Aiming to accelerate progress in AI development and responsible AI adoption, Alphabet Inc.'s Google will combine teams responsible for building AI models across Google Research and Google DeepMind, CEO Sundar Pichai said. It said in a memo to employees Thursday. Pichai said in the memo that all AI “work will reside in Google DeepMind.”
The reorganization “will expand our ability to deliver powerful AI to our users, partners, and customers,” Pichai said in the memo. “This centralizes compute-intensive model building in one place and establishes a single point of access for PAs looking to leverage these models to build generative AI applications. It simplifies development.”
The Google DeepMind team will be led by Demis Hassabis, the memo said.
Exactly one year ago, Google merged two AI research teams, Google Research's Brain team and DeepMind, to form Google DeepMind. Pichai said in an April 2023 memo that this focused team, backed by Google's computing power, “will significantly accelerate our progress in AI.” The Gemini model was created by Google DeepMind.
Google Research is the research arm of Google, specializing in AI and computer science, with key focus areas including AI/ML, responsible human-centered technology, science and social impact, computing paradigms, and algorithms. We develop next-generation technology to help your products. And optimization.
What's new now?
Google is currently consolidating all of its AI units into one to “simplify structure and improve speed and execution.” For example, we are integrating Google Research's Brain team and DeepMind's team to help accelerate Gemini models. Integrate your ML infrastructure and team of ML developers for faster decisions, smarter compute allocation, and better customer experiences. and bring our search team under one leader,” the memo added.
The move also gives Google Research a clear mandate to continue investing in three key areas that align with Google's mission: computing systems, fundamental ML and algorithms, and applied science and society. , Pichai said.
“Combining all of Google's AI teams, including Google Research and DeepMind, into one division under Google DeepMind will help us develop AI across the company,” said AI expert and co-founder Pradeepta Mishra. “This is likely to reflect a strategic move aimed at streamlining and optimizing deployment.” of cybersecurity company Data Safeguard.
Additionally, Google is reaffirming its commitment to responsible AI adoption by strengthening the role and accountability of its Responsible AI team. The team focused on Responsible AI within the Google Research team will move to Google DeepMind to be closer to “where models are built and scaled,” the memo states.
Similarly, the “Other Responsibilities Team'' is transitioning to our core “Trust and Safety'' team, and the company is investing further in “AI Testing and Evaluation'' to improve product accuracy and responsiveness. I am. “These changes create clearer responsibility and accountability at all levels when building and deploying, and strengthen the feedback loop between models, products, and users,” Pichai said in the memo. added.
Google recognizes the potential of AI and is formally promoting collaboration between its AI departments, software, and computing platforms. “So we are formalizing the collaboration between DSPA and P&E, bringing the teams together into a new PA called Platforms & Devices.”
Having a unified team across platforms and devices will allow Google to deliver high-quality products and experiences to its users and partners, Pichai said. This will help us accelerate the Android and Chrome ecosystem and deliver the best innovations to our partners faster, like we did with Circle to Search with Samsung. Internally, it also speeds up decision-making. ”
How Google can help
Unifying teams eliminates redundancy and fosters closer collaboration between researchers and developers. This could accelerate the development cycle for new AI products and features. All of these moves “will help us become more focused and clear on our mission,” Pichai said in the memo.
“Having one central unit makes decisions about resource allocation and project prioritization more efficient,” Mishra added. The environment surrounding AI continues to be fiercely competitive. Data Safeguard's Mishra said the integration will allow Google to “stay ahead of the curve by accelerating innovation.”
“By consolidating R&D under one roof, Google has the potential to create more integrated and impactful AI products across its platforms (Search, Assistant, etc.). This can lead to more efficient resource utilization and potentially cost savings,” said Mishra.
