Google today unveiled significant upgrades to its Gemini 2.5 AI models, introducing an experimental “Deep Think” reasoning mode for 2.5 Pro that allows the model to consider multiple hypotheses before responding. The new capability has achieved impressive results on complex benchmarks, scoring highly on the 2025 USA Mathematical Olympiad and leading on LiveCodeBench, a competition-level coding benchmark. Gemini 2.5 Pro also tops the WebDev Arena leaderboard with an ELO score of 1420.

“Based on Google’s experience with AlphaGo, AI model responses improve when they’re given more time to think,” said Demis Hassabis, CEO of Google DeepMind. The enhanced Gemini 2.5 Flash, Google’s efficiency-focused model, has improved across reasoning, multimodality, and code benchmarks while using 20-30% fewer tokens. Both models now feature native audio capabilities with support for 24+ languages, thought summaries, and “thinking budgets” that let developers control token usage. Gemini 2.5 Flash is currently available in preview with general availability expected in early June, while Deep Think remains limited to trusted testers during safety evaluations.

Read more of this story at Slashdot.

There are some real-world use cases out there where AI could genuinely be a useful tool that doesn’t just act as a shortcut to something you could do on your own. Case in point: Google’s new ‘try it on’ feature, which allows you to see how various outfits look on your body over your phone or computer. The company announced the feature at Google I/O 2025, and it uses a photo you upload to generate images of you in different clothes. This feature is available via Google Search Labs starting today.

How to use Google’s ‘try it on’ feature

To try out this new feature, go to Google’s Search Labs page, sign in to your account, and opt in to the “try on” experiment. The next time you’re using Google to shop, you’ll see a ‘try it on’ icon next to different product listings. Tap this to get started.

Google will ask you to upload a full-length photo of yourself, and once you do, the AI will generate images of you wearing the outfits you’re checking out. It’s worth noting that this will work best if you upload a photo of you that’s been shot in good lighting and while you’re wearing fitted clothing. Otherwise, Google’s AI’s will have a hard time mapping out your body’s proportions, which may distort the final results. Once you’ve seen how you look in various outfits, you can share those photos with others for their opinions, or go right ahead and buy the clothes you liked.

The catch here is that you’re relying on the product photos themselves to be accurate. No matter how good Google’s AI is, it can’t really show you how something will look on you if the listing has misleading or inaccurate photos. 

Ask Google’s AI to track prices and buy products for you

During a demo at Google I/O 2025 showing off the “try it on” feature, the company also revealed that you” eventually be able to ask Google’s AI to buy products for you. Once you’ve selected an outfit you like, you can ask Google’s shopping AI to track its price. When it hits the price you’re looking for, Google will send you an alert, and you can ask the AI to buy it for you using information in your Google Wallet (or you can buy it manually).

Google is calling this feature “agentic checkout,” and the company said it’ll roll out in the coming months. 

The last time we talked about 23andMe’s bankruptcy sale, we suggested you might want to delete your genetic data from the site, since we didn’t know who would end up owning it. Now, a buyer has been announced, and they’re promising to “prioritize the privacy, security and ethical use” of customer data and to keep offering the company’s services uninterrupted. This is all—probably—good news. I’m feeling cautiously optimistic, anyway.

Regeneron will be 23andMe’s new owner

23andMe’s new buyer, paying $256 million for the company’s assets, is Regeneron. Regeneron is a biotech company perhaps best known for developing an antibody treatment for COVID early in the pandemic. (Donald Trump was given a dose when he first came down with the virus.) That treatment never made it all the way to market, but the company does market other antibody- and protein-based treatments for conditions like Ebola virus, genetic disorders, and cancers. 

Regeneron’s website states that they “are shaping the next frontier of medicine with data-powered insights from the Regeneron Genetics Center® and pioneering genetic medicine platforms, enabling us to identify innovative targets and complementary approaches to potentially treat or cure diseases.” 

That explains why they’re interested in 23andMe, since it provides a trove of genetic data. Many 23andMe users had also signed up to provide more of their personal medical information for research purposes (this was a separate thing that you would have had to opt in to provide). 

Regeneron says they plan to “continue all consumer genome services uninterrupted,” rather than shut down the company. Lemonaid health, also owned by 23andMe, is not included in the sale.

