OpenAI PULLS MODEL, HALTS TRAINING As Danger LOOMS

There's, as usual, a lot going on in the wild world of AI, and we're happy to be joined by Garrison Lovely, who is a journalist focused on these things, and author of the new book that is actually out now. It is called Obsolete: The AI Industries's Trillion Race to Replace Us and How to Stop It. This book is

literally flying off the shelves. Um, Garrison, we will put a link in the description of the best place to find it at this point because it's actually out of stock in a number of places. So sign of how much interest there is in the topic in general and this book in particular. So congrats and great to see

you Garrison. >> Thank you. And we are working on getting more copies to everybody ASAP. >> I'm sure that you are. I'm sure that your people are certainly working on that. Um we want to get to some of the specifics of the book which focuses on the threat to as the title implies make

human labor obsolete. But I wanted to start with some of the latest news here with regard to open AI. If we can put C1B up on the screen, this is a pretty drastic move that they are making. They are scrapping the release of their new AI model over safety concerns. Um, they said they were having all sorts of

issues. They didn't get into too much specifics around this latest model. Of course, we know a lot of the safety incidents, not all, but a lot of the safety incidents in recent months have originated from OpenAI. And in addition, Sam Alman has announced that they have not only scrapped this release, they have stopped training new models

internally at all. So, um, what is going on here, Garrison? Are we getting the full story about what is happening internally at OpenAI? >> Yeah, I mean, so the big picture is that OpenAI has been racing even harder than usual to make models more and more capable. And they're using this technique called reinforcement learning