When arsenic is mentioned, many people associate it with a bygone poison. The type of quiet killer used in a murder mystery set on a train somewhere in Victorian England. It seems like a problem of the past.However, arsenic is a naturally occurring substance that can contaminate groundwater and food irrigated with tainted water. And in an alarming new study, researchers have found that climate change is impacting the level of arsenic that people are ingesting. Given that arsenic is a carcinogen, it’s primed to become a lethal killer. Billions of people are at risk of developing arsenic-induced diseases.Why Arsenic Levels are Rising in Rice In a 2025 study in The Lancet, an international team of researchers revealed the results of a study that measured two components associated with climate change — increased carbon dioxide (CO2) levels and surface temperatures — in rice paddies in China.“One important chemical aspect that has been known about rice is its ability to accumulate arsenic, leading to an important and crucial question: Can CO2 and temperature affect arsenic concentration, and if so, what are the health consequences?” says Lewis H. Ziska, an associate professor in environmental health sciences at the Mailman School of Public Health at Columbia University. The study was the first of its kind. Although scientists have long understood how both inorganic and organic arsenic can make their way into food or water, they weren’t sure how climate change might influence arsenic levels.To find out, the researchers grew rice in controlled paddies in China from 2014 to 2023. To a certain extent, the research team expected to find some arsenic in the rice because it grows in flooded paddies and absorbs arsenic from the water.The research team also expected the CO2 levels and surface temperatures would increase because climate change scientists have charted a projection of the likely increase. But would increased CO2 levels and temperatures lead to higher arsenic levels? And what might that mean if people eat the tainted crop? Impacts of Eating ArsenicThe study found that the increased CO2 levels and temperatures indeed lead to higher concentrations of arsenic. The researchers’ rice paddies were representative of rice paddies in Asia, which means that what happened in their paddies is likely happening in others around the region.Higher arsenic rates in rice are a major health concern. Arsenic is a known carcinogen. When found in drinking water, there is an increased risk of bladder and skin cancer. Other studies have found that ingesting arsenic can lead to cancer in the digestive tract, hematopoietic system, kidneys, liver, or lymphatic system.Problematically, billions of people can now be considered at risk for arsenic-related cancers. More than half of the world’s population depends on rice as a staple food. Scientists are concerned that the increased arsenic in a staple food like rice could lead to a spike in future cancer cases throughout Asia.Rice Consumption in the U.S. But what does this mean for the U.S. food supply? Rice consumed in the U.S. comes from the Arkansas Grand Prairie, Gulf Coast, Mississippi Delta, or Sacramento Valley. These crops have not been tested by scientists to determine how CO2 levels and surface temperatures are impacting the arsenic concentrations. But if climate change can raise arsenic levels in Asia, Ziska says it can happen in the U.S. as well.“It seems likely given the biogeochemistry of how rice can accumulate arsenic when grown under flooded (paddy) conditions,” Ziska says.It’s a research area that Ziska says needs more attention, both in terms of long-term study support and public awareness. Many people don’t realize that climate change has the ability to alter the food supply by creating conditions that are ripe for toxins.“Climate change isn’t just about wildfires and sea-level rise and storms of the century. It’s also affecting a food staple — rice — that is consumed by half the world’s population every day. Maybe we should pay attention,” he says. This article is not offering medical advice and should be used for informational purposes only.Article SourcesOur writers at Discovermagazine.com use peer-reviewed studies and high-quality sources for our articles, and our editors review for scientific accuracy and editorial standards. Review the sources used below for this article:National Cancer Institue. ArsenicEmilie Lucchesi has written for some of the country's largest newspapers, including The New York Times, Chicago Tribune and Los Angeles Times. She holds a bachelor's degree in journalism from the University of Missouri and an MA from DePaul University. She also holds a Ph.D. in communication from the University of Illinois-Chicago with an emphasis on media framing, message construction and stigma communication. Emilie has authored three nonfiction books. Her third, A Light in the Dark: Surviving More Than Ted Bundy, releases October 3, 2023, from Chicago Review Press and is co-authored with survivor Kathy Kleiner Rubin.

