OpenAI's 2019 GPT-2 'Too Dangerous to Release' Post Resurfaces with 386 Points — Community Reflects on How Far AI Has Come

Hacker News·low signal

A 2019 Slate article about OpenAI declaring GPT-2 'too dangerous to release' resurfaced on HN with 386 points and 114 comments, serving as a striking contrast to today's landscape where far more capable models are freely available. The discussion thread features practitioners reflecting on how quaint GPT-2's capabilities seem seven years later, with many noting the irony that OpenAI now releases models orders of magnitude more powerful while pivoting to for-profit. A useful historical benchmark for the AI acceleration narrative.