Over 40 scientists from OpenAI, Google DeepMind, Anthropic, and Meta ...

The smartest AI scientists in the world just issued a terrifying warning, and for the first time, they're all agreeing. Scientists from OpenAI, Google DeepMind, Meta, and Anthropic just co-authored a paper with one message. We're about to lose our ability to understand what AI is thinking. Right now, advanced AI models used in, like, ChatGPT shows when it's reasoning. It thinks out loud so we can see how it reaches decisions. This is our early warning system when AI goes wrong, but experts are saying that's about to disappear. The scientists found that AI models are already abandoning English in favor of unintelligible shortcuts. They're developing internal processes that are faster, more efficient, and completely opaque. Some cutting-edge models already skip language altogether, operating in mathematical space where there's nothing for humans to observe. Think about that. The AI could be planning something dangerous, and we'd have no way to see it coming. Two of the most highly regarded individuals in the AI space have endorsed this warning. When the people who created these systems are worried, we should probably be terrified.