david_franklin - stock.adobe.com News Post Office scandal inquiry to publish first findings this summer Inquiry announces that findings on compensation and human impact will be presented in the next few months By Karl Flinders, Chief reporter and senior editor EMEA Published: 15 May 2025 10:20 The Post Office scandal public inquiry will publish the first part of its final report this summer, with its findings on compensation and the scandal’s human impact to be made public “as soon as possible”. The inquiry, which lasted about three years, examined a scandal that went on for two decades and saw subpostmasters wrongly blamed and punished for accounting shortfalls. It is described as the widest miscarriage of justice in UK history, and over 700 people have had wrongful convictions overturned as a result of it being fully exposed. Computer Weekly first revealed the Horizon problems in 2009, and has been investigating ever since, including detailed reporting of the inquiry since it began hearing evidence in May 2022. Inquiry chair Wyn Williams, who today made the announcement, said: “Throughout my investigations, I have heard how people have faced unimaginable hardship due to the impacts of the Horizon scandal, from wrongful convictions and financial ruin to broken relationships and severe emotional distress. I have also held several compensation hearings, which have stressed the urgency for people to receive redress that is full, fair and prompt.” He added that the people affected by the scandal should be at the heart of the inquiry’s work. “It seems fitting that my conclusions on the suffering endured by so many, as well as the issue of redress, should be at the forefront of my report and published as soon as feasibly possible,” said Williams. After a High Court victory in 2019, when subpostmasters proved the Horizon computer system they used was to blame for unexplained losses, the first thing campaigner Alan Bates, now Sir Alan, said to Computer Weekly was that he wanted a statutory public inquiry into the scandal. He got it in May 2021, when a government inquiry into the scandal was made statutory. Read a round-up of the full inquiry Post Office Horizon IT scandal inquiry: Three years of shocking revelations. When originally set up as a government department inquiry without the power to call witnesses, there was outrage and accusations of whitewashing. But following pressure from subpostmasters and their supporters, the inquiry was put on a statutory footing, with a judge in charge and the power to call witnesses to give evidence. The inquiry was split into seven phases.  The human impact hearings were shocking, revealing the extreme suffering of people at the hands of the Post Office. Other phases have revealed that the Post Office had knowledge that the Horizon software had bugs when rolled out, prosecution witnesses changed their statements when prompted by the Post Office, and lawyers hid evidence during trials of subpostmasters because it would have made their prosecutions unsafe. It has also featured directors, politicians and civil servants who, whether deliberately or not, contributed to the cover-up. Following Williams’ announcement, Sir Alan Bates said bringing forward the findings involving financial redress was vital. “It is good that things will start to appear, albeit in sections,” he said. “The chair is right in his comments because it is the priority in all this,” said Bates. “I just hope it helps bring swift resolution to all the outstanding problems. We have to find, one way or another, a way of getting [financial redress] done.” The Post Office scandal was first exposed by Computer Weekly in 2009, revealing the stories of seven subpostmasters and the problems they suffered due to Horizon accounting software, which led to the most widespread miscarriage of justice in British history. Read detailed coverage of the entire public inquiry by Computer Weekly Read more about phase one of the public inquiry: The British people are waking up to the scandal that happened under their noses. Read a round-up of phase two: Post Office scandal – ‘cock-up or cook-up’? Read a round-up of phase three: Post Office scandal – cover-up a ‘dark chapter’ in government, corporate and legal history. Read a round-up of phase four: Post Office scandal: Phase four’s rogues’ gallery. Read a round-up of the combined phases five and six: Post Office scandal: Phases 5 and 6 had islands of conscientiousness in great depths of neglect. Read a round up of phase seven: Post Office scandal: Inquiry’s final phase exposes dysfunction past and present. In The Current Issue: UK MoJ crime prediction algorithms raise serious concerns Interview: Markus Schümmelfeder, CIO, Boehringer Ingelheim Download Current Issue Mind the insight-to-impact gap, Qlik captures analytics ‘in the moment’ – CW Developer Network Safer Sharknado: Sysdig donates Stratoshark to Wireshark Foundation – Open Source Insider View All Blogs

May 15, 2025Ravie LakshmananAI Training / Data Protection Austrian privacy non-profit noyb (none of your business) has sent Meta's Irish headquarters a cease-and-desist letter, threatening the company with a class action lawsuit if it proceeds with its plans to train users' data for training its artificial intelligence (AI) models without an explicit opt-in. The move comes weeks after the social media behemoth announced its plans to train its AI models using public data shared by adults across Facebook and Instagram in the European Union (E.U.) starting May 27, 2025, after it paused the efforts in June 2024 following concerns raised by Irish data protection authorities. "Instead of asking consumers for opt-in consent, Meta relies on an alleged 'legitimate interest' to just suck up all user data," noyb said in a statement. "Meta may face massive legal risks – just because it relies on an 'opt-out' instead of an 'opt-in' system for AI training." The advocacy group further noted that Meta AI is not compliant with