Google CEO Sundar Pichai delivered a comprehensive overview of the company's progress and future plans at the Google I/O 2024 event. The central theme of his address was the evolution of artificial intelligence (AI) and its increasing integration into Google's products and services.
Pichai reflected on Google's investment in AI over the past decade, highlighting innovations across research, product, and infrastructure. He emphasized that we are still in the early stages of the AI platform shift, but sees immense potential for creators, developers, startups, and users.
A year ago, Google introduced Gemini: a frontier model designed to be natively multimodal from inception. This model can reason across text, images, video, code, and more. Since then, Google has launched its first Gemini models which have demonstrated state-of-the-art performance on every multimodal benchmark.
The company has also been incorporating Gemini’s capabilities across its products in innovative ways. Today all of Google's 2-billion user products utilize Gemini. New experiences have also been introduced including mobile interaction with Gemini through an app now available on Android and iOS.
One transformation that stands out is in Google Search where billions of queries have been answered as part of their Search Generative Experience. Users are utilizing it to search in new ways by asking complex questions or even searching with photos.
Google Photos is another product that has benefited from Gemini. Launched almost nine years ago, people use it to organize their most important memories - more than 6 billion photos and videos uploaded every day. With Gemini’s help, users can now search their memories more efficiently.
Google Workspace is another area where Gemini is making an impact. It helps users summarize all recent emails from a specific source or provide highlights from long recordings among other things.
In addition to these advancements, Pichai announced that they are introducing Gemini 1.5 Flash tomorrow – a lighter-weight model built for scale optimized for tasks where low latency and cost matter most. He also mentioned the progress made on video and image generation with Veo and Imagen 3, and the introduction of Gemma 2.0, their next generation of open models for responsible AI innovation.
Pichai announced the upcoming launch of Google's 6th generation tensor processing units (TPUs), Trillium, which will be available to Cloud customers in late 2024. Trillium delivers a 4.7x improvement in compute performance per chip over the previous generation, TPU v5e.
Google is also investing heavily in its founding product, Search. With each platform shift, they have delivered breakthroughs to help answer users' questions better. The Gemini Era will take this to a whole new level combining infrastructure strengths, latest AI capabilities, high bar for information quality, and decades of experience connecting users to the richness of the web.
Finally, Pichai highlighted Google's commitment to a responsible approach to AI development. They are developing an innovative technique called AI-assisted red teaming that draws on Google DeepMind's gaming breakthroughs like AlphaGo to improve their models.