Vinson·Li

Index

On ai-safety

  1. "Too dangerous to release"

    OpenAI trained a much bigger language model and is holding back the full version. The unicorn story is impressive. What's actually dangerous is cheap, plausible text at scale.

    2 min