Google is rapidly expanding its AI search capabilities, as reflected in the announcements it made Tuesday at its Google I/O developer conference. The search giant announced the general availability of AI Mode, its chatbot-format AI search product; some changes to its AI Overviews search results; and its plans to add new visual and agentic search features this summer.  Google’s biggest announcement in the realm of search was the general availability of its AI Mode, a chatbot-style search interface that allows users to enter a back-and-forth with the underlying large language model to zero in on a complete and satisfying answer. “AI Mode is really our most powerful version of AI search,” Robby Stein, Google’s VP of Product for Search, tells Fast Company. The tool had been available as an experimental product from Google Labs. Now it’s a real product, available to all users and accessible within various Google apps, and as a tab within the Google mobile app.  AI Mode is powered by Gemini 2.5, Google’s most formidable model, which was developed by DeepMind. The model can remember a lot of data during a user interaction, and can reason its way to a responsive answer. Because of this, AI Mode can be used for more complex, multipart queries. “We’re seeing this being used for the more sophisticated set of questions people have,” Stein says. “You have math questions, you have how-to questions, you want to compare two products—like many things that haven’t been done before, that are probably unique to you.”  The user gets back a conversational AI answer synthesized from a variety of sources. “The main magic of the system is this new advanced modeling capability for AI Mode, something called a query fan-out where the model has learned to use Google,” Stein says. “It generates potentially dozens of queries off of your single question.” The LLM might make data calls to the web, indexes of web data, maps and location data, product information, as well as API connections to more dynamic data such as sports scores, weather, or stock prices.  New shopping tools  Google also introduced some new shopping features in AI Mode that leverage the multimodal and reasoning capabilities of the Gemini 2.5 models. Google indexes millions of products, along with prices and other information. The agentic capability of the Gemini model lets AI Mode keep an eye out for a product the user wants, with the right set of desired features and below a price threshold that the user sets. The AI can then alert the user with the information, as well as a button that says “buy for me.” If the user clicks it the agent will complete the purchase.  Google is also releasing a virtual clothing try-on function in AI Mode. The feature addresses perhaps the biggest problem with buying and selling apparel online. “It’s a problem that we’ve been trying to solve over the last few years,” says Lilian Rincon, VP of Consumer Shopping Products. “Which is this dilemma of [where] users see a product but they don’t know what that product will look like on them.” Virtual Try-on lets a user upload a photo of themself, then the AI shows the user what they’d look like in any of the billions of clothing products Google indexes. The feature is powered by a new custom image generation model for fashion that understands the nuances of the human body and how various fabrics fold and bend over the body type of the user, Rincon says. Google has released Virtual Try-on as an experimental feature in Google Labs.  New features coming to AI Mode this summer Google says it intends to roll out further enhancements to AI Mode over the summer.  For starters, it’s adding the functionality of its previously announced Project Mariner (an AI agent prototype that works with the Chrome browser) to AI Mode. So the LLM will be able to control the user’s web browser to access information from websites, fill out and submit forms, and use websites to plan and book tasks. Google is going to start by enabling the AI to do things like book event tickets, make restaurant reservations, and set appointments for local services.  The user can give the AI agent special instructions or conditions, such as “buy tickets only if less than $100, and only if the weather (if its an outdoor event) forecast looks good.” The AI will not only find the best ticket prices to a show, but will also submit the data needed to buy the ticket for the user. (The user gets final sign-off, of course.)  Google will be adding a new “deep search” function in which the model might access, and reason about, hundreds of online, indexed, or AI data sources. The model might spend several minutes thinking through the completeness of its answer, and perhaps make additional data queries. The end result is a comprehensive research report on a given topic.   