Google is bringing two of its leading AI groups under one roof. The new organization, Google DeepMind, joins Google Brain with DeepMind and puts large-scale multimodal models at the center of its plans.
A new structure for Google’s AI teams
DeepMind was founded in 2010 by Demis Hassabis and Shane Legg. Google acquired it in 2014, though it remained largely independent. Google Brain, meanwhile, was the company’s in-house AI team.
In March, The Information reported that Google Brain and DeepMind were joining forces on the secret Gemini project. The announced merger makes that combination official: Hassabis will lead Google DeepMind as CEO, while Jeff Dean, formerly the head of Google Brain, becomes chief scientist.
Dean will report directly to Sundar Pichai, Alphabet and Google’s CEO. Pichai said Dean and Hassabis would guide the direction of AI research and lead strategic technical projects, beginning with a series of powerful multimodal AI models.
Scale and multimodality are central
Multimodal AI refers to systems designed to work across different forms of information. The article identifies this direction as a focus for Google DeepMind, alongside large-scale model development. Pichai said combining the teams with Google’s computing resources would significantly accelerate AI development.
The merger builds on work the two groups have already done. Google Brain and DeepMind teams developed influential technologies including Transformer and Deep Reinforcement Learning. The article also notes that many scientists involved in that work have since left the companies, with some starting AI businesses of their own.
For Google, the new structure is part of a wider realignment. The company has developed the Bard chatbot, offered AI models in the cloud and in Workspaces, and is looking to change search through its ‘Magi’ AI project. These efforts point to AI being pursued across several parts of the business.
A crowded field, with unresolved questions
Google’s emphasis on multimodal models follows a direction also being pursued by OpenAI and Germany’s Aleph Alpha. Jeff Dean had presented a similar vision for Google through Pathways. The article frames this focus as an economic bet: despite uncertainty over whether issues such as hallucinations can be overcome, companies across industries show strong interest in the technology.
That interest is reflected in the range of companies named in the article: OpenAI, Nvidia, AI21Labs, Amazon, and Aleph Alpha. The merger gives Google a single organization to drive its AI research and products, but it does not settle whether large language models’ weaknesses are temporary or fundamental.
Pichai believes Google can solve problems affecting Bard and similar systems, the article says, citing a recent interview with 60 Minutes. It also points to German data protection authorities’ administrative proceedings against OpenAI as another challenge Google may face beyond hallucinations.
What happens to research beyond multimodal AI?
A central open question is how the new organization will divide its attention. The article asks whether Google DeepMind will have the resources to continue supporting open-source projects such as AlphaFold 2 and advance research outside multimodal models.
That question matters because the merger is both a consolidation of teams and a sharper statement of priority. Google is combining expertise and computing resources to move faster on AI, while the breadth of the organization’s ambitions leaves its future support for other research areas uncertain.