Google reorganized DeepMind to remove 90-person AI responsibility teams

On 27 August, according to the Wall Street Journal, Google plans to move some 90-person AI duty teams out of DeepMind in early September, as part of a further integration of DeepMind into the Google system, as well as human resources and policy teams out of DeepMind. The AI Accountability Team focuses on front-line AI risks, including chemical, biological, radiological and nuclear risks of testing models, and on the psychological impact of user-AI interactions and chat robots. Some staff are concerned that moving out of DeepMind will increase the distance from the Gemini model development team and may reduce the ability to conduct independent research and to identify front-line model risks. Google, for its part, indicated that the team would retain DeepMind access, numeracy and staff resources, and other teams responsible for model behaviour, privacy and security would remain in DeepMind. This adjustment is at a time of continuous change in DeepMind management and talent。