Gf
Sergey Brin on the Future of AI & Gemini
@ Google for Developers
05/24/25
Takeaways
SB
Sergey Brin
05/24/25
@ Google for Developers
The general sentiment at I/O is incredibly positive, reflecting a significant amount of progress across Google's models and products.
SB
Sergey Brin
05/24/25
@ Google for Developers
I was surprised by many announcements at I/O, as I was unaware of about 30% of them due to my focus on Gemini.
SB
Sergey Brin
05/24/25
@ Google for Developers
The reception of the new features has been great, but it takes time for people to explore and understand them fully.
SB
Sergey Brin
05/24/25
@ Google for Developers
My primary focus is on the core text model of Gemini, as I believe it will lead to advancements in self-improvement and AI development.
SB
Sergey Brin
05/24/25
@ Google for Developers
Generative media models are impressive, but my attention is mainly on Gemini because of its potential for significant breakthroughs in AI.
SB
Sergey Brin
05/24/25
@ Google for Developers
The integration of audio in Veo enhances the richness of the experience, making it more compelling than just visual elements alone.
SB
Sergey Brin
05/24/25
@ Google for Developers
The native audio feature in Gemini has been in development for over a year, and its release has taken longer than expected due to various challenges.
SB
Sergey Brin
05/24/25
@ Google for Developers
Language models are currently the primary focus of AI development, which was not anticipated 15 years ago when we were exploring physical grounding in AI.
SB
Sergey Brin
05/24/25
@ Google for Developers
The surprising interpretability of thinking models allows us to understand how they reach their conclusions, which was not expected in earlier AI developments. The fact that AI models can articulate their reasoning provides a level of comfort regarding safety, even though it's not infinite comfort.
SB
Sergey Brin
05/24/25
@ Google for Developers
Architecturally, there's a significant similarity among various AI models, including video diffusion models like Veo and text language models, which share a lot of core elements.
SB
Sergey Brin
05/24/25
@ Google for Developers
The shift in model training is moving from a focus on pretraining to a greater emphasis on post-training, which is becoming a more substantial part of the overall training process.
SB
Sergey Brin
05/24/25
@ Google for Developers
The ability to allow AI models to reason for extended periods, rather than just providing quick answers, can lead to significantly better outcomes for complex questions.
SB
Sergey Brin
05/24/25
@ Google for Developers
Google has always been an AI company at its core, focusing on large-scale data and analysis, which positions it well for the current shift towards AI innovation.
SB
Sergey Brin
05/24/25
@ Google for Developers
The launch of the 2.5 Pro model represents a significant leap forward for Google, showcasing advancements in AI capabilities and performance across various benchmarks.
SB
Sergey Brin
05/24/25
@ Google for Developers
The recent introduction of the 2.5 Flash model, which ranks highly on performance leaderboards, demonstrates Google's ongoing commitment to AI innovation and development.