Google's new "Ask YouTube" feature. Image: GoogleGoogle used its I/O developer conference on Tuesday to showcase how deeply it plans to embed AI across its products, including new models and a conversational search feature for YouTube.Why it matters: Google comes to the AI race with arguably the deepest set of assets, but also the largest amount of surface area to defend against a world of AI-native apps.Here's what Google has announced so far:YouTubeA new "Ask YouTube" feature allows users to make conversational search queries and get results across Shorts and traditional videos. In most cases, results include both text answers and the videos from where they are drawn.It's being tested now and will be more broadly available this summer.AI modelsGoogle DeepMind head Demis Hassabis announced Gemini Omni, which is designed to translate between different forms of media, including text, audio, images and video. Its initial release, however, is focused solely on generating video from combinations of text, images, audio and video. "Anything becomes a canvas for creating entirely new realities," Hassabis said, calling the model a key step towards so-called artificial general intelligence.What they're saying: "It's been an intense year," CEO Sundar Pichai said, kicking off the event in Mountain View, Calif.Pichai noted that Google is now processing 3.2 quadrillion tokens per month, a key measure of AI usage. That's up from 480 trillion tokens per month a year ago and 9.7 trillion tokens per month two years ago."We knew how profound AI would be to advancing our mission and improving people's lives at scale."Catch-up quick: AI has been the focus of the conference in recent years. Last week, the company made a number of announcements related to Android, including previewing the Googlebook, the Gemini-infused successor to the Chromebook that combines Android and Chrome OS.Editor's note: The keynote is just getting started. Check back for frequent updates.