At last year’s I/O, Google revealed its Project Astra, a prototype of a universal AI assistant that can see, hear, and reason, and converse with the user, out loud, in real time. The assistant taps into search in several ways. A user could show the assistant an object in front of the phone camera and ask for more information about it, which the agent would get from the web. Or the assistant might be shown a recipe, and help the user shop for the ingredients.  Google also plans to launch enhanced personalization features to AI search as a way of delivering more relevant search results. “The best version of search is one that knows you well,” Stein says. For example, AI Mode and AI Overviews soon might consult a user’s search history to use past preferences to inform the content of current queries. That’s not all. Google also intends to consult user data from other Google services, including Gmail, to inform searches, subject to user opt-in.  Finally, the company will add data visualizations to search results, which it believes will help users draw meaning from data returned in search results. It will start by modeling sports and financial data this summer, Stein says.  AI Overviews now reaches almost all Google users AI Overviews is Google’s original AI search experience. For some types of search queries, users see an AI-generated narrative summary of information synthesized from various web documents and Google’s information graphs. Stein says Google is now making AI Overviews available to 95 more countries and territories, bringing the total to around 200, and in 40 languages. Google claims that AI Overviews, its generative AI search experience, now has 1.5 billion users.  Where search is concerned, Google is a victim of the “inventor’s dilemma.” It built a massive business placing ads around its search results, so it has a good reason to keep optimizing and improving that experience, rather than pivoting toward new AI-based search, which nobody has reliably monetized with ads yet. Indeed Google’s core experience still consists of relatively short queries and results consisting of ranked websites and an assortment of Google-owned content. But development of AI search products and functions seems to be accelerating. Google is protecting its cash cow (traditional search with ads) while keeping pace with the chatbot search experiences offered by newcomers like OpenAI’s ChatGPT.  But it’s more than that. Google’s VP and Head of Search Liz Reid suggests that we may be looking at the future of Google Search—full stop. “I think one of the things that’s very exciting with AI Mode is not just that it is our cutting-edge AI search, but it becomes a glimpse of what we think can be more broadly available,” Reid tells Fast Company. “And so our current belief is that we’re going to take the things that work really well in AI Mode and bring them right to the core of search and AI Overviews.” 

Commonly available glow-in-the-dark materials are made from phosphors, which are chemicals like zinc sulfide or strontium aluminate, that manufacturers mix with paint or plastics. These phosphors absorb energy from sunlight or even artificial light, then give it off over time; that process, known as phosphorescence, is what causes them to glow. For some objects, it is not desirable to rely on phosphors, since they need to be "charged" (i.e. regularly exposed to light in order to produce their glowing powers). For example, there are gunsights and watch faces made with tritium, a radioactive isotope of hydrogen, which glows continuously. The radioactivity emitted by a tritium gunsight or watch face is so weak that it cannot penetrate human skin, thus is deemed to be safe; manufacturers are quick to point out that you encounter more radiation just walking around on our planet. TEC Accessories, an EDC gear company, sells this line of Isotope Tritium Fobs. These are glow-in-the-dark key fobs that, you guessed it, rely on tritium for that always-on glow. There is a bit of a catch: TEC Accessories can sell you the key fob--but they can't actually sell you the tritium. Tritium is illegal to sell in the U.S., where the company is based. Thus they've got an entire page on their website describing how to skirt U.S. laws to get your hands on the stuff. Once you've ordered and received your vial of tritium, you can pop it into their key fob, and off you go. You're probably wondering: Aren't radioactive materials dangerous? As with gunsights and watch faces, the answer is that it might be—if you ingested it or inhaled it. Contained within a vial that is inserted into a titanium key fob, you'd have to go to a lot of trouble to do either of those things.You'd also have to go to a lot of trouble to order tritium. But darn if these things don't look cool. The fobs run $14.50 to $80 a pop.

