Subvolts take Google details Gemini's native multimodal design and the future of vision-centric AI assistants. Jul 2, 2025 - Developers
- Learn about the technical considerations for building natively multimodal AI and improving video understanding capabilities.
- Business
- Understand Google's strategy for multimodal AI and the potential for proactive AI assistants in future products.
- Everyday users
- Discover how AI is evolving to understand visual information, leading to potentially more helpful and intuitive AI assistants.
This episode provides a solid interview with a Google product lead discussing Gemini's core multimodal architecture and future AI assistant vision.
Gemini's Multimodality
Audio streams from Google AI. Your place is remembered on this device.
From Subvolts
Summary and analysis by Subvolts editorial, AI-assisted.
Logan Kilpatrick interviews Ani Baddepudi on Google AI's "Gemini's Multimodality" podcast episode. They discuss Gemini's design as a natively multimodal model, the future of AI assistants, and the concept of "everything is vision." The conversation covers technical aspects like video and image understanding, token representations, and higher FPS video sampling.
This episode explains how Google built its Gemini AI to understand images and videos from the start, not just text. It explores how AI might become more helpful by seeing and understanding the world around us, like a proactive assistant.
This episode offers insight into Google's strategic approach to building multimodal AI from the ground up. It's for anyone interested in the technical underpinnings of advanced AI systems and the future direction of AI assistants. The discussion on video understanding and proactive AI aligns with current frontier lab efforts to create more integrated and capable AI experiences.
- HostLogan Kilpatrick
- GuestAni BaddepudiGemini Model Behavior Product Lead, Google
- Guest: Ani Baddepudi, Gemini Model Behavior Product Lead
- Show: Google AI: Release Notes
- Length: 44 minutes
- Key Concept: Gemini built as a natively multimodal model from day one
- Future Vision: Moving towards a world where "everything is vision"
Ani Baddepudi, Gemini Model Behavior Product Lead, joins host Logan Kilpatrick for a deep dive into Gemini's multimodal capabilities. Their conversation explores why Gemini was built as a natively multimodal model from day one, the future of proactive AI assistants, and how we are moving towards a world where "everything is vision." Learn about the differences between video and image understanding and token representations, higher FPS video sampling, and more.
Chapters:
0:00 - Intro
1:12 - Why Gemini is natively multimodal
2:23 - The technology behind multimodal models
5:15 - Video understanding with Gemini 2.5
9:25 - Deciding what to build next
13:23 - Building new product experiences with multimodal AI
17:15 - The vision for proactive assistants
24:13 - Improving video usability with variable FPS and frame tokenization
27:35 - What’s next for Gemini’s multimodal development
31:47 - Deep dive on Gemini’s document understanding capabilities
37:56 - The teamwork and collaboration behind Gemini
40:56 - What’s next with model behavior
Watch on YouTube: https://www.youtube.com/watch?v=K4vXvaRV0dw
- multimodal
- Ani Baddepudi
- Able to understand and process different types of information, like text, images, and video, all at once.
- token representations
- Ani Baddepudi
- How AI breaks down different types of data (like words or image parts) into smaller pieces for processing.
- FPS
- Ani Baddepudi
- Frames Per Second, a measure of how many still images (frames) are shown each second to create the illusion of motion in video.
27 minKoray Kavukcuoglu on frontier models, coding agents, and building AGI
WithKoray Kavukcuoglu
39 minIntroducing Gemini Robotics 2
WithCarolina Parada, Stuart Bowers, Kanishka Rao, Jie Tan
42 minGemini co-leads on project origins and what's next
WithJeff Dean, Koray Kavukcuoglu, Noam Shazeer, Oriol Vinyals
40 minQuestions people ask
- Why was Gemini built as a natively multimodal model?
- Gemini was designed from day one to be natively multimodal, meaning it can process and understand multiple types of information simultaneously.
- What are the future implications of "everything is vision"?
- This concept suggests a future where AI assistants can perceive and understand visual information, leading to more proactive and integrated AI experiences.
Page generated Oct 19, 2026. Summaries are Subvolts' own; the audio, artwork and notes belong to Google AI.