News

Flying petabytes of AI data is China’s latest workaround for strict U.S. chip controlsPhysically smuggling hard drives now bypasses surveillance and digital firewalls across multiple jurisdictionsGPU-rich Malaysian data centers are becoming ground zero for offshore Chinese AI trainingAs the United States continues to tighten export restrictions on advanced AI chips, such as those produced by Nvidia, Chinese AI companies are turning to a workaround that feels almost analog in today’s digital world.Rather than relying on online transfers or sanctioned hardware, some firms are physically transporting vast datasets on hard drives across borders.A report from the Wall Street Journal claims four Chinese tech workers recently flew into Malaysia, each carrying 15 high-capacity hard drives, totaling an astonishing 4.8 petabytes of data, intended for training large language models.Big data continues to enter China despite restrictionsUS restrictions have made it increasingly difficult to acquire high-end AI GPUs through legal channels.Although Nvidia maintains, “there is no evidence of chip diversion,” on-the-ground reports suggest otherwise, with a black market for smuggled Nvidia GPUs thriving in China.Some of these chips are reportedly entering the country through subsidiaries and partners in neighboring nations.However, that route is becoming more expensive and riskier due to heightened scrutiny and diplomatic pressure from Washington on these intermediary countries.As a result, companies are shifting tactics: rather than importing restricted chips, they are exporting massive volumes of training data.This is a complex and resource-intensive process. Companies carefully plan the physical transportation of data, distributing drives to avoid detection by customs.They also rent GPU-rich servers in third-party countries such as Malaysia to process the data.One example involves a Chinese firm that used its Singapore-registered subsidiary to sign a data center contract. However, the Malaysian partner later insisted on local registration to avoid regulatory pressure, as Singapore began tightening its own controls.Despite increasing efforts by US agencies, enforcement gaps and logistical loopholes continue to be exploited. While shipping petabytes of data on hard drives may seem outdated, it sidesteps bandwidth limitations and digital surveillance.The use of hard drives, ranging from large SSDs arrays to high-capacity external HDDs, is central to these covert transfers.Still, it raises a question: why not use magnetic tape, especially given that modern LTO-10 formats can store up to 30TB uncompressed and 75TB compressed?The answer likely lies in practicality. Tape solutions require specialized read/write hardware and lack the plug-and-play convenience of high-end HDDs commonly used today.You might also likeHere are the best mobile workstations around todayWe've rounded up the best external hard drives available nowExperts warn clicking "unsubscribe" on that boring email could actually be a security risk

I'm currently in the middle of testing my first robotic lawn mower, and I'm a shell of the home tech reviewer I used to be. It has been, to put it politely, a humbling journey, and I'm not even at the end of it yet.When I was offered my first lawnbot, I assumed it'd be roughly a similar experience to testing today's best robot vacuums. I've been reviewing those for around a year now, and for the most part, getting them set up has been a doddle: plug in your dock, spend five minutes on a 'quick mapping' run, and you're done. There are things you can do to personalize the setup and improve performance, but they're not essential. You don't need to be a tech whizz to manage it.The demands robot mowers and robot vacuums are similar: both have to navigate logically over an area, covering all the ground and avoiding obstacles, and then find their way back to their charge dock. However, it turns out a lawnbot is not just an outdoor robovac with some added razor blades; the big, crucial difference is in the navigation process. Broadly, the majority of robot vacuums navigate using LiDAR, whereas robot lawn mowers rely on satellites.The model in question is the Mammotion Yuka Mini, but I would have had the same problems with almost any lawnbot (Image credit: Future)The model I'm reviewing is a Mammotion Yuka Mini, and it predominantly uses GPS and an RTK receiver for navigation. While that won't be exactly the same for all models – older bots might still use a boundary wire, and some of the very newest models, like the Eufy E15 robot lawn mower, rely on cameras and don't need an RTK receiver – it's a very common category of lawnbot, so my experience here will be applicable to a range of models from different brands.I haven't quite got everything working as I want it to yet, but I have learned some important lessons thus far. Here are some key things to know before you embark on your own