TECNO brings a full suite of AI-powered devices to COMPUTEX 2025, showcasing how its self-developed edge-side AI model transforms everyday technology interactions. The company returns to Taipei with products spanning laptops, smart glasses, and wearables under its “Mega Leap with AI” theme. Designer: TECNO The centerpiece of their showcase is the new MEGABOOK S16 AI PC, complemented by the world’s lightest 14-inch OLED laptop weighing just 899 grams. These devices represent TECNO’s vision for AI that works seamlessly both online and offline, addressing the growing need for intelligent computing that adapts to users rather than forcing adaptation. MEGABOOK S16: Flagship AI Performance in a Premium Package The MEGABOOK S16 integrates TECNO’s self-developed edge-side AI model, enabling AI functionality even without internet connectivity. Powered by Intel’s Core i9-13900HK processor with 14 cores and 20 threads reaching 5.4 GHz turbo speeds, the system delivers substantial computational power for demanding AI applications. Despite its performance capabilities, the S16 maintains a surprisingly portable profile at just 1.3kg and 14.9mm thick. The all-metal chassis houses TECNO’s first 16-inch display in the flagship laptop line, responding to user demand for larger screens without sacrificing mobility. The system particularly excels in multitasking scenarios where AI assistance proves most valuable. Users can seamlessly switch between creative work, productivity tasks, and entertainment without the performance degradation typically associated with running multiple demanding applications. MEGABOOK S14: Redefining Ultralight Computing Perhaps more impressive from an engineering standpoint is the MEGABOOK S14, which achieves a remarkable 899g weight while incorporating a 2.8K OLED display. TECNO offers the system in two variants: one with Qualcomm’s Snapdragon X Elite compute platform and another with Intel’s Core Ultra 9 processor. The magnesium-alloy chassis contributes to the ultralight design without compromising structural integrity. For users requiring additional graphics performance, TECNO provides an external graphics dock with NVIDIA GPU options that transforms the ultraportable into a creative workstation or gaming system. The S14 represents TECNO’s first OLED implementation in a laptop, delivering 2.8K resolution with a 120Hz refresh rate and 91% screen-to-body ratio. The display carries TÜV Rheinland eye comfort certification for extended viewing sessions. K-Series: Accessible AI Computing TECNO also showcases its K15S and K14S models, representing new size options in its entry-level lineup. The K15S features an all-metal design with a 15.6-inch display, Intel Core i5-13420H processor, and expandable memory up to 32GB. Despite its more accessible positioning, the K15S incorporates a substantial 70Wh battery with 65W GaN fast charging technology, addressing a common pain point in the category. The system includes a full-sized keyboard with numeric keypad and four-level backlighting for productivity in various lighting conditions. AI Capabilities Across the Lineup All MEGABOOK models now feature TECNO’s upgraded AI model with DeepSeek-V3, enhancing offline capabilities while enabling comprehensive online AI searches through a Personal GPT function. The system offers six core AI functionalities designed to streamline common workflows. The AI Gallery, which TECNO claims is a world-first on Windows, connects wirelessly with TECNO smartphones for photo backup, smart album creation, and image searches. The Ella AI Assistant manages tasks and schedules, while AI PPT localizes and completes presentations using TECNO’s AI sources. For professionals, the AI Meeting Assistant provides real-time transcription with speaker identification and key point extraction. The system also includes AI Drawing tools for creative applications. AI Glasses: Smartphone-Grade Photography in Eyewear Moving beyond computing, TECNO introduces its first AI Glasses series with two models: AI Glasses and AI Glasses Pro. Both incorporate a 50MP camera system using the same OV50D sensor, ISP, and imaging algorithms found in TECNO’s flagship CAMON 40 Premier smartphone. The glasses feature a “SmartSnap” function that recognizes scenes and automatically generates captions for social sharing. The AI Info function compiles notifications from multiple apps into concise reports, while real-time translation supports over 100 languages. The Pro model adds WaveGuide AR display technology co-developed with Meta-Bounds, featuring a MicroLED screen with 30° field of view and 1500 nits brightness. This enables navigation overlays, meeting translations, and other augmented reality applications. Both models offer approximately 8 hours of mixed use on a 30-minute charge of their 250mAh batteries. The standard model features an aviator design, while the Pro adopts a browline style. Ecosystem Integration and Market Positioning TECNO emphasizes the interconnectivity of its AI ecosystem through OneLeap technology, which enables multi-screen sharing, file transfer, and cross-device collaboration between MEGABOOK laptops, smartphones, and tablets. This approach addresses a common friction point for users working across multiple devices, allowing content and context to follow the user rather than remaining siloed on individual devices. TECNO positions its AI ecosystem as democratizing advanced technology for emerging markets, with a presence in over 70 markets across five continents. Their “Stop At Nothing” brand philosophy guides product development toward accessible innovation. The comprehensive lineup demonstrates TECNO’s commitment to AI as a transformative technology rather than a marketing checkbox. By developing its own edge-side AI model, the company maintains control over the user experience while ensuring functionality even in regions with inconsistent connectivity. For users seeking to experience TECNO’s vision of AI-enhanced computing, the company’s booth at COMPUTEX 2025 (N1302) showcases all products through May 23rd.The post TECNO Unveils Comprehensive AI Ecosystem at COMPUTEX 2025 first appeared on Yanko Design.

Google’s new chatbot-style AI Mode search experience, previously an experiment, is launching for US users. Publishers and marketers will have to adjust their search strategies once again.

The tech giant is taking its next big step in artificial intelligence by adding interactive capabilities to its flagship product.

