英国 AI Security Institute 与 Google DeepMind 深化研究合作
Deepening our partnership with Google DeepMind
英国 AI Security Institute 宣布与 Google DeepMind 启动新的研究合作，签署研究 MOU，在既有前沿模型测试合作基础上开展双向协作，包括共享数据与想法、联合发表以及团队专项合作。首批联合研究方向为监控思维链以推进对齐、研究社会情感对齐以理解 AI 系统如何影响人类福祉，以及研究 AI 对经济活动的影响。
Today we’re launching a new research partnership with Google DeepMind, marking the next step in our collaboration over the past two years. This builds on our existing work together on frontier model testing and forms an important part of DSIT’s new broader partnership with Google DeepMind.
The research MOU announced today will enable deeper two‑way collaboration: shared access to data and ideas, joint publications, and dedicated work between our teams to advance the science of AI safety and security.
Our initial joint research areas include:
- Monitoring chains‑of‑thought for alignment, expanding on our collaboration earlier this year.
- Investigating socio‑affective alignment to better understand how AI systems interact with human wellbeing.
- Studying AI’s impact on economic activity to inform long‑term labour‑market understanding.
We look forward to building on two years of productive partnership and sharing the results of this expanded research with the wider community.
来源：英国 AI Security Institute：Blog · aisi.gov.uk