lawnbot journey1. RTK receiver placement is keyMost robot lawn mowers, including my review model, come with an RTK receiver. This is a key part of how lawnbots navigate without boundary wires – it essentially helps adjust satellite positioning data so it's correct to within a few centimeters rather than a few meters. After all, that kind of margin could be the difference between a lawn and the middle of the driveway or flowerbed.An RTK receiver is shaped like the top of a large mushroom, with an antenna protruding from its base. The Yuka's instructions told me this could be mounted on a pole and stuck into the lawn, or attached to the side of my house. Option one seemed by far the easiest – not least because I was testing at my parents' home, and my dad keeps an especially close guard of his power tools and wouldn't take kindly to new holes appearing in the side of his house.The final location for the lawnbot's RTK receiver (Image credit: Future)Unfortunately, option two is the superior choice if you want your lawnbot to work effectively. The RTK receiver relies on a direct line of sight to a range of satellites in the sky. That means it needs to be it in a lofty, fixed position with a wide view of the sky in a number of different directions, without things like trees and structures looming above it.While attaching the RTK receiver to your home (or, in my case, the wooden side of a freestanding garage) is a bit of a job, putting it in the lawn is trickier than it initially seems, too. The receiver requires a power source, which limits where it can go, and you also need to be sure the bot isn't going to mow over its own power cable when it's in place.2. Your charge dock needs a good, fixed positionSeparate from the RTK receiver, the lawnbot has a charge dock, which is where it returns to charge its batteries. Finding a good, fixed spot for the charge dock to live is surprisingly difficult.The ground needs to be flat and hard. It needs to be in a position from which the lawnbot can access the bits of grass that require cutting. It needs to be relatively near a power outlet, and positioned so that it won't mow its own charge cable. It may or may not need to be within reach of the house Wi-Fi (I'm still investigating this part). And because the bot needs to have a direct line of sight to at least some satellites while it's docked, it can't go under a shelter or tucked away between structures. Long story short, I have yet to find anywhere that works in my test yard.One of my regular reviewers had less trouble finding a home for his bots' charge docks (Image credit: Future)Initially, I tried the charge station on a concreted area by the back door, but in this spot, the robot kept telling me the dock had been moved and it couldn't find its way back to it. I think that's because on concrete it's impossible (or, at least difficult) to attach the charge station to the ground, and the dock was shifting about a bit when the lawnbot was being remote controlled / placed in and out.3. Yards are full of unexpected obstacles Since beginning my setup process, I've been struck by how many unexpected obstacles have made themselves known. The lawnbot has become tangled in the netting over fruit plants. It digs itself into a trench whenever it's asked to make its way over a pathway of – admittedly quite deep – pebbles. The less precise mapping and positioning has knock-on effects too: there's a grassy pathway that's perhaps two lawnbots wide that my review model refuses to mow because it's too narrow. I've also had to leave a chunky border along the edge of a shallow brook that runs along the side of the yard, for fear the lawnbot might try to take a dip.None of these things are particularly surprising, and I don't think they represent failings on the part of the lawnbot either. But I would urge you to take a moment to consider the layout of your yard, and what might pose a challenge, before purchasing one of these.The results of the Mammotion Yuka Mini's efforts in one area of my test yard (Image credit: Future)Is it worth it? At time of writing, I've got to the point where the lawnbot will happily mow two unusually shaped lawn areas in my test yard, but it won't return to its dock without help. Even getting this far has been a challenge, and I'd be lying if I said I hadn't considered chucking it into the neighbor's pond and being done with it. But at this point I'm actually feeling quite positive about the technology.You see, the Mammotion just does such a good job of mowing the lawn. The results are neat and it doesn't miss any spots, and it will even amble up and down relatively steep inclines without complaining. It's so much less effort than doing it manually. I don't even mind having to remote-control it over the narrow strips that it refuses to map, because it's still something I can do from my phone, while sitting on a deck lounger.If I can find somewhere for the dock to live, so I don't need to remote control or carry the bot to and from its base, I'll be very happy with it indeed. And maybe at that point, I will forget the pain of getting it set up.You might also like...Read our full Mammotion LUBA 2 review for a closer look at a sister botHere's more on how robot lawn mowers workShould I buy a robot lawn mower? Here are the pros and cons