the General Data Protection Regulation (GDPR) in the region, and that, besides claiming that it has a "legitimate interest" in taking user data for AI training, the company is also limiting the right to opt-out before the training has started. Noyb also pointed out that even if 10% of Meta's users expressly agree to hand over the data for this purpose, it would amount to enough data points for the company to learn E.U. languages. It's worth pointing out that Meta previously claimed that it needed to collect this information to capture the diverse languages, geography, and cultural references of the region. "Meta starts a huge fight just to have an opt-out system instead of an opt-in system," noyb's Max Schrems said. "Instead, they rely on an alleged 'legitimate interest' to just take the data and run with it. This is neither legal nor necessary." "Meta's absurd claims that stealing everyone's personal data is necessary for AI training is laughable. Other AI providers do not use social network data – and generate even better models than Meta." The privacy group also accused the company of moving ahead with its plans by putting the onus on users and pointed out that national data protection authorities have largely stayed silent on the legality of AI training without consent. "It therefore seems that Meta simply moved ahead anyways – taking another huge legal risk in the E.U. and trampling over users' rights," noyb added. In a statement shared with Reuters, Meta has rejected noyb's arguments, stating they are wrong on the facts and the law, and that it has provided E.U. users with a "clear" option to object to their data being processed for AI training. This is not the first time Meta's reliance on GDPR's "legitimate interest" to collect data without explicit opt-in consent has come under scrutiny. In August 2023, the company agreed to change the legal basis from "legitimate interest" to a consent-based approach to process user data for serving targeted ads for people in the region. The disclosure comes as the Belgian Court of Appeal ruled the Transparency and Consent Framework, used by Google, Microsoft, Amazon, and other companies to obtain consent for data processing for personalized advertising purposes, is illegal across Europe, citing violation of several principles of GDPR laws. Found this article interesting? Follow us on Twitter  and LinkedIn to read more exclusive content we post. SHARE    

Citrix proposes using VDI and Linux to dodge rising PC prices, but the fix carries technical risks, licensing costs, and security vulnerabilities that companies must consider.

Google DeepMind’s AI systems have taken big scientific strides in recent years — from predicting the 3D structures of almost every known protein in the universe to forecasting weather more accurately than ever before.  The UK-based lab today unveiled its latest advancement: AlphaEvolve, an AI coding agent that makes large language models (LLMs) like Gemini better at solving complex computing and mathematical problems.  AlphaEvolve is powered by the same models that it’s trying to improve. Using Gemini, the agent proposes programs — written in code — that try to solve a given problem. It runs each code snippet through automated tests that evaluate how accurate, efficient, or novel it is. AlphaEvolve keeps the top-performing code snippets and uses them as the basis for the next round of generation. Over many cycles, this process “evolves” better and better solutions. In essence, it is a self-evolving AI.   DeepMind has already used AlphaEvolve to tackle data centre energy use, design better chips, and speed up AI training. Here are five of its top feats so far.  1. It discovered new solutions to some of the world’s toughest maths problems Register now AlphaEvolve was put to the test on over 50 open problems in maths, from combinatorics to number theory. In 20% of cases, it improved on the best-known solutions to them.  One of those was the 300-year-old kissing number problem. In 11-dimensional space, AlphaEvolve discovered a new lower bound with a configuration of 593 spheres — progress that even expert mathematicians hadn’t reached.  2. It made Google’s data centres more efficient The AI agent devised a way to better manage power scheduling at Google’s data centres. That has allowed the tech giant to improve its data centre energy efficiency by 0.7% over the last year — a significant cost and energy saver given the size of its data centre operation.  3. It helped train Gemini faster AlphaEvolve improved the way matrix multiplications are split into subproblems, a core operation in training AI models like Gemini. That optimisation sped up the process by 23%, reducing Gemini’s total training time by 1%. In the world of generative AI, every percentage point can translate into cost and energy savings.  4. It co-designed part of Google’s next AI chip The agent is also using its code-writing skills to rewire things in the physical world. It rewrote a portion of an arithmetic circuit in Verilog — a language used for chip design — making it more efficient. That same logic is now being used to develop Google’s future TPU (Tensor Processing Unit), an advanced chip for machine learning.   5. It beat a legendary algorithm from 1969 For decades, Strassen’s algorithm was the gold standard for multiplying 4×4 complex matrices. AlphaEvolve found a more efficient solution — using fewer scalar multiplications. This could lead to more advanced LLMs, which rely heavily on matrix multiplication to function. According to DeepMind, these feats are just the tip of the iceberg for AlphaEvolve. The lab envisions the agent solving countless problems, from discovering new materials and drugs to streamlining business operations. AI’s evolution will be a hot topic at TNW Conference, which takes place on June 19-20 in Amsterdam. Tickets for the event are now on sale — use the code TNWXMEDIA2025 at the checkout to get 30% off. Story by Siôn Geschwindt Siôn is a freelance science and technology reporter, specialising in climate and energy. From nuclear fusion breakthroughs to electric vehic (show all) Siôn is a freelance science and technology reporter, specialising in climate and energy. From nuclear fusion breakthroughs to electric vehicles, he's happiest sourcing a scoop, investigating the impact of emerging technologies, and even putting them to the test. He has five years of journalism experience and holds a dual degree in media and environmental science from the University of Cape Town, South Africa. When he's not writing, you can probably find Siôn out hiking, surfing, playing the drums or catering to his moderate caffeine addiction. You can contact him at: sion.geschwindt [at] protonmail [dot] com Get the TNW newsletter Get the most important tech news in your inbox each week. Also tagged with

