Category: Technology

Google has started testing a reasoning model called Deep Think for Gemini 2.5 Pro, the company has revealed at its I/O developer conference. According to DeepMind CEO Demis Hassabis, Gemini’s Deep Think uses “the latest cutting-edge research” that gives the model the capability to consider multiple hypotheses before responding to queries. Google says it got an “impressive score” when evaluated using questions from the 2025 United States of America Mathematical Olympiad competition. However, Google wants to take more time to conduct safety evaluations and get further input from safety experts before releasing it widely. That’s why it’s making Deep Think initially available to trusted testers via the Gemini API first in order to get their feedback first. 

The company has also introduced a better Gemini 2.5 Flash model, which is optimized for speed and efficiency. It’s now more efficient than before, uses fewer tokens and has scored higher in benchmarks for reasoning, multimodality, code and long context than its predecessor. It will be generally available in early June. For now, the improved Gemini 2.5 Flash is available as a preview via Google AI Studio for developers, via Vertex AI for enterprise customers and via the Gemini app for other users. 

While most of the efficiency gains covered on the I/O stage were focused on 2.5 Flash, Google did announce that it’s bringing the 2.5 Flash concept of “Thinking Budgets” to its more advanced 2.5 Pro model. This feature will let you balance tokens spent vs. accuracy and speed of output.

Separately, Google is bringing Project Mariner into the Gemini API and Vertex AI, as well. Project Mariner is Google’s Gemini-powered AI agents that can navigate pages on the web browser to complete tasks for users. The company will roll the agents out more broadly this summer so that developers can experiment with them. In addition, the company is releasing new previews for text-to-speech on both 2.5 Pro and 2.5 Flash models via the Gemini API, with support for two voices in 24 languages. 

This article originally appeared on Engadget at https://www.engadget.com/ai/google-introduces-the-deep-think-reasoning-model-for-gemini-25-pro-and-a-better-25-flash-174531020.html?src=rss 

One of the new AI features that Google has announced for Search at I/O 2025 will let you discuss what it’s seeing through your camera in real time. Google says more than 1.5 billion people use visual search on Google Lens, and it’s now taking the next step in multimodality by bringing Project Astra’s live capabilities into Search. With the new feature called Search Live, you can have a back-and-forth conversation with Search about what’s in front of you. For instance, you can simply point your camera at a difficult math problem and ask it to help you solve it or to explain a concept you’re having difficulty grasping. 

Search will also give you links to more resources you can explore. To be able to access Search Live, you’ll have to tap the Live icon in Google Lens or in AI Mode, which the company has just rolled out in the US after making it available to Labs users earlier this month. AI Mode is a new chatbot Google built directly into Search that can break down your query into subtopics so it can serve you more thorough results with “hyper-relevant content.” The new AI Mode tab will show up in Search and in the Google app in the coming weeks. 

AI Mode will be able to help you analyze and visualize numbers and data, as well, by generating interactive graphics for you. If you want to compare two different sports teams, for instance, Google will be able to generate a graph showing their performance using its own real-time sports information. The company will also be able to create real-time graphs for finance queries. 

Soon, AI Mode will be able to give you personalized responses based on your past activities. If you ask it where you can eat while visiting a specific city, for example, it can name restaurants with outdoor seating if you’ve booked an outdoor table in the past. It can also give you suggestions based on your hotel and flight bookings. In addition, AI Mode will come with a Shopping experience that will let you try on clothes virtually by uploading a photo of yourself. When you’re done, Google’s new checkout feature will be able to buy the clothes you want in your stead from their respective websites with Google Pay. 

This article originally appeared on Engadget at https://www.engadget.com/ai/google-search-live-will-let-you-ask-questions-about-what-your-camera-sees-174532593.html?src=rss 

As part of this year’s announcements at its I/O developer conference, Google has revealed its latest media generation models. Most notable, perhaps, is the Veo 3, which is the first iteration of the model that can generate videos with sounds. It can, for instance, create a video of birds with an audio of their singing, or a city street with the sounds of traffic in the background. Google says Veo 3 also excels in real-world physics and in lip syncing. At the moment, the model is only available for Gemini Ultra subscribers in the US within the Gemini app and for enterprise users on Vertex AI. It’s also available in Flow, Google’s new AI filmmaking tool. 

