The 5 Google I/O Announcements That Actually Matter

Google’s annual I/O developer conference, which kicked off on Tuesday in Mountain View, California, was full of announcements of new AI features, hardware and tools. Our reporters were on the ground covering every update on CNET’s live blog.

Here's a recap of some of the highlights.

Gemini Spark task automation

AI was front and center at Google I/O this year, and the headline announcement was Google’s new “personal AI agent," called Gemini Spark. Similar to Anthropic’s Claude Cowork or Microsoft 365 Copilot, Gemini Spark will run 24/7 in the background to offload tasks.

It can comb through your various Google software accounts like Gmail, Docs and chats to perform tasks like compiling info from your documents into weekly emails to your team or sending out a friendly reminder to anyone who hasn’t RSVP'd to your party.

Spark will roll out to test users immediately, with a beta version expected for Google AI Ultra subscribers in the US in the next week. For everyone else, Google says that it will be available in Chrome later this summer.

Ask YouTube AI-powered Search

Google’s AI search features are coming to YouTube. Called Ask YouTube, the feature allows you to make ultraspecific queries on the platform, like “how to change the oil on a 2019 Subaru Outback.” If the initial results don’t show you what you were looking for, you can ask follow-up questions that further refine your results. Ask YouTube will even jump straight to the relevant part of the video.

Google

You can try the feature right now if you’re a Premium member in the US aged 18 and up. Google says it will roll out to more users in the near future.

Google Expands AI Identification Tool

Google is also attempting to make it easier to identify AI. Google's SynthID invisible watermark system was released late last year, but it had been limited to the Gemini app. Now, Google will be able to detect AI-generated images, videos and audio with the watermark in Google Chrome and Google Search. You’ll be able to use Google's Circle to Search feature, which allows you to right-click on an image and ask if any part of it was generated with AI.

Along with its own Gemini AI model, Google has partnered with ChatGPT-maker OpenAI, AI voice generator ElevenLabs and Nvidia on its SynthID system.

"It's great to see the cross-industry collaboration," Google CEO Sundar Pichai said. "We are looking forward to expanding to more partners and setting the standard of transparency for the AI era."

Docs Live brain dumps

There’s no shortage of AI tools that will spit out volumes of text, but they can often take some back and forth to get what you need. Google’s new Docs Live feature works directly in Google Docs, and can turn your random thoughts -- spoken or written -- into a piece of coherent writing. If you give it permission, it will even use your connected Google accounts (Gmail, Drive and Chat) and the web to fine-tune the results even more.

Docs Live will be available to Google AI subscribers when it launches this summer, but it will be limited to those on the AI Pro ($20 per month) or Ultra ($100 or $200 per month) tiers.

So many smart glasses

If you’ve been interested in the world of smart glasses, Google debuted a slew of new options at its I/O conference this year. CNET’s resident wearable tech expert, Scott Stein, was on hand to test them all, and he came away impressed with the new options.

“They could be the best smart glasses out there,” Stein wrote. “They could even help redefine what smart glasses can be. And they're the first glasses I've seen that really feel ready to work with the apps and services we have on phones already.”

There are valid privacy concerns about wearing smart glasses, and it sounds like Google is still working through them. Google's head of XR, Shahram Izadi, told Stein that the company plans to go into more detail on data privacy on its glasses at a fall event. For now, Izadi says all of Google’s smart glasses are equipped with a bystander LED indicator that lets people around you know when a camera or microphone is active.

“Given the privacy aspects, you have to design with privacy in mind from the very, very beginning and leverage some of the standards that have been set in the AI world," Izadi told Stein. “We do need to raise the bar, for sure."