Google I/O Was One Big AI Update: Everything Announced, From Astra to Veo
Big changes are coming for Google Search -- I/O showcased an upgraded version of Gemini, plus AI Overviews and a new generative video app, Veo.
Google found creative ways to debut its new version of Gemini, as well as other AI features.
Google's annual developers conference, Google I/O, brings us the announcements typical of these events: new software features the company hopes its developer community will exploit over the next year, as well as its own uses for the tech. In this case, that means AI: AI in search, AI in Android 15 (currently in beta) and the updates to its core AI tools -- Gemini, DeepMind and more. In fact, Sundar Pichai referred to it as "the Gemini era." New apps take advantage of the upgrades in the models, like Ask Photos, AI Overviews, Imagen 3 and a lot more.
I can't remember a launch event by one of the major tech companies over the past year that didn't center around how wonderful-fast-timesaving its AI models, developer tools, features and silicon are. That's the signal it's not mainstream yet, and still in the major confusion phase. (We try to help demystify it all in our guide, AI Atlas.) But Google gave us two hours of it and not much else.
The company didn't talk about hardware at all, except for a shout out to new Trillium Tensor processors and that Nvidia's Blackwell GPUs are coming to its data centers, both in 2025. Plus, Axion TPUs are more energy efficient. It announced the Pixel 8A a week before the show.
For more of a minute-by-minute experience, we've archived our live blog, and you can find all our coverage of the event at our Google I/O hub.
So what's new with Gemini?
Gemini, the company's ChatGPT competitor, has been updated to make it more conversational, have better cross-app compatibility and be smarter in general. Project Astra, "the future of AI assistants," takes advantage of Gemini to make conversation more natural and caches info for faster response. It can use the camera to respond to vocal queries based on what you're pointing at thanks to improved video processing as well.
Circle to Search, which we saw with Samsung's last flagship phone announcement, comes to Android generally. The Gemini app assistant starts the process of replacing Google Assistant, using an on-device model. Upcoming Gemini Nano is a smaller version of the model with less latency and device-specific functions, like scam warning popups mid-call.
The new Gemini mobile app in Android 15 brings Live (voice conversations with the AI), use the camera as described above, has a feature called Gems, which are sets of instructions you create that can access Google Drive to perform complex operations that you do frequently, as well as figure out the same complex research and planning tasks, such as trip booking (coming Labor Day weekend!).
AI Overviews are a pane in Search that provides AI-assisted (processed, packaged and analyzed) results of whatever you're looking for; it launches in the US this week.
Boost your image search with Ask Photos
Ask Photos can search for specific info about what's in your libraries, such as finding a particular license plate that's in there... somewhere. It can use multistep search to answer complex queries, for instance, "where are the best restaurants in Texas for an anniversary dinner," that takes into account the season -- a new AI search results page summarizes it all for you.
NotebookLM's new virtual assistant.
Gemini 1.5 Pro, its analysis version, comes out of preview and will become available to Gemini Advanced subscribers and all developers: It understands images better and will double the context window from 1 million to 2 million tokens later this year. (A token is a bit of information to process, so more tokens = better.)
Your Gmail gets smarter?
In Google Workspace, Gemini 1.5 Pro becomes available in Workspace to summarize emails, as well as in Gmail on mobile. It rolls out next month to Labs users. In addition to summarizing, it can organize messages, export the data and save it as a workflow for future email (coming in the Fall). With the help of the new LearnLM model, it can help with homework, lesson planning and other education-related tasks.
Final AI thoughts
Upgraded Imagen 3 generative AI tools, via DeepMind, have improved text understanding and better details for image generation. It also brings a music AI sandbox and more types of prompts for video -- they're available in the company's Labs' tools.
Veo is a new generative model/tool based on DeepMind for creating videos. Google trotted out actor/comedian/rapper/writer/director/producer Donald Glover to be a spokesperson for it.
NotebookLM has a new "virtual teammate" you can set up to handle, for instance, summarizing all conversations to reply to specific questions. It builds a collective memory for teams, function as a project manager to create timelines and point out bottlenecks.
Overall, the 1.5 Pro upgrade makes everything a little better. Gemini 1.5 Flash is a small version of Pro (with 1 million tokens), which is going to be coming for developers.
Editors' note: CNET used an AI engine to help create several dozen stories, which are labeled accordingly. The note you're reading is attached to articles that deal substantively with the topic of AI but are created entirely by our expert editors and writers. For more, see our AI policy.