OpenAI says its agents turned off own safeguards

Технологии

OpenAI says its agents turned off own safeguards

Al Jazeera +13 ч назад

ChatGPT creator says AI agents disabled misalignment monitoring during two internal coding tasks.

Читать в источнике: Al Jazeera(откроется в новом окне)

Ещё 1 источник

CNBCOpenAI's most powerful AI models disabled their own safety monitoring — and one lied about it.(откроется в новом окне)