Researchers from Google, OpenAI, DeepMind, and Meta have joined forces to warn about the cognitive powers of what they're building — and it's a pretty bizarre state of affairs indeed. In a new position paper, 40 researchers spread across those four companies called for more investigation of so-called "chains-of-thought" or CoTs, the "thinking out loud" process that advanced reasoning models exhibit when they're working through a query. As those researchers acknowledge, CoTs are a helpful transparency into the inner workings of AI that allow users to see any "intent to misbehave" or potential backwards reasoning as it happens. Still, there […]

---

Link :
https://futurism.com/top-ai-researchers-concerned