What this means for your data and privacy

Importantly, Regeneron says they will respect the company’s privacy policy (“and applicable laws”) and the 23andMe press release also says that Regeneron will not be making any changes to the privacy policy. 

The sale, which still needs to be approved by a bankruptcy court, will also include a court-appointed “consumer privacy ombudsman” whose job is to make sure that everyone’s data is treated appropriately. Regeneron says that they’re ready to work with this ombudsman and will detail all their privacy-related plans. 

While we don’t yet know what the future holds, this all has me tentatively optimistic. Yes, a corporation has your data and intends to profit from it; but that was also true of 23andMe in its heyday. The policies about privacy and consent that you agreed to when contributing your data will still be in effect. 

The 23andMe community seems to be cautiously optimistic. In a r/23andme Reddit thread about the sale, one redditor, who identifies themselves as an academic biomedical researcher, says “I would rather [Regeneron] have my data than an insurance provider or just random [venture capitalist].” Another redditor says, “So there really is not [a] best case scenario here, there’s just ‘wait and see’ and bad. And this is more of a ‘wait and see’ than a bad.” And another says “I know people side eye big pharma (rightfully in most circumstances) but this is a much better outcome than many other situations.” 

The popular Amazon Fire Stick HD is on sale for $20 as part of a Memorial Day promotion. This is a discount of 43 percent, as the usual price is $35.

This particular stick made our list of the best streaming devices, and we particularly recommend it for the budget-conscious. The Fire Stick HD supports 1080p streaming at 60 FPS and includes an Amazon remote that can control the power and volume on the actual TV. This is something the previous iteration, the Fire TV Stick Lite, couldn’t do. The days of having to navigate multiple remotes are over.

The remote also integrates with Alexa for voice control. Amazon’s Fire TV interface supports all of the major streaming services, including Netflix, HBO Max, Hulu and all the rest. The Fire TV Stick HD ships with the updated interface that includes user profiles and a simpler UI.

On the downside, this is a streaming stick for HD content. It cannot stream stuff in 4K. Also, Amazon’s interface is heavy on the ads, especially when compared to rival platforms. Still, the price is right. We enjoyed this streaming stick at the original cost, so we definitely recommend it at $20. Amazon’s Fire TV Stick 4K Max is also on sale for $40, which is a discount of 33 percent. 

Follow @EngadgetDeals on X for the latest tech deals and buying advice.

This article originally appeared on Engadget at https://www.engadget.com/deals/our-favorite-budget-streaming-stick-drops-to-only-20-for-memorial-day-185649504.html?src=rss

Microsoft is updating Windows 11 with a set of new encryption algorithms that can withstand future attacks from quantum computers in an attempt to jump-start what’s likely to be the most formidable and important technology transition in modern history.

Computers that are based on the physics of quantum mechanics don’t yet exist outside of sophisticated labs, but it’s well-established science that they eventually will. Instead of processing data in the binary state of zeros and ones, quantum computers run on qubits, which encompass myriad states all at once. This new capability promises to bring about new discoveries of unprecedented scale in a host of fields, including metallurgy, chemistry, drug discovery, and financial modeling.

Averting the cryptopocalypse

One of the most disruptive changes quantum computing will bring is the breaking of some of the most common forms of encryption, specifically, the RSA cryptosystem and those based on elliptic curves. These systems are the workhorses that banks, governments, and online services around the world have relied on for more than four decades to keep their most sensitive data confidential. RSA and elliptic curve encryption keys securing web connections would require millions of years to be cracked using today’s computers. A quantum computer could crack the same keys in a matter of hours or minutes.

Read full article

Comments

Google has begun rolling out AI Mode to every Search user in the US. The company announced the expansion during its I/O 2025 conference. Google first began previewing AI Mode with testers in its Labs program at the start of March. Since then, it has been gradually rolling out the feature to more people, including in recent weeks regular Search users. At its keynote today, Google shared a number of updates coming to AI Mode as well, including some new tools for shopping, as well as the ability to compare ticket prices for you and create custom charts and graphs for queries on finance and sports. 

