examples that individuals had discovered, the worst issues folks had been seeing within the wild. We form of assessed every of them and talked about how we must always repair it.
Jan Leike: Typically it’s one thing that’s gone viral on Twitter, however we have now some individuals who really attain out quietly.
Sandhini Agarwal: A whole lot of issues that we discovered had been jailbreaks, which is certainly an issue we have to repair. However as a result of customers must attempt these convoluted strategies to get the mannequin to say one thing dangerous, it isn’t like this was one thing that we utterly missed, or one thing that was very stunning for us. Nonetheless, that’s one thing we’re actively engaged on proper now. Once we discover jailbreaks, we add them to our coaching and testing information. All the information that we’re seeing feeds right into a future mannequin.
Jan Leike: Each time we have now a greater mannequin, we need to put it out and take a look at it. We’re very optimistic that some focused adversarial coaching can enhance the scenario with jailbreaking lots. It’s not clear whether or not these issues will go away solely, however we expect we will make plenty of the jailbreaking much more tough. Once more, it’s not like we didn’t know that jailbreaking was attainable earlier than the discharge. I believe it’s very tough to essentially anticipate what the actual security issues are going to be with these techniques when you’ve deployed them. So we’re placing plenty of emphasis on monitoring what persons are utilizing the system for, seeing what occurs, after which reacting to that. This isn’t to say that we shouldn’t proactively mitigate security issues after we do anticipate them. However yeah, it is rather arduous to foresee all the things that may really occur when a system hits the actual world.
In January, Microsoft revealed Bing Chat, a search chatbot that many assume to be a model of OpenAI’s formally unannounced GPT-4. (OpenAI says: “Bing is powered by considered one of our next-generation fashions that Microsoft personalized particularly for search. It incorporates developments from ChatGPT and GPT-3.5.”) Using chatbots by tech giants with multibillion-dollar reputations to guard creates new challenges for these tasked with constructing the underlying fashions.
