What just happened, what is happening next

The current best estimates of the rate of improvement in Large Language Models show capabilities doubling ever 5 to 14 months. This incredibly rapid pace was on my mind as I put the finishing touches on my book (available at every major bookstore and now, with its bestseller status, apparently discounted 33% at Amazon) back at the end of December. Given that I knew it would release in April, I wrote it in a way that would be relatively timeless, but I also had to make a couple forecasts on where the future would be heading. Fortunately, I think I predicted correctly, and, as a result, the book still encompasses the current state-of-the-art. But that doesn’t mean that nothing has changed in the last four months. So, I wanted to go over the last few months of AI and take stock of what we have learned about what AI models can do, and especially, what AI might do soon. At the time I finished the book, there was one frontier model, GPT-4. Now have three GPT-4 class models, all of roughly si

The current best estimates of the rate of improvement in Large Language Models show capabilities doubling ever 5 to 14 months. This incredibly rapid pace was on my mind as I put the finishing touches on my book (available at every major bookstore and now, with its bestseller status, apparently discounted 33% at Amazon) back at the end of December. Given that I knew it would release in April, I wrote it in a way that would be relatively timeless, but I also had to make a couple forecasts on where the future would be heading. Fortunately, I think I predicted correctly, and, as a result, the book