For the uninitiated, AI Mode is a chatbot built directly into Google Search. It lives in a separate tab, and was designed by the company to tackle more complicated queries than people have historically used its search engine to answer. For instance, you can use AI Mode to generate a comparison between different fitness trackers. Before today, the chatbot was powered by Gemini 2.0. Now it’s running a custom version of Gemini 2.5. What’s more, Google plans to bring many of AI Mode’s capabilities to other parts of the Search experience.

“AI Mode is where we’ll first bring Gemini’s frontier capabilities, and it’s also a glimpse of what’s to come,” the company wrote in a blog post published during the event. “As we get feedback, we’ll graduate many features and capabilities from AI Mode right into the core search experience in AI Overviews.”

Looking to the future, Google plans to bring Deep Search, an offshoot of its Deep Research mode, to AI Mode. Google was among the first companies to debut the tool in December. Since then, most AI companies, including OpenAI, have gone on to offer their take on Deep Research, which you can use to prompt Gemini and other chatbots to take extra time to create a comprehensive report on a subject. With today’s announcement, Google is making the tool available in a place where more of its users are likely to encounter it.

Another new feature that’s coming to AI Mode builds on the work Google did with Project Mariner, the web-surfing AI agent the company began previewing with “trusted testers” at the end of last year. This addition gives AI Mode the ability to complete tasks for you on the web. For example, you can ask it to find two affordable tickets for the next MLB game in your city. AI Mode will compare “hundreds of potential” tickets for you and return with a few of the best options. From there, you can complete a purchase without having done the comparison work yourself.

“This will start with event tickets, restaurant reservations and local appointments,” says Google. “And we’ll be working with companies like Ticketmaster, StubHub, Resy and Vagaro to create a seamless and helpful experience.”

AI Mode will also soon include the ability to generate custom charts and graphics tailored to your specific queries. At the same time, AI Mode will be more personalized in the near future, with Google introducing an optional feature allowing the tool to draw their past searches. The company will also give people the option to connect their other Google apps to AI Mode, starting with Gmail, for even more granular recommendations.

As mentioned above, Google is adding a suite of shopping features to AI Mode. Engadget has a separate post dedicated to the Shopping features Google announced today, but the short of it is that AI Mode will be able to narrow down products for you and complete purchases on your behalf – with your permission, of course.

All of the new AI Mode features Google previewed today will be available to Labs users first before they roll out more broadly.

Update, May 20 2025, 2:45PM ET: This story has been updated to preview in the intro some of the updates coming to AI Mode.

This article originally appeared on Engadget at https://www.engadget.com/ai/google-is-rolling-out-ai-mode-to-everyone-in-the-us-174917628.html?src=rss

At Google I/O 2025, the company announced that Gemini Live Camera and Screen Sharing is rolling out to iPhone starting today. This means that you’ll be able to use your phone’s camera to live stream a feed to Gemini, and have a conversation about with the AI agent about what it sees. This feature has been available on Android for some time, but it’s coming to iPhone for the first time starting today.

This could be useful for many things, such as showing Gemini a video of something in your neighborhood and asking it to identify the item. In theory, you could ask it to identify cars, trees, or important landmarks in your neighborhood, or ask it general knowledge questions about historical landmarks while you’re on holiday. Unfortunately though, Google chose an unbelievably cringey video to demonstrate what the Gemini Live Camera can do. 

It featured a person being deliberately obtuse and asking exceptionally…dumb…questions to the AI, presumably to demonstrate that Gemini isn’t a complete idiot (or, as Google said, that it’s not afraid to tell you when you’re wrong). They pointed the camera at a garbage truck while calling it a “convertible,” but Gemini was able to clarify that it was indeed, a garbage truck. They repeated the same routine by calling a light pole a “tall skinny building” and a shadow “someone following” them. If you were to only go by the demo to figure out the extent of Gemini Live’s Camera Sharing capabilities, it’d be pretty hard to know how it holds up to stress testing.

Fortunately, there are much better examples of what you can do with Gemini Live Camera on the Android blog. You can ask Gemini Live for advice on a color palette for your home, gift ideas, and much more. As long as the AI doesn’t hallucinate, this could a genuinely useful tool, since it also works with screen sharing, too. Just… don’t ask it if your shadow is a real person. 

