(Credit: Rob Pegoraro)
MOUNTAIN VIEW, Calif.—The price of entry for many Google searches is only some of your attention and some of your data. But most AI has to run in data centers that require ever-growing amounts of electricity. So, if you want to play, you’ll have to pay.
For those willing to sign up for one of its service plans, Google is rolling out a suite of new or updated AI-driven tools, built on the new 2.5 version of the company’s Gemini AI platform. Demis Hassabis, CEO of Google DeepMind, calls it "our most intelligent model."
“We want to build an AI assistant that's personal, proactive and powerful,” said Josh Woodward, VP of Google Labs and Gemini, during a pre-I/O briefing.
CEO Sundar Pichai opened the I/O keynote by emphasizing how the company is rushing to get these tools into use–having shipped more than 20 AI products and features since last year’s I/O. "We want to get our best models into your hands and into our products ASAP,” he said. “We are shipping faster than ever."
Putting AI to Work in Workspace
Subscribers to Google Workspace are in line to get some new features in Gmail, the core of that service's bundle. For example, “personalized smart replies” go beyond the canned answers you opt for to answer a message to draw on your past correspondence and Google Docs to adapt to “your typical tone.”
"Gemini can do almost all the work for me,” Pichai said in his keynote. “It uses my favorite adjective, 'exciting,'" he commented after showing how Gmail wrote a set of trip directions for him. (It’s unclear I need an AI-generated version of the sarcasm that already gets me in enough trouble when it comes organically from my own typing fingers.)
An "inbox cleanup" option, meanwhile, lets you automate mail management tasks by, for example, telling Gmail to delete messages from particular senders after a set interval. Microsoft Outlook.com users should need no introduction to this concept, as that free webmail service has offered its “sweep” feature for more than a decade using merely software.
Starting next quarter, you’ll be able to link relevant supporting material to a Google Doc and have Gemini limit its writing assistance to the content of those documents, an implicit nod to concerns over AI hallucinations. And next month, the Google Vids tool that lets you turn a set of Google Slides into a video presentation will add an option to have an AI-generated avatar narrate the clip.
Google is also rolling out live translation in Google Meet video calls, at first between English and Spanish; that will initially require at least Google’s AI Pro plan, but the company says Workspace subscribers will start getting access to it later this year.
Go With the Flow–and the Imagen, Veo, Lyria, and Jules
Google is updating a variety of AI content-creation tools, starting its contributions to the AI image-generation and video-generation genres. Imagen 4 offers what Google calls finger image textures and improved typography, and is available via Workspace.
Veo 3, meanwhile, not only generates video but can now handle audio, taking this tool out of the silent-movie era. Google showed off Veo’s capabilities in a series of AI-generated clips played before the I/O keynote.
Imagen, Veo and the music-generation tool Lyria all watermark their output with SynthID, a standard for identifying AI-generated content. Google is also launching a SynthID detector site at which people can upload a questionable item to see if it shows signs of being SynthID-marked.
For more ambitious moviemaking, Google now offers Flow, which can create complete cinematic shots from prompts as specific as requesting the visual style of a specific type of camera and then stitch them together into longer material.
Using Flow will cost a little more than the $7/month/user cost of a “Business Starter” Workplace subscription: You’ll need to sign up for one of Google’s two higher-end AI plans: AI Pro (formerly AI Premium), $19.99/month, or AI Ultra, at $249.99 a month.
The latter includes 30TB of storage, access to all of the above tools and access to Google’s latest Project Mariner agentic AI to automate common web tasks.
One last AI-creation tool, however, is free for now: Jules, which Woodward called an "asynchronous coding agent" that lets developers debug and optimize code that they’ve uploaded to a private, secure Google Cloud virtual machine. Google’s post says the company will announce pricing once Jules exits beta.