DigiNews

Tech Watch by Johan Denoyer

← Back to articles

OpenAI says its new model GPT-2 is too dangerous to release (2019)

Quality: 8/10 Relevance: 9/10

Summary

OpenAI did not release the full GPT-2 model due to safety concerns, releasing a smaller version while withholding data and training code. The article discusses the model's capabilities, the debate among researchers about withholding results, potential misuse like false information and impersonation, and the broader question of how to govern powerful AI tools.

🚀 Service construit par Johan Denoyer