As the pace accelerates in the development of artificial intelligence, and more particularly in generative AI, the tech giants are responding in their own way. This week, Sundar Pichai, CEO of Google and Alphabet, announced a number of structural changes aimed at improving speed and execution across the company.
It is in the form of note to employees that Sundar Pichai spoke out. Quickly returning to the group's flagship event – Cloud Next – he wanted to reaffirm the need to simplify the structure. Last year, the Brain team at Google Research was grouped with the DeepMind teams, thereby accelerating the development of the Gemini models.
Advertisement
Research and DeepMind teams merge
At the same time, Google has been unifying its ML infrastructure and ML developer teams for faster decisions, smarter compute allocation, and better customer experience. Finally, the research teams were grouped under a single leader.
Today marks another milestone as the model building teams at Research and Google DeepMind will merge. All of this work will now be grouped within DeepMind, specifies Sundar Pichai.
A clear roadmap towards AI and quantum
Advertisement
With this change, the Google boss hopes to give his Research division clear guidelines to continue investing in fundamental and applied computer science research in three key areas: computer systems (including quantum), fundamental algorithms and science applied. Sundar Pichai adds that James Manyika, in his role as head of research at Tech & Society, will continue to oversee Google Research.
An entity called Platforms & Devices is created
In his overall vision, Sundar Pichai believes that the firm must “locate at the intersection of hardware, software and AI”. To do this, he announces formalizing the merger between the Devices and Services PA (DSPA) team – which manages Pixel, Nest and Fitbit hardware – and the Platform and ecosystems (P&E) team – which notably manages Android, Chrome, etc. . – in order to merge these two teams into an entity called “Platforms & Devices”.
Google Research teams specializing in computational photography and device intelligence will also be transferred within this organization, it is specified, “to bring deep AI expertise across platforms and devices.” This team will – among its many objectives – be responsible for boosting the Android and Chrome ecosystems.
Google wants to be reassuring about its place in the generative AI market
Note that this consolidation of teams working on artificial intelligence is not trivial. The announcement is intended to be both reassuring about Google's ability to remain a serious competitor in the race for generative AI, but also in its ability to be transparent and committed to the development of responsible AI.
The latest episode around Gemini's hallucinations caused a lot of noise internally. For the record, last February, Gemini users reported malfunctions in the generation of images representing human beings. Obviously, the systems put in place by Google to counterbalance representation biases (typically by only generating white people) have caused excessive correction, giving the opposite result.
Several Internet users have thus watch that it was very difficult to get the model to generate an image of a white person, even in a context that left no room for interpretation, such as a king of England in the Middle Ages. Jack Krawzyck, the creator of Google Bard – now renamed Gemini – then apologized on : “We are aware that Gemini has inaccuracies in some historical image generation representations, and we are working to resolve this issue immediately.” The tool was then paused to allow time to make some adjustments.
Do you want to stay up to date on the latest news in the artificial intelligence sector? Register for free to the IA Insider newsletter.
Selected for you