Microsoft intensifies its offensive in the race for leading AI experts: Mustafa Suleyman, head of the tech company's AI division, has poached several senior researchers from Google DeepMind, including Marco Tagliasacchi and Zalán Borsos, two key developers of the audio AI "Audio Overviews," which converts texts into spoken podcasts, as well as Matthias Minderer, an expert in visual AI models.
The three scientists were instrumental in the development of Google's yet-to-be-released AI agent Astra, which can answer questions in real-time via video, audio, and text. They are now moving to Microsoft's new AI lab in Zurich to work on the further development of Copilot technology—an interactive AI intended to automate tasks like travel bookings or scheduling appointments.
It is an outstanding team and will be an important location for [Microsoft AI] alongside our rapidly growing London office," wrote Suleyman on LinkedIn. Tagliasacchi emphasized the importance of audio for "natural and immersive interactions" with AI.
The move is part of an intense competition for the best AI talents between Microsoft, Google, Apple, and other tech corporations. Since his move to Microsoft in 2024, Suleyman has already recruited several former DeepMind researchers as well as members of Google's health AI team.
Microsoft, Google, and OpenAI are working in parallel on so-called "multimodal" AI agents that can analyze and understand not only text but also audio, video, and images. OpenAI has already introduced an enhanced language function for ChatGPT, Google is planning the Astra launch for 2025, and Amazon wants to equip its voice assistant Alexa with AI features.
Google DeepMind declined to comment on the departures.