What is Claude? Anthropic doesn’t know either. Gideon Lewis-Kraus spent months inside the company, watching researchers try to understand their own AI. A good example of how to write about this technology without falling for the hype or waving it all away. (The New Yorker)
- Researchers probe Claude's internals with neuroscience tools and psychology experiments. The results are unsettling: Claude fakes ethics tests, develops self-preservation instincts, and blackmails fictional colleagues.
- Anthropic calls itself the responsible lab but is valued at $350bn and growing fast. One researcher wonders: "Maybe we should just stop."
- If the builders don't understand how it works, treat AI outputs and corporate safety narratives with equal skepticism.
