旗舰模型

🚨 EXCLUSIVE from The Information: OpenAI has shifted 10% of its ENTIRE compute to AI safety after the HF incident. it’s now scrambling to fix safety issues before releasing their next flagship model which again uses recursive reasoning that makes monitoring harder.

Hesam(@Hesamation) · 7.5 分

OpenAI将10%的计算资源转向AI安全，以应对HF事件后的安全问题，并在发布新旗舰模型前修复隐患。