Google has released (PDF) a new analysis of its AI's environmental impact, showing that it has cut the energy use of AI text queries by a factor of 33 over the past year. Each prompt now consumes about 0.24 watt-hours -- the equivalent of watching nine seconds of TV. An anonymous reader shares an excerpt from an Ars Technica article: "We estimate the median Gemini Apps text prompt uses 0.24 watt-hours of energy, emits 0.03 grams of carbon dioxide equivalent (gCO2e), and consumes 0.26 milliliters (or about five drops) of water," they conclude. To put that in context, they estimate that the energy use is similar to about nine seconds of TV viewing. The bad news is that the volume of requests is undoubtedly very high. The company has chosen to execute an AI operation with every single search request, a compute demand that simply didn't exist a couple of years ago. So, while the individual impact is small, the cumulative cost is likely to be considerable. The good news? Just a year ago, it would have been far, far worse. Some of this is just down to circumstances. With the boom in solar power in the US and elsewhere, it has gotten easier for Google to arrange for renewable power. As a result, the carbon emissions per unit of energy consumed saw a 1.4x reduction over the past year. But the biggest wins have been on the software side, where different approaches have led to a 33x reduction in energy consumed per prompt. The Google team describes a number of optimizations the company has made that contribute to this. One is an approach termed Mixture-of-Experts, which involves figuring out how to only activate the portion of an AI model needed to handle specific requests, which can drop computational needs by a factor of 10 to 100. They've developed a number of compact versions of their main model, which also reduce the computational load. Data center management also plays a role, as the company can make sure that any active hardware is fully utilized, while allowing the rest to stay in a low-power state. The other thing is that Google designs its own custom AI accelerators, and it architects the software that runs on them, allowing it to optimize both sides of the hardware/software divide to operate well with each other. That's especially critical given that activity on the AI accelerators accounts for over half of the total energy use of a query. Google also has lots of experience running efficient data centers that carries over to the experience with AI. The result of all this is that it estimates that the energy consumption of a typical text query has gone down by 33x in the last year alone.

Read more of this story at Slashdot.

BrianFagioli writes: Google is preparing to bring its television platforms in line with the rest of Android. Starting August 1, 2026, both Google TV and Android TV will require app updates that include native code to provide 64-bit support. The move follows similar requirements for phones and tablets, and it paves the way for upcoming 64-bit TV devices.

Read more of this story at Slashdot.

Phone manufacturers rarely acknowledge that IP68 water resistance degrades over time, but Google has broken that silence with advertising disclaimers for its Pixel 10 Pro Fold. The fine print explicitly warns that water and dust protection "will diminish or be lost over time due to normal wear and tear, device repair, disassembly or damage." The company further notes that liquid damage voids warranties despite IP68 certification at manufacture.

Read more of this story at Slashdot.

An anonymous reader quotes a report from The Verge's Victoria Song: The original Pixel Watch was late to the game. For years, there had been rumors of a Google smartwatch that never materialized. Then, when it finally arrived, it was a quintessential first-gen device, with thicc bezels, dismal battery life, and a host of quirks that needed ironing out. My DMs were full of people wondering when the watch would be unceremoniously dumped into Google's infamous product graveyard. A part of me wondered if Google was going to spend the next decade playing catch-up. Fast forward to 2025, and I'm holding the Pixel Watch 4 at Google's office in New York City. On the surface (and my wrist), it doesn't look like much has changed. But after fiddling with a few menus, watching some demos, and talking over the updates, it's evident that Google has a clear vision about where smartwatches are going. [...] Starting with hardware, the Pixel Watch 4 has a new domed "Actua 360" display -- as in, the display itself, not just the glass, is also domed. What this translates to is about 10 percent more visible screen space, 15 percent thinner bezels, and a 50 percent increase in maximum brightness to 3,000 nits. On a table, there's a lineup of the Pixel Watch 2, 3, and 4 with the flashlight app turned on. Side-by-side, the improvements are striking. Material 3 Expressive in Wear OS 6also helps emphasize the Pixel Watch's roundness. (No squircles here, folks.) The widgets have more rounded edges, and each screen has been redesigned to be more glanceable, fitting more complications. It's not Liquid Glass, but there are subtle animations when flitting through menus that call your attention to the Pixel Watch's rain droplet-inspired design. Altogether, it's a design tweak that makes senseandis aesthetically pleasing. Google also says battery life has improved. The 41mm watch gets an estimated 30 hours on a single charge, while the 45mm gets 40 hours. That can stretch up to two days in battery saver mode for the smaller watch and three days for the larger one. I couldn't test that at a hands-on, but I did get to see the improved fast charging in action. As with theGalaxy Watch 8, Gemini has a big presence on the Pixel Watch 4. It replaces Google Assistant and is capable of more complex queries -- even if none have been able to blow my mind yet. But, in a bid to make interacting with Gemini as smooth as possible, the speaker and haptic engines have also been updated so you can hear and interact more easily. There's also a new raise-to-talk gesture that lets you speak to Gemini without having to use the wake word. The processor has been upgraded to the Snapdragon W5 Gen 2 to enable more on-device AI features, as well, like smart replies. On the Pixel Watch 4, you'll get more smart reply options to texts that refer to the content of your conversations. They're not confined to the default Messages app, either. But the major AI update this time around is a Gemini-powered health coach that's slated to arrive alongside a revamped Fitbit app in October. ... The gist is the health coach will act more like a personal trainer than a Captain Obvious summary generator. If you sleep poorly, it'll adjust workout suggestions. (This is also why Google is also introducing an improved sleep algorithm.) You can tell it that you've been injured, and that too will be taken into consideration when generating weekly fitness plans. [...] Another big first is the Satellite SOS mode. If you're without your phone and in a remote area with no signal, you can still call emergency services. (So long as you have the LTE version of the watch.) The big thing here is that there's no extra subscription cost. The watch will also feature more accurate dual-frequency GPS -- a nice update given that I've had issues with the Pixel Watch's GPS maps in the past. The Pixel Watch 4 is priced at $349.99 and is available for pre-order now.

