Google has combined two of its best-known artificial intelligence groups into a new unit called Google DeepMind. The company says the organization will bring research and product work closer together as it seeks to build more capable AI systems and products across Google and Alphabet.
A new structure for Google’s AI work
Google DeepMind joins the DeepMind team with Google Brain, which was part of Google Research. DeepMind co-founder and CEO Demis Hassabis will lead the new unit. In a blog post, he said it would work closely with Google’s product areas to deliver AI research and products.
The reorganization also establishes a scientific board to oversee research progress and the unit’s direction. Koray Kavukcuoglu, DeepMind’s VP of research, will lead it. Eli Collins, formerly VP of product at Google Research, will become Google DeepMind’s VP of product. Google Brain lead Zoubin Ghahramani will join the research leadership team and report to Kavukcuoglu.
Jeff Dean, a Google Brain co-founder, will become chief scientist for both Google Research and Google DeepMind. He will report to Google CEO Sundar Pichai. The arrangement links the two research organizations through a shared senior science role while keeping them as separate divisions.
Google Research keeps a separate focus
Google Research will remain an independent division, reporting to Google SVP of technology and society James Manyika. Its stated remit covers fundamental advances in computer science, including algorithms and theory, privacy and security, quantum computing, health, climate and sustainability, and responsible AI.
That distinction matters because the announcement describes different responsibilities for the two groups. Google DeepMind is being assembled to connect AI talent with Google’s computing power, infrastructure and product areas. Google Research will continue work across a broader set of scientific fields.
Hassabis framed the combination as a way to bring research capability and resources together, with the goal of producing new advances and products. Google CEO Sundar Pichai also presented the move as part of an effort to develop general AI more safely and responsibly.
Two research histories meet
DeepMind was co-founded in 2010 by Hassabis, Shane Legg and Mustafa Suleyman. Google acquired the company for $500 million in 2014. The lab later became known for AI that could beat human professional Go players and for predicting structures of complex proteins.
Google Brain began in 2011 as a part-time research collaboration involving Dean, Google researcher Greg Corrado and Stanford University professor Andrew Ng. Its work included TensorFlow, a framework for creating AI models, and improvements to Google Translate.
The teams had worked together for years, but the source describes occasional tension between DeepMind and its corporate parent, Alphabet. Senior DeepMind managers reportedly sought more autonomy, including an independent legal structure for sensitive research. Those talks were called off in May 2021.
The consolidation comes amid a tougher AI race
Bringing the groups together comes as Google faces pressure to compete with Microsoft and OpenAI in AI products that can generate revenue. The Information reported that Google Brain engineers and DeepMind employees were working on Gemini, software intended to compete with OpenAI.
Google had also made recent AI product moves, including Bard, which the source says were poorly received. It describes the company as responding to Microsoft’s collaboration with OpenAI on Bing Chat and investing resources in Magi, a set of AI-enabled search features with a task force of over 160 people, newly formed as of this year.
The reorganization brings together teams with distinct histories and capabilities under one AI-focused unit. Its stated ambition is to speed research and product development while emphasizing responsible development. Whether the new structure delivers that acceleration will depend on how effectively those teams and resources work together.