Big Technology Podcast · Alex Kantrowitz

How Google DeepMind Operates & Experiments — With Lila Ibrahim and James Manyika

·50 min·3 clips

Google Labs was rebooted three years ago to build AI-first products from Google research.

1. Big Technology Podcast centers on how Google DeepMind operates, experiments, and ships AI products from a live Davos stage. 2. Host Alex Kantrowitz speaks with Lila Ibrahim, Chief Operating Officer of Google DeepMind, and James Manyika, SVP of Research Labs, Technology, and Society at Google. 3. The episode asks how Google DeepMind balances ambitious research, product shipping, and responsible AI across Google. 4. Lila Ibrahim describes DeepMind’s mission as “build AI responsibly to benefit humanity” and says the team starts with big problems. 5. She says the company structures research around ambitious goals without telling people how to do the work. 6. Ibrahim compares the operating model to Bell Labs, the Apollo program, and Pixar. 7. She says interdisciplinary teams matter, naming a bioethicist, a computer scientist, and a neuroscientist. 8. Ibrahim says DeepMind is comfortable exploring, then deciding whether to pause a project or “double down.” 9. She gives learning science as an example and says Gemini absorbed that work after Google DeepMind decided it was ready. 10. She says the organization is a mix of top-down direction from Demis Hassabis and bottom-up initiative from researchers. 11. Manyika says the Gemini program emerged when Google Brain and DeepMind were put together three years ago. 12. He says Gemini now underlies Search, Google Workspace, NotebookLM, and other products. 13. Manyika says Google DeepMind is acting as the company’s “AI innovation engine.” 14. He adds that Google does not simply farm models out to product teams, but works closely with those teams and their customers. 15. He says the goal is to test models before they reach products so they can be refined for specific uses. 16. Manyika says the latest Gemini versions now ship on roughly a five- to six-month cycle. 17. The conversation turns to Google Labs, which Manyika says was rebooted three years ago for AI-first products. 18. He traces NotebookLM back to a five-person project called Tailwind and says citations are a key feature. 19. He says audio overviews began as a way to hear paper summaries while driving and later became podcast-style discussions. 20. The episode closes with AI in education, where Ibrahim and Manyika discuss guided learning, teacher productivity, and a Northern Ireland pilot that saved about 10 hours per teacher per week.

As heard by us

A concise look at how Google keeps experimental ideas moving between employee projects and Google Labs.

Google DeepMind's operating model comes through most clearly in the question of how Google still makes space for ideas that do not fit neatly inside a job description.

Read the full review in PlayNext →

Why you'd press play

You want a behind-the-scenes listen on how Google's old 20% time spirit still shows up in new experiments.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Big Technology Podcast