Read more of this story at Slashdot.

Google announced Gemini for Home, a new AI-powered voice assistant that will replace Google Assistant on Nest smart speakers and displays starting in October. Powered by Gemini's advanced reasoning and conversational capabilities, it promises more natural interactions, complex task handling, and features like Gemini Live for back-and-forth conversations. The Verge reports: According to a blog post by Anish Kattukaran, chief product officer of Google Home and Nest, using Gemini for Home will "feel fundamentally new." He says the new voice assistant leverages the "advanced reasoning, inference and search capabilities" of Google's AI models, along with adaptations for the home that allow for more natural interactions to complete more complex tasks. In short, it should be an assistant that can better understand context, nuance, and intention -- a complete change from its predecessor. For example, Kattukaran says Gemini for Home can accurately respond to requests like "turn off the lights everywhere except my bedroom," "play that song from this year's summer blockbuster about race cars," or "set a timer for perfectly blanched broccoli." It will also create lists, calendar entries, and reminders more easily than before, he says. Another big upgrade is that Gemini Live will be part of Gemini for Home, bringing more conversational back-and-forth voice interactions to Google Home without needing to repeatedly say "Hey Google." Kattukaran says this will allow for more detailed and personalized help -- from cooking ("I have spinach, eggs, cream cheese, and smoked salmon in the fridge. Help me make a delicious meal") to brainstorming how to buy a new car or figuring out how to fix your dishwasher, as well as more creative tasks like generating bedtime stories. [...] Google hasn't announced pricing for the paid tier of Gemini for Home, but Gemini Live, with its more advanced capabilities, is a likely candidate for a premium plan.

Read more of this story at Slashdot.

Sous le coude pour lecture ultérieure.
(Permalink)

IBM and Google report they will build industrial-scale quantum computers containing one million or more qubits by 2030, following IBM's June publication of a quantum computer blueprint addressing previous design gaps and Google's late-2023 breakthrough in scaling error correction. Current experimental systems contain fewer than 200 qubits. IBM encountered crosstalk interference when scaling its Condor chip to 433 qubits and subsequently adopted low-density parity-check code requiring 90% fewer qubits than Google's surface code method, though this requires longer connections between distant qubits. Google plans to reduce component costs tenfold to achieve its $1 billion target price for a full-scale machine. Amazon Web Services quantum hardware executive Oskar Painter told FT he estimates useful quantum computers remain 15-30 years away, citing engineering challenges in scaling despite resolved fundamental physics problems.

Read more of this story at Slashdot.

Google is rolling out a new feature called "Preferred Sources" in the U.S. and India, which allows users to select their preferred choice of news sites and blogs to be shown in the Top Stories section of Google's search results. From a report: Enabling this feature means you will see more content from the sites you like, the company says. When users search for a particular topic, they will see a "star" icon next to the Top Stories section. They can tap on that icon and start adding sources by searching for them. Once you select the sources, you can refresh the results to see more content from your selected sources. Google said that for some queries, users will also see a separate "From your sources" section below the Top Stories section.

Read more of this story at Slashdot.

South Korea postponed a decision for the second time this year on Friday regarding Google's request to export detailed mapping data to overseas servers, which would enable full Google Maps functionality in the country. The inter-agency committee extended the deadline from August to October to allow further review of security concerns and consultations with industry stakeholders. South Korea remains one of only a handful of countries alongside China and North Korea where Google Maps fails to function properly, unable to provide directions despite displaying landmarks and businesses. Tourism complaints increased 71% last year, with Google Maps accounting for 30% of all app-related grievances, while local industry groups representing 2,600 companies report 90% opposition to Google's request due to fears of market domination by the US tech company.

Read more of this story at Slashdot.