Macworld Apple’s newest iPad is on sale for its best price yet, which means you can get Apple’s budget tablet for just $299. That’s a savings of $50 you can put towards something else you’ve been waiting to buy. There are about a million things you can do with your new tablet, starting with streaming all your favorite shows and movies. The 11-inch Liquid Retina Display will ensure you’ll get gorgeous images in vivid color. It also comes with 128GB of storage space—twice that of the previous model. Considering how large some apps and games are, that’s a welcome change across iPads. We gave the new iPad a 4-star rating, appreciating just how bright the screen is, as well as the overall high-quality design and workmanship of the tablet. Even though the A16 doesn’t support Apple Intelligence, we still appreciated the strong performance Apple’s most affordable tablet delivers. Oh, and did we mention the battery life? Our tests involve running a TV show on a loop with the screen set to max brightness and with True Tone disabled. The tablet lasted for six hours and ten minutes before shutting down, which means you’ll likely get more than 10 hours of real-world use. That’s plenty to get through a busy day of gaming and video watching. Apple deals don’t stick around for long, so our advice is to hurry up and get your own A16 iPad for $299 before the deal vanishes. Save $50 on the iPad A16Buy now at Amazon

Do you want to make a podcast from notes you record on your iPhone? You can, as Google has introduced an iOS version (and an Android version) of its popular NotebookLM tool, which can do this, among other things. The news follows hot on the heels of speculation that Apple may try to overcome shortcomings in its own AI development by opening up its platform to third-party AI services in addition to ChatGPT and Apple Intelligence. It may be relevant to point out that Apple this week made it possible to use Google Translate instead of Apple’s own Translate app on iPhones. What is NotebookLM? Notebook LM has won a ton of praise since it appeared. It is a really useful document summarization system that is very handy for researchers — and can even turn topics you write about into engaging and thought-provoking podcasts. The service achieves this through use of Google’s Gemini genAI system, which seems to be improving rapidly when it comes to focused tasks. “We’ve received a lot of great feedback from the millions of people using NotebookLM, our tool for understanding and engaging with complex information. One of the most frequent requests has been for a mobile app — for listening to Audio Overviews on the go, asking questions about sources in the moment, and sharing content directly to NotebookLM while browsing,” said Google when it announced the new apps. Making a podcast on your iPhone NotebookLM has been available as a web app, and now also as an app for iOS and Android. Once installed, you can use the mobile app to create new notes and access those you may already have created via your Google account. You can also add new sources to notes and create podcasts of those notes. But one of the best new features is the ability to get involved in the podcast/conversation.  Tap the Join button and you can interact with the AI-generated hosts, asking them questions or steering the conversation. It’s remarkable, particularly if you are still trying to explain which song you want Siri to play in Apple Music.  It shows the extent to which Apple’s AI services are playing catch-up and may also be why Apple’s management is thinking about opening up the company’s platform to third-party AI services. Is this how Apple will make AI services optional? The move to make Google Translate an option for users shows how that may be done. Just as Apple is being forced to permit users to choose between browsers in Europe, the options tool for Translation lets you select which service to use for that.  Finding a way to offer these choices while preserving platform integrity is easier said than done. Apple has admitted to having thousands of engineers tasked with figuring out how to make that possible.  But as the company moves forward with developing solutions that deliver such choice, it is also creating the template we will probably see it follow as it moves to offer up support for different forms of AI services on its devices. With that in mind, it is likely that, as AI services introduce apps for Apple’s systems, the company will introduce a new setting in which users will be able to choose what service to use. It is possible that Apple will need to keep Apple Intelligence as the first point of contact, acting as a kind of concierge for queries, which it then directs to an appropriate AI. Users will then select which service will offer the default AI. What about people who don’t want to use these services? For enterprise users, this poses additional challenges. To avoid data leaks, not every business will be prepared to authorize employees to use every available genAI model. That challenge implies that Apple will also need to build APIs for Mobile Device Management to enable IT to switch off access to these third-party genAI models for managed devices. The problem with access must therefore logically extend to app-based control. IT will want to be able to prevent people from using apps such as NotebookLM on managed devices, presumably by setting restrictions on the use of certain apps. It also seems viable to expect Apple to offer up an additional choice — one in which users are given the opportunity to select to stay with a purely Apple experience. After all, that should also be an option for those who like it, right? You can follow me on social media! Join me on BlueSky,  LinkedIn, and Mastodon.