Google is testing the NotebookLM feature Audio Overviews in SearchThe feature will offer short, AI-generated audio summaries for certain queries.The feature uses Gemini models to deliver podcast-style explanations with clickable links.I've been a fan of the Audio Overviews feature in Google's NotebookLM since I first experimented with it last year. Now, it's coming to Google Search, currently only as a test in the Labs, but it brings a more bite-sized version of the AI-generated "podcasts" that I like in NotebookLM.Once you’ve opted in through Labs, you’ll start seeing a little prompt on some search results pages saying, “Generate Audio Overview.” Tap that, wait about 30 to 40 seconds, and out comes a compact audio clip of around five minutes, sometimes less, that explains what you looked up in the form of two AI-generated voices having a discussion. Not too deep, but not one-sentence shallow either. Think of a middle ground between “Wikipedia rabbit hole” and “I read the headline only."While you listen, the audio player stays docked in your results page, showing clickable links to the sources the AI pulled from. You can keep browsing, tap into related articles, or just listen and absorb. If you like what you hear, you can give it a thumbs up. If it's egregiously wrong, the thumbs down is there too.Though similar to what NotebookLM does with its Audio Overviews, the Search version has one major difference. NotebookLM only uses documents you upload, YouTube videos, and websites you specifically link to. Google Search’s version pulls from public web content. That can be good or bad, depending on what you look up. Something straightforward and scientific might be fine, but a discussion about the best movie ever might get a different audio track every time you look. Here's an example I recorded a clip from.AI podcast searchesIt's hardly perfect, and while the voices are good, they are still AI voices. You also might notice it parroting phrases straight from someone’s Reddit post. But it is listenable and, as Google points out, hands-free, with the option to adjust the speed of the speakers and the links there to provide more context. You can speed it up or slow it down, skip around, or follow the links as you go. It's AI-enhanced Search, not a new audiobook.For now, not every search will offer to create an Audio Overview. You also have to be in the U.S. and sign up for Labs right now. But, I'd expect it to have a general release pretty soon. Then you can ask how lithium-ion batteries work or why Roman concrete is still standing, and get a nice mini discussion from digital characters.Think of it as how video summaries and image carousels brought new dimensions to how we take in information online. Audio Overviews are another aspect of that and a win for auditory learners or people with visual impairments, With OpenAI and Perplexity and a dozen AI search engines nipping at its heels, Google needs whatever tricks it can muster to stand out and an AI podcast as the answer to a serach is definitely one way to be unique, at least for now.You might also likeOne of my favorite AI tools is getting an iPhone app, and here's why you should install itI used NoteBookLM to help with productivity - here’s 5 top tips to get the most from Google’s AI audio toolGoogle will turn those long documents into your next favorite podcastI fed NotebookLM a 218-page research paper on string theory and the podcast results were mind-blowing

Amazon to invest AU$20 billion in Australia between now and 2029New data centers in Sydney and Melbourne are part of the planThree new solar farms are also crucial to the investmentAmazon has confirmed plans to invest AU$20 billion into Australia between 2025 and 2029, with expanded data center infrastructure in Sydney and Melbourne accounting for a large part of the upcoming plans.The company says this multibillion-dollar pledge is Australia's "largest publicly-announced global technology investment."In its announcement, Amazon declared its support for Australia's "vision to improve productivity and grow the economy through AI innovation."Amazon makes "largest" tech investment into AustraliaCiting figures from the Australian Government’s Department of Industry, Science and Resources, Amazon noted that AI and automation could contribute up to AU$600 billion annually to Australia’s GDP by 2030.As part of the cash injection and to support its expanding infrastructure network, Amazon also confirmed it will build three new solar farms in Victoria and Queensland to add more than 170 megawatts of clean