The getting is good (optimizing unattended runs)
ai
A researcher shared observations on LessWrong about testing frontier AI models when left running unattended, finding significant variation in stability across models. Some struggled within hours of continuous operation, while others demonstrated extended reliability. Per the post, LessWrong discussions on AI improvements have historically influenced frontier model development, with the author suggesting an opportunity for the community to contribute methods for improving accuracy and reliability.
Source: https://www.lesswrong.com/posts/xcEuZhuZGeExoYazj/the-get...
Listen to this story
Hear this and more stories in a personalized audio briefing.
Open The Chonkerton