The day after his inauguration in January, President Donald Trump announced Stargate, a $500 billion initiative to build out AI infrastructure, backed by some of the biggest companies in tech. Stargate aims to accelerate the construction of massive data centers and electricity networks across the US to ensure it keeps its edge over China. This story is a part of MIT Technology Review’s series “Power Hungry: AI and our energy future,” on the energy demands and carbon costs of the artificial-intelligence revolution. The whatever-it-takes approach to the race for worldwide AI dominance was the talk of Davos, says Raquel Urtasun, founder and CEO of the Canadian robotruck startup Waabi, referring to the World Economic Forum’s annual January meeting in Switzerland, which was held the same week as Trump’s announcement. “I’m pretty worried about where the industry is going,” Urtasun says.  She’s not alone. “Dollars are being invested, GPUs are being burned, water is being evaporated—it’s just absolutely the wrong direction,” says Ali Farhadi, CEO of the Seattle-based nonprofit Allen Institute for AI. But sift through the talk of rocketing costs—and climate impact—and you’ll find reasons to be hopeful. There are innovations underway that could improve the efficiency of the software behind AI models, the computer chips those models run on, and the data centers where those chips hum around the clock. Here’s what you need to know about how energy use, and therefore carbon emissions, could be cut across all three of those domains, plus an added argument for cautious optimism: There are reasons to believe that the underlying business realities will ultimately bend toward more energy-efficient AI. 1/ More efficient models The most obvious place to start is with the models themselves—the way they’re created and the way they’re run. AI models are built by training neural networks on lots and lots of data. Large language models are trained on vast amounts of text, self-driving models are trained on vast amounts of driving data, and so on. But the way such data is collected is often indiscriminate. Large language models are trained on data sets that include text scraped from most of the internet and huge libraries of scanned books. The practice has been to grab everything that’s not nailed down, throw it into the mix, and see what comes out. This approach has certainly worked, but training a model on a massive data set over and over so it can extract relevant patterns by itself is a waste of time and energy. There might be a more efficient way. Children aren’t expected to learn just by reading everything that’s ever been written; they are given a focused curriculum. Urtasun thinks we should do something similar with AI, training models with more curated data tailored to specific tasks. (Waabi trains its robotrucks inside a superrealistic simulation that allows fine-grained control of the virtual data its models are presented with.) It’s not just Waabi. Writer, an AI startup that builds large language models for enterprise customers, claims that its models are cheaper to train and run in part because it trains them using synthetic data. Feeding its models bespoke data sets rather than larger but less curated ones makes the training process quicker (and therefore less expensive). For example, instead of simply downloading Wikipedia, the team at Writer takes individual Wikipedia pages and rewrites their contents in different formats—as a Q&A instead of a block of text, and so on—so that its models can learn more from less. Training is just the start of a model’s life cycle. As models have become bigger, they have become more expensive to run. So-called reasoning models that work through a query step by step before producing a response are especially power-hungry because they compute a series of intermediate subresponses for each response. The price tag of these new capabilities is eye-watering: OpenAI’s o3 reasoning model has been estimated to cost up to $30,000 per task to run.   But this technology is only a few months old and still experimental. Farhadi expects that these costs will soon come down. For example, engineers will figure out how to stop reasoning models from going too far down a dead-end path before they determine it’s not viable. “The first time you do something it’s way more expensive, and then you figure out how to make it smaller and more efficient,” says Farhadi. “It’s a fairly consistent trend in technology.” One way to get performance gains without big jumps in energy consumption is to run inference steps (the computations a model makes to come up with its response) in parallel, he says. Parallel computing underpins much of today’s software, especially large language models (GPUs are parallel by design). Even so, the basic technique could be applied to a wider range of problems. By splitting up a task and running different parts of it at the same time, parallel computing can generate results more quickly. It can also save energy by making more efficient use of available hardware. But it requires clever new algorithms to coordinate the multiple subtasks and pull them together into a single result at the end.  