energy to the grid. That's on top of the eight solar and wind projects it already invests in across New South Wales and the two above states.With all 11 sites in operation, Amazon expects to generate more than 1.4 million megawatt hours of clean energy annually."This is the largest investment our country has seen from a global technology provider, and is an exciting opportunity for Australia to build AI capability using secure, resilient infrastructure," said Australian Prime Minister Anthony Albanese.The announcement comes with the usual side dishes – bold claims of widespread upskilling and economic contributions.Claiming to have trained over 400,000 Australian citizens since 2017, the company pledged to work with educational institutions, industry partners and the government for further training opportunities.AWS CEO Matt Garman concluded: "AI is a once-in-a-generation transformation, and Amazon is pleased to be empowering all Australians to innovate at scale through this investment."The news comes just days after AWS committed $20 billion to expand cloud computing infrastructure in Pennsylvania, and a further $10 billion commitment for North Carolina.You might also likeCheck out the best cloud hosting providersAWS reveals European Sovereign Cloud to allay fears about US providersKeep your data online with the best cloud storage and best cloud backup tools

Meta has just announced it's partnering with Oakley on something newMost likely, this is the long-rumored Oakley smart glassesWe'll find out more on June 20Oakley and Meta are officially partnering on something – I mean, it’s smart glasses, right? – and we’ll be finding out everything about it in just a few days on June 20, 2025.Meta’s CTO, Andrew Bosworth, took to social media to share a bombastic promo video featuring the Meta and Oakley logos.🤘🏼 pic.twitter.com/dvJ0y6afDQJune 16, 2025This announcement has been a long time coming. We first heard that Meta has been developing Oakley versions of its Ray-Ban Meta smart glasses back in January, and it makes a lot of sense.The Ray-Ban smart glasses are among Meta’s most popular gadgets, and as another EssilorLuxotica brand, Oakley, is in a prime position to try and get itself in on the smart glasses action too.What is strange, however, is that Meta isn’t making this reveal at Meta Connect in September, as it normally would. Instead, according to Oakley’s website, “The next evolution is coming on June 20.”As for what that evolution might look like, according to the leaks, the Oakley specs will be based on its Sphaera glasses. The specs of the glasses would seemingly be similar to the current model of Ray-Bans (ie, camera, no display), but would shift the camera to the center of the glasses frame, and include features to appeal to cyclists and athletes.(Image credit: Oakley)Now, technically, Meta hasn’t specifically said we’re getting Oakley smart glasses, but given the previous leaks and that Oakley is a glasses brand, it is all but certain that’s indeed what we’re getting.Bringing the fight to Android XRThis may have always been the plan, but given how unusual the timing is, I can’t help but feel this is a reaction to Google’s recent Android XR debut.Five brands are working with Google to make Android XR glasses, including Samsung, Xreal, and Gentle Monster. Yet, until now, Meta has only had one partner.Although Meta is hardly on the back foot, its key advantage is that its glasses are not only already available but also dominate the smart glasses space. And by launching its Oakley specs sooner rather than later, Meta’s not giving Android XR time to breathe.We’re not expecting the first Android XR specs to launch until 2026, and that could cause trouble for the new XR operating system if the space is already saturated with great Meta specs.(Image credit: Google)Another advantage of launching now is that it gives Meta’s other rumored smart glasses project space to own Meta Connect.That’s, of course, its leaked smart glasses which will feature a screen. According to rumors, you'll be able to see apps and notifications with a system that sounds identical to everything Google announced for its AR glasses.Again, while Google was first to showcase next-gen consumer AR specs, Meta is seemingly set to be the first mainstream company to actually launch a pair, with that announcement still presumably set for Meta Connect 2025.We’ll have to wait and see what Meta announces on June 20 and in September, but its Ray-Ban smart glasses continue to be one of my favorite gadgets of the past couple of years, and I’m ecstatic that it's finally getting some hardware updates.You might also likeHands on: I tried Google's Android XR prototype and they can't do much but Meta should still be terrified‘I made a lot of mistakes with Google Glass’: Google’s Sergey Brin admits missteps but says Android XR has a bright future for one big reasonGoogle's Veo 3 marks the end of AI video's 'silent era'