AI discussions at Google were more focused on improving search rather than contemplating the long-term implications of AGI.

Google's recent advancements in AI, particularly with their Gemini model, demonstrate their commitment to reclaiming leadership in the AI space after OpenAI's initial success.

I used to feel good about working for Google because they were responsible and cautious with AI, but I would feel less happy working for them today due to their military AI involvement.

In early 2023, I realized AI development was outpacing our ability to control it, influenced by the impressive capabilities of ChatGPT and my work on analog computation at Google.

Google's risk aversion regarding AI stems from concerns about regulatory backlash and the potential for AI to produce offensive content.

Google's Gemini 2.5 Pro is now the leading model across all categories, showcasing significant advancements in AI capabilities.

AGI is becoming more central to Google's vision, and I sometimes describe Google DeepMind as the engine room of Google.

I was really excited about the developer keynote at Google I/O, where we discussed how we're enabling multimodal capabilities, like taking images to code and using AI to help with rapid coding and idea exploration.

It's so awesome to meet developers from around the world and share the latest from Google DeepMind, especially seeing the excitement in the Gemini community, which is vibrant right now with many developers building with our models and asking for more features.