Flow brings Veo, Imagen and Gemini together to create cinematic clips and scenes. Users can describe the final output they want in natural language, and Flow will go to work making it for them. The new tool will only be available to Google AI Pro and Ultra subscribers in the US for now, but Google says it will roll out to more countries soon. 

While the company has released a brand new video-generating model, it hasn’t abandoned Veo 2 just yet. Users will be able to give Veo 2 images of people, scenes, styles and objects to use as reference for their desired output in Flow. They’ll have access to camera controls that will allow them to rotate scenes and zoom into specific objects for Flow, as well. Plus, they’ll be able to broaden their frames from portrait to landscape if they want to and add or remove objects from their videos. 

Google has also introduced its latest image-generating model, Imagen 4, at the event. The company said Imagen 4 does fine details like intricate fabrics and animal fur with “remarkable clarity” and excels at generating both photorealistic and abstract images. It’s also significantly better at rendering typography than its predecessors and can create images in various aspect ratios with resolutions of up to 2K. Imagen 4 is now available via the Gemini app, Vertex AI and in Workspace apps, including Docs and Slides. Google said it’s also releasing a version of Imagen 4 that’s 10 times faster than Imagen 3 “soon.” 

Finally, to help people identify AI-generated content, which is becoming more and more difficult these days, Google has launched SynthID Detector. It’s a portal where users can upload a piece of media they think could be AI-generated, and Google will determine if it contains SynthID, its watermarking and identification tool for AI art. Google had open sourced its watermarking tool, but not all image generators use it, so the portal still won’t be able to identify all AI-generated images. 

This article originally appeared on Engadget at https://www.engadget.com/ai/googles-veo-3-ai-model-can-generate-videos-with-sound-174541183.html?src=rss 

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.

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.

Separately, 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.

This article originally appeared on Engadget at https://www.engadget.com/ai/ai-mode-is-now-rolling-out-to-everyone-in-the-us-174917212.html?src=rss 

Google’s Chrome browser is the latest major product from the company to get its own built-in Gemini features. Today at Google I/O, the company detailed its plans to bring its AI assistant to Chrome.

While Gemini can already distill information from websites, having the assistant baked into Chrome allows it to provide insights and answer questions about your open tabs without ever having to move to a different window or application. Instead, Gemini lives in a new menu at the top of your browser window as well as in the taskbar.

The company envisions its assistant as being able to help out with tasks that may normally require switching between several open tabs or scrolling around to different parts of a web page. For example, Google showed off how Gemini can give advice about potential modifications for dietary restrictions while looking at a recipe blog. Gemini in the browser could also come in handy while shopping as it can answer specific questions about products or even summarize reviews.

To start, Gemini will only be able to answer queries about a single open tab, but the company plans to add multi-tab capabilities in a future update. This would allow the assistant to synthesize info across multiple open tabs and answer even more complex questions. Gemini in Chrome will also have Gemini Live capabilities, for anyone more comfortable conversing with the assistant using their voice. The company also teased a future update that will allow Gemini to actually scroll through web pages on your behalf, like asking it to jump to a specific step in a recipe. (Notably, all this is separate from Google’s other web-browsing AI, Project Mariner, which is still a research prototype.)

Gemini is starting to roll out to Chrome users on Mac and Windows today, beginning with AI Pro and AI Ultra subscribers in the United States. The company hasn’t indicated whether it plans to bring similar features to Chromebooks or Chrome’s mobile app.

This article originally appeared on Engadget at https://www.engadget.com/ai/google-is-bringing-gemini-to-chrome-so-it-can-answer-questions-about-your-open-tabs-174903787.html?src=rss 

Google Chrome has announced a feature for its built-in password manager that it claims will let users instantly change passwords compromised in data breaches. Google Password Manager already alerts you when your credentials have appeared in a data breach, and partially automates the process of changing your password, but — until now — you still had to go through the steps manually for each of your online accounts.

The Automated Password Change feature, announced at today’s Google I/O keynote presentation, goes a step farther. It will apparently let you generate a new password and substitute it for the old one with a single click, without ever seeing a “Create New Password” page. The feature only works on participating websites. Google is currently in talks with developers to expand the range of sites that will support one-click password changes, with plans for a full rollout later in 2025.

Automated Password Change was discovered as far back as February by eagle-eyed software diggers, but was limited to the early developer-only builds made public as Chrome Canary. At that time, it was located in the “AI Innovations” settings menu, though it’s not yet clear how AI figures in the process.