The largest, most powerful models won’t be used all the time, either. There is a lot of talk about small models, versions of large language models that have been distilled into pocket-size packages. In many cases, these more efficient models perform as well as larger ones, especially for specific use cases. As businesses figure out how large language models fit their needs (or not), this trend toward more efficient bespoke models is taking off. You don’t need an all-purpose LLM to manage inventory or to respond to niche customer queries. “There’s going to be a really, really large number of specialized models, not one God-given model that solves everything,” says Farhadi. Christina Shim, chief sustainability officer at IBM, is seeing this trend play out in the way her clients adopt the technology. She works with businesses to make sure they choose the smallest and least power-hungry models possible. “It’s not just the biggest model that will give you a big bang for your buck,” she says. A smaller model that does exactly what you need is a better investment than a larger one that does the same thing: “Let’s not use a sledgehammer to hit a nail.” 2/ More efficient computer chips As the software becomes more streamlined, the hardware it runs on will become more efficient too. There’s a tension at play here: In the short term, chipmakers like Nvidia are racing to develop increasingly powerful chips to meet demand from companies wanting to run increasingly powerful models. But in the long term, this race isn’t sustainable. “The models have gotten so big, even running the inference step now starts to become a big challenge,” says Naveen Verma, cofounder and CEO of the upstart microchip maker EnCharge AI. Companies like Microsoft and OpenAI are losing money running their models inside data centers to meet the demand from millions of people. Smaller models will help. Another option is to move the computing out of the data centers and into people’s own machines. That’s something that Microsoft tried with its Copilot+ PC initiative, in which it marketed a supercharged PC that would let you run an AI model (and cover the energy bills) yourself. It hasn’t taken off, but Verma thinks the push will continue because companies will want to offload as much of the costs of running a model as they can. But getting AI models (even small ones) to run reliably on people’s personal devices will require a step change in the chips that typically power those devices. These chips need to be made even more energy efficient because they need to be able to work with just a battery, says Verma. That’s where EnCharge comes in. Its solution is a new kind of chip that ditches digital computation in favor of something called analog in-memory computing. Instead of representing information with binary 0s and 1s, like the electronics inside conventional, digital computer chips, the electronics inside analog chips can represent information along a range of values in between 0 and 1. In theory, this lets you do more with the same amount of power.  SHIWEN SVEN WANG EnCharge was spun out from Verma’s research lab at Princeton in 2022. “We’ve known for decades that analog compute can be much more efficient—orders of magnitude more efficient—than digital,” says Verma. But analog computers never worked well in practice because they made lots of errors. Verma and his colleagues have discovered a way to do analog computing that’s precise. EnCharge is focusing just on the core computation required by AI today. With support from semiconductor giants like TSMC, the startup is developing hardware that performs high-dimensional matrix multiplication (the basic math behind all deep-learning models) in an analog chip and then passes the result back out to the surrounding digital computer. EnCharge’s hardware is just one of a number of experimental new chip designs on the horizon. IBM and others have been exploring something called neuromorphic computing for years. The idea is to design computers that mimic the brain’s super-efficient processing powers. Another path involves optical chips, which swap out the electrons in a traditional chip for light, again cutting the energy required for computation. None of these designs yet come close to competing with the electronic digital chips made by the likes of Nvidia. But as the demand for efficiency grows, such alternatives will be waiting in the wings.  It is also not just chips that can be made more efficient. A lot of the energy inside computers is spent passing data back and forth. IBM says that it has developed a new kind of optical switch, a device that controls digital traffic, that is 80% more efficient than previous switches.    3/ More efficient cooling in data centers Another huge source of energy demand is the need to manage the waste heat produced by the high-end hardware on which AI models run. Tom Earp, engineering director at the design firm Page, has been building data centers since 2006, including a six-year stint doing so for Meta. Earp looks for efficiencies in everything from the structure of the building to the electrical supply, the cooling systems, and the way data is transferred in and out. For a decade or more, as Moore’s Law tailed off, data-center designs were pretty stable, says Earp. And then everything changed. With the shift to processors like GPUs, and with even newer chip designs on the horizon, it is hard to predict what kind of hardware a new data center will need to house—and thus what energy demands it will have to support—in a few years’ time. But in the short term the safe bet is that chips will continue getting faster and hotter: “What I see is that the people who have to make these choices are planning for a lot of upside in how much power we’re going to need,” says Earp. One thing is clear: The chips that run AI models, such as GPUs, require more power per unit of space than previous types of computer chips. And that has big knock-on implications for the cooling infrastructure inside a data center. “When power goes up, heat goes up,” says Earp. With so many high-powered chips squashed together, air cooling (big fans, in other words) is no longer sufficient. Water has become the go-to coolant because it is better than air at whisking heat away. That’s not great news for local water sources around data centers. But there are ways to make water cooling more efficient. One option is to use water to send the waste heat from a data center to places where it can be used. In Denmark water from data centers has been used to heat homes. In Paris, during the Olympics, it was used to heat swimming pools.   