The company says it has found a new window into how its models arrive at answers. We spoke with senior editor Will Douglas Heaven about it. Anthropic--currently the world's most valuable AI company, with a nearly $1 trillion valuation--has a reputation for publishing strange and heady research. It's looking into whether AI models can feel pain, for example, and will sometimes cut off chatbot conversations if it suspects users are "abusing" the model. One niche that Anthropic spends more time and money on than other AI companies is called mechanistic interpretability, which means looking inside the complex math of an AI model to learn why it comes up with one particular output and not another. It's complicated stuff; there are millions of data points that might contribute to any result, and wading through them can look more like word salad than anything useful.

For Bill Gates, artificial intelligence is the most important invention since the internet or the personal computer. For Google boss Sundar Pichai, it will have a more profound impact than electricity and fire. Already, though, there are signs the AI revolution may be faltering. Since OpenAI released its landmark GPT-4 system in March 2023, new large language models like Google's Gemini have offered only incremental improvements. GPT-5 could change this tomorrow, of course.

Everyone from chief executive officers to product managers to venture capitalists wants to understand machine learning better. They know it has the potential to take their software to the next level. They feel the excitement around it. They've read the TechCrunch or Fortune articles, and they've maybe even done a quick linear regression or two. But the primary issue that many product leaders grapple with when it comes to machine learning is that they want programs that don't just crunch the numbers, but can also think for them.