This feature builds on password health functionality that Google has been steadily incorporating into Chrome since it released the Password Checkup extension in 2019, recognizing that compromised credentials are a common vector for cybercrime. People often reuse the same short, memorable password on multiple websites. If hackers steal a credential database from one weakly defended site and dump it on the dark web, other cybercriminals can try the leaked usernames and passwords on more secure sites — like online banks and cash apps — until one fits.

The best way to prevent this is to use a password manager to generate and save a different strong password for every account you make, even ones you don’t think will handle sensitive information. If you haven’t done this, the second-best prevention is to monitor password data breaches and immediately change any password that gets leaked. If Automated Password Change works as advertised, it’ll make that crisis response a lot more convenient.

This article originally appeared on Engadget at https://www.engadget.com/cybersecurity/google-chrome-previews-feature-to-instantly-change-compromised-passwords-175051933.html?src=rss 

At I/O today, Google pitched creators on a new app for “AI filmmaking”: Flow. Combining all of Google’s recent announcements and developments across AI-powered services, including Veo (video), Imagen (images) and Gemini, the company bills Flow as a storytelling aid ‘built with creatives.” If it sounds familiar, this is the advanced version of VideoFX, previously a Google Labs experiment.

It says it’s aimed at helping storytellers to explore ideas and create clips and scenes, almost like storyboards and sketches in motion. Google’s generally impressive Veo 2 model seems to form the core of Flow, able to extend footage and create video that “excel(s) at physics and realism”, although I’m not sure many agree with that..

You can use Gemini’s natural language skills to construct and tweak the video output, and creatives can pull in their own assets or create things with Imagen through simple text input. What’s notable is the ability to integrate your creations and scenes into different clips and scenes with consistency. While the early demo footage we saw was impressive, it still had a not-so-faint AI-slop aroma.

There are further film-making tools, too. Flow will also feature direct control over the movement of your ‘camera’, and even choose camera angles. You can also edit and extend shots, adding different transitions between AI-generated videos. Creating video with Veo is often a piecemeal process, but Flow will have its own asset management system to organize assets and even your prompts. These richer controls and editing abilities could make for more compelling creations in time. Let’s not forget: It’s been less than a year since that very weird Toys R’Us ad.

Google buddied up with several notable filmmakers to attempt to legitimize collaborate on these still-early steps into AI video creation, including Dave Clark, Henry Daubrez and Junie Lau. It says it offered creatives early access to the tools, and folded in their insights and feedback into what is now called Flow.

Flow is now available to AI Pro and AI Ultra subscribers in the US, and will roll out to other countries soon. Pro users will get Flow tools outlined so far and 100 generations each month. With the Ultra sub, you’ll get unlimited generation and early access to Veo 3, with native audio generation.

This article originally appeared on Engadget at https://www.engadget.com/ai/google-filmmaking-tool-flow-ai-generated-video-175212520.html?src=rss 

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 

Amazon will now deliver a bunch of new products via drone, as mentioned in a company blog post. This includes stuff like smartphones, video doorbells and food thermometers. This is thanks to the Federal Aviation Administration (FAA) giving approval to several new item categories.

Newly-available items include Apple iPhones, AirTags and AirPods, in addition to Samsung Galaxy smartphones, Ring doorbells and Alpha Grillers Instant Read Food Thermometers. All told, customers can now choose from more than 60,000 items for drone delivery.

These deliveries will typically arrive within an hour, assuming the package is under five pounds and the customer lives in a drone service area. Unfortunately, these areas are extremely limited. As of this writing, Amazon only offers drone deliveries to parts of Texas and Arizona. The company plans on expanding to the UK and Italy in the near future.

Weather also plays a part when determining the availability of a drone delivery. Amazon won’t offer the service “during heavy winds, or during heavy rain,” the company wrote in another blog post. It isn’t being overly cautious here. The drones can get into serious trouble when trying to navigate foul weather, as shown by recent crashes that forced the company to put the platform on hold for a while.

Last year, the FAA gave Amazon permission for its drones to fly beyond the visual line of sight (BVLOS). This has allowed the company to use its newest MK30 model to make deliveries within an hour.

This article originally appeared on Engadget at https://www.engadget.com/transportation/amazon-drones-can-now-deliver-smartphones-airtags-and-more-164413100.html?src=rss