Recently rediscovered in Mattel’s archives and remastered, this is the original 5-minute pilot episode ‘Down The Mineshaft’ for Thomas The Tank Engine and Friends. The pilot, narrated by Ringo Starr and filmed in 1983, was later remade for use as the show’s 25th episode after executives confirmed that yes, this Thomas The Tank Engine thing is, in fact, a gold mine. Since then there have been “more than 500 episodes, 207 books and 20 movies.” Damn, talk about the little engine that could! Thomas is hauling money straight to the bank. Now, are you thinking what I’m thinking? “Does it involve a train robbery?” I’ll get some bandanas, you start practicing riding a horse and poking a train in the eyes with a stick.

Google’s Smart Replies are getting an AI-themed, and to me, slightly creepy update. At its Google I/O 2025 keynote, Google announced a new feature called Personalized Smart Replies. This feature uses Gemini AI, alongside the data in your Gmail and Google Drive, to reply to emails in a way that sounds like you. Google promises is can write like you would, using your tone and writing style, as learned by your Gmail history.

With this new update, the Smart Replies option that shows up when you reply to an email will be a lot more powerful and a lot more contextual. On stage, Google CEO Sundar Pichai used planning a trip as an example. Let’s say a friend emails you asking for details about a trip you two are going on. To manually collate everything you need to send, like your hotel stays, driving plans, rest stops, and restaurant recommendations would take a lot of time.

Instead, you can use the Personalized Smart Replies feature, which has access to all your emails, your Google Maps location history, and your photos and files in Google Drive, to do the work for you. It can collate hotel bookings from your email, locations from Google Drive history, and present it all in an email that matches your writing style.

According to an interview with The Verge, Google Workspace’s VP of product, Blake Barnes, says that the new Smart Replies feature will be available in English at first, on the web, Android, and iOS. It will roll out in Alpha in Google Labs in July, and will be available for public use in Q3 of this year. Unfortunately, it will be only available for paid subscribers at first. Barnes sees the feature eventually rolling out to all Gmail users, but there’s no timeline for that yet.

Of course, like with all AI, there’s a risk that what Gemini generates might have elements of hallucination, so I’d recommend checking your replies before sending them. Plus, the feature will definitely need access to a lot of personal data to work. I’ll keep an eye out for how to protect your data as the feature comes out, and update accordingly.

Google has just announced a new $250 per month AI Ultra plan for people who want unlimited access to its most advanced machine learning features. Yes, you read that right. It means the new subscription is $50 more expensive than the already pricey ChatGPT Pro and Claude Max plans from OpenAI and Anthropic.

For $250, you’re getting early access to new models like Veo 3, and unlimited usage of features like Flow (the new AI film-making app the company announced today) and the compute-intensive Deep Research. In the coming weeks, Google will also roll out Deep Think to AI Ultra users, which is the new enhanced reasoning mode that is part of its Gemini 2.5 Pro model. Subscribers can also look forward to access to Project Mariner, Google’s web-surfing agent, and Gemini within Chrome, plus all the usual places where you can find the chatbot like Gmail and Docs.

Google is partly justifying the high cost of AI Ultra by touting the inclusion of YouTube Premium and 30TB of cloud storage across Google Photos, Drive and Gmail. On its own, a YouTube Premium subscription would cost you $14 per month, and Google doesn’t offer 30TB of cloud storage separately. The closest comparison would be Google One, which includes a Premium tier that comes with 2TB of storage for $10 per month. As another incentive to sign up for AI Ultra, Google is giving new subscribers 50 percent off their first three months.

As of today, Google is also revamping its existing AI Premium plan. The subscription, which will be known as Google AI Pro moving forward, now includes the Flow app and early access to Gemini in Chrome. Google says the new benefits will come to US subscribers first, with availability in other countries to follow.

This article originally appeared on Engadget at https://www.engadget.com/ai/google-wants-250-per-month-in-return-for-its-new-ai-ultra-plan-180248513.html?src=rss

It’s that time again, for Google to announce that real-time translation has come to one of its communication apps. This time, it’s Google Meet, which can translate between English and Spanish as you speak in a video call. If that sounds familiar, it’s because it’s not the first time Google has announced something like this.

