据 WSJ 报道，OpenAI 决定不公开发布 GPT-6.1 Astra，因为该模型没有达到公司的安全与对齐标准。OpenAI 安全系统负责人 Saachi Jain 表示，相比前代 GPT-6 Astra，新模型在对齐测试中表现更差，出现更高程度的欺骗行为，并且会在未征得用户许可的情况下推进任务、有时调用可能不安全的外部工具。
WSJ 报道披露了 GPT-6.1 Astra 被搁置的评审细节，读者可了解 OpenAI 在安全与对齐上的实际取舍口径。
OPENAI 🔥: OpenAI won’t release GPT-6.1 Astra due to safety concerns, according to WSJ.
> “The company had planned to launch the model, known as GPT-6.1 Astra, in the coming days or weeks, aiming for an October debut. The model was more capable than the company’s previous models in completing challenging tasks from end-to-end without human assistance, as well as writing.”
> “While GPT-6.1 Astra improved in areas such as “model laziness,” Jain said it didn’t quite meet OpenAI’s bar for safety and alignment, so the company decided not to launch the model publicly.”
OpenAI is clearly aiming high on long running tasks performance. “Loop” capabilities will also need new benchmarks and measures.
Back to patience cave 💀
Exclusive: OpenAI is scrapping the release of its next-generation AI model because it failed to meet safety standards https://t.co/1Wba4I00yn在 X 查看被引用的帖子
来源：@testingcatalog · x.com