upvote
OpenAI has been pulling this marketing trick for years. Remember how GPT-3 was too dangerous to release? It's also probably bad PR if script kiddies have access to GPT model with no guardrails even if it doesn't enable any significant attacks.
reply
I suppose you meant GPT-2, but for years? Did they say the same about subsequent models?
reply