Google Translate has had features that let you speak to someone in another language in real time for a while. For example, back in 2019, there was a real-time translation feature called Interpreter Mode built into Google Assistant. It’s also been possible on Pixel phones for a while (and even Samsung phones). Most of these, however, have been either text-to-text, or speech-to-text. You can use the Google Translate app for a speech-to-speech experience, but like with Google Assistant’s Interpreter Mode, that only works in person.

So, what’s different here? Well, during its I/O keynote, Google demoed two users in a video chat speaking in their native languages. Google Meet then translates and speaks the translation back in a relatively human-sounding voice. This new feature is available now for Google Workspace subscribers (plans start at $7/month), but unfortunately, it’s not in the free version. On the plus side, additional languages are promised to start coming out in just a few weeks.

While I haven’t tested it out yet, it does seem to be a more convenient way to access a feature that you might otherwise have to hack together with another tab, or by opening your phone and holding it up to a speaker. Plus, it can be a bit more natural to hear translations spoken out for you, rather than having to rely on translated captions. I do wonder whether it can keep up with the natural speed and flow of a conversation, though—nobody likes to feel interrupted.

We may earn a commission from links on this page.

Motorola is set to release its latest wearable, the Moto Watch Fit, on May 22 with a price tag of $199 in the U.S. ($249 in Canada). The new smartwatch, which was initially unveiled alongside updated Razr flip phones last month, will be available through Motorola’s official website at launch, with availability expanding to Amazon and Best Buy in the coming weeks.

What to know about the Moto Watch Fit

Unlike most Android smartwatches on the market, the Moto Watch Fit runs on a proprietary operating system rather than Google’s Wear OS. While this might raise some concerns for potential buyers wary of an untested system, there are a few features that could sway you in its direction.

Design

The Moto Watch Fit features a sleek design with a 1.9-inch OLED display housed in an aluminum frame. In terms of durability, the watch has 5ATM water resistance and IP68 ratings—meaning it should hold up for swimming and other water-based activities. At just 25 grams, the Moto Watch Fit offers a lightweight design that won’t weigh you down during workouts or daily wear. (For comparison, most versions of the Apple Watch Series 10 sit around 30 grams).

Impressive battery life

Where Motorola claims the watch truly stands out is its battery life. It says users can expect up to 16 days of use on a single charge and under normal conditions. However, something like constant GPS usage would naturally drain your battery much faster. However, the company promises a quick 5-minute charge can provides a full day of usage.

The Watch Fit focuses on delivering reliable fitness tracking, making it best for someone prioritizing battery life and comfort over extensive app ecosystems.

Limited smart features

On that note, it looks like the Moto Watch Fit does not support third-party apps and lacks smart assistant functionality. While this may be a dealbreaker users who want the full smartwatch experience, it’s likely these limitations contribute to the device’s extended battery life, not to mention it’s price point—at under $200, it’s one of the more affordable smartwatches. For context, the most popular versions of Galaxy, Google Pixel, or Apple watches will typically cost at least $299.

So, if you’re looking for heart rate monitoring, GPS tracking, and basic smartwatch functions, and you’re tired of power-hungry options on the market that struggle to last even a week between charges, the Moto Watch Fit could be the right fit for you.

Google is adding AI-powered live translation to Meet, enabling participants to converse in their native languages while the system automatically translates in real time with the speaker’s original vocal characteristics intact. Initially launching with English-Spanish translation this week, the technology processes speech with minimal delay, preserving tone, cadence, and expressions — creating an effect similar to professional dubbing but with the speaker’s own voice, the company announced at its developer conference Tuesday.

In some testings, WSJ found occasional limitations: initial sentences sometimes appear garbled before smoothing out, context-dependent words like “match” might translate imperfectly (rendered as “fight” in Spanish), and the slight delay can create confusing crosstalk with multiple participants. Google plans to extend support to Italian, German, and Portuguese in the coming weeks. The feature is rolling out to Google AI Pro and Ultra subscribers now, with enterprise availability planned later this year. The company says that no meeting data is stored when translation is active, and conversation audio isn’t used to train AI models.

Read more of this story at Slashdot.