Water can also serve as a type of battery. Energy generated from renewable sources, such as wind turbines or solar panels, can be used to chill water that is stored until it is needed to cool computers later, which reduces the power usage at peak times. But as data centers get hotter, water cooling alone doesn’t cut it, says Tony Atti, CEO of Phononic, a startup that supplies specialist cooling chips. Chipmakers are creating chips that move data around faster and faster. He points to Nvidia, which is about to release a chip that processes 1.6 terabytes a second: “At that data rate, all hell breaks loose and the demand for cooling goes up exponentially,” he says. According to Atti, the chips inside servers suck up around 45% of the power in a data center. But cooling those chips now takes almost as much power, around 40%. “For the first time, thermal management is becoming the gate to the expansion of this AI infrastructure,” he says. Phononic’s cooling chips are small thermoelectric devices that can be placed on or near the hardware that needs cooling. Power an LED chip and it emits photons; power a thermoelectric chip and it emits phonons (which are to vibrational energy—a.k.a. temperature—as photons are to light). In short, phononic chips push heat from one surface to another. Squeezed into tight spaces inside and around servers, such chips can detect minute increases in heat and switch on and off to maintain a stable temperature. When they’re on, they push excess heat into a water pipe to be whisked away. Atti says they can also be used to increase the efficiency of existing cooling systems. The faster you can cool water in a data center, the less of it you need. 4/ Cutting costs goes hand in hand with cutting energy use Despite the explosion in AI’s energy use, there’s reason to be optimistic. Sustainability is often an afterthought or a nice-to-have. But with AI, the best way to reduce overall costs is to cut your energy bill. That’s good news, as it should incentivize companies to increase efficiency. “I think we’ve got an alignment between climate sustainability and cost sustainability,” says Verma. ”I think ultimately that will become the big driver that will push the industry to be more energy efficient.” Shim agrees: “It’s just good business, you know?” Companies will be forced to think hard about how and when they use AI, choosing smaller, bespoke options whenever they can, she says: “Just look at the world right now. Spending on technology, like everything else, is going to be even more critical going forward.” Shim thinks the concerns around AI’s energy use are valid. But she points to the rise of the internet and the personal computer boom 25 years ago. As the technology behind those revolutions improved, the energy costs stayed more or less stable even though the number of users skyrocketed, she says. It’s a general rule Shim thinks will apply this time around as well: When tech matures, it gets more efficient. “I think that’s where we are right now with AI,” she says. AI is fast becoming a commodity, which means that market competition will drive prices down. To stay in the game, companies will be looking to cut energy use for the sake of their bottom line if nothing else.  In the end, capitalism may save us after all. 

The Gemini app for iOS is being updated with extra AI smarts, including a better-performing Gemini 2.5 Flash and a new Gemini Live camera.Google's Gemini Live camera will integrate with Google Calendar in future - Image Credit: GoogleGoogle I/O is the search giant's annual gathering to show off the company's latest innovations. As usual, a lot of it is about artificial intelligence, and some of it applies to iOS.The Gemini app for iOS is the beneficiary of some of the announcements, with updates to the app rolling out to users from Tuesday onwards. Continue Reading on AppleInsider | Discuss on our Forums

In this week's curated employer highlight from Archinect Jobs, we are featuring five architecture and design firms with current openings in New York City, Los Angeles, and Miami. For even more opportunities, visit the Archinect job board and explore our active community of job seekers, firms, and schools. New York City-based design firm Roman and Williams Buildings and Interiors has three exciting job opportunities: a Sr. Designer, Architecture - Residential with a minimum of five to eight years of experience and is fluent in AutoCAD, an Architectural Designer - Residential with a minimum of two to five years of experience and is fluent in AutoCAD and SketchUp, and a Sr. Interior Designer, Residential with a minimum of seven to ten years of experience and is fluent in AutoCAD, Adobe Creative Suite, and Microsoft Office. The Standard NY Hotel by Roman and Williams Buildings and Interiors.STUDIO V Architecture is in search of a Project Manager in New York City with seven-plus years of...