Topic
1 article
Someone predicted two years ago that simply making AI models bigger would stop working. The GPT-4 to GPT-5 gap suggests they were right. What comes next?