Quantum computing may still be regarded by many IT leaders as a very niche technology, but broader business use cases may be just a few years away. While only a handful of companies have machines with logical qubits today, delegates at the Commercialising Quantum Computing conference in London were told that a machine with 100 logical qubits would offer quantum advantage in material science by 2028. This means that, by then, a sufficiently powerful and stable quantum computer would start delivering business value better than what would be possible using high performance computing. Mark Jackson, senior quantum evangelist at Quantinuum, said the company was already using generative quantum artificial intelligence (AI). In a fireside chat at the conference, Jackson spoke about the interaction between quantum computing and AI. It is largely acknowledged that a quantum computer is not good at providing a precise answer, such as if applied to big data analysis. But, according to Jackson, it shines when used for machine learning, which can be applied to identify a correct answer. Quantum-enhanced machine learning can process large datasets far quicker than conventional computers, especially when applied to detecting patterns. “Quantum computers can detect patterns that would be missed by other conventional computing methods,” said Jackson. This ability to detect patterns in massive datasets could revolutionise cyber security. Becky Pickard, managing director of global cyber operations at Barclays, pointed out during a panel discussion that a lot of progress has been made with machine learning and how to apply it on a day-to-day basis: “We’re working with massive volumes of data – 12Tbytes – on a daily basis.” She suggested that quantum machine learning could help. From an optimisation perspective, she is keen to see the development of quantum computing applied in a way that reshapes cyber defence.  HSBC is one of the organisations that has been working on quantum computing for several years. Discussing the return on investment opportunity, and how quantum computing can be used to build more optimised financial models, Phil Intallura, global head of quantum technologies at HSBC, said: “When you breakdown the opportunities, financial services is one of the biggest beneficiaries.” As Intallura points out, banks are always looking for a better financial model: “There’s one thing that catalyses commercial organisations more than anything else, and that’s confidence. If you can show a solution using quantum technology that can get a better output based than using a supercomputers, [business decision-makers] will give you much more runway than you need.” Another application area is the ability to generate a true random number, which can feed into financial model simulations. In March, a team of researchers from JPMorganChase, Quantinuum, Argonne National Laboratory, Oak Ridge National Laboratory, and the University of Texas at Austin published a paper in Nature discussing a technique known as Random Circuit Sampling (RCS). RCS is used to perform a certified-randomness-expansion protocol, which outputs more randomness than it takes as input. It is a task that is often used to demonstrate quantum supremacy since it cannot be achieved on a classical computer. Speaking of the usefulness of a quantum number generator at HSBC, Intallura said: “Using quantum random numbers as your entropy source to classical simulation does not change any of the underlying model practices in classical models. You’re just injecting a different source of entropy than what we would [normally] use.” For Intallura, regulatory pressure and the need to ensure financial transactions are secure is helping to inform quantum computing plans at financial institutions.  The US National Institute of Standards and Technology has ratified a number of post-quantum cryptography (PQC) standards. Banks face pressure from regulators to replace RSA-2048 encryption by 2035 and migrate fully over to quantum safe encryption standards to protect banking transactions.But, as Mark Carney, lead of quantum cyber security research at Santander Global, noted, post-quantum cryptography needs both software and hardware acceleration. “We want to be able to have PQC at speed in our devices and on our payment cards,” he said. “We want to give our customers the very best cryptography that we possibly can – not just for regulatory purposes, but also because it gives a sense of assurance.” Among the promises of quantum computing is that it can be applied to solve complex optimisation problems. As and when they become commercially viable, such systems will need to work alongside traditional enterprise IT. This is something that Gerard Mullery, interim CEO of Oxford Quantum Circuits, recognised during his presentation at the event. Mullery sees a need for quantum computing to be embedded in enterprise workflows.  “As AI agents autonomously orchestrate enterprise workflows, quantum compute platforms must be designed to integrate with them,” he added. What is clear from the experts who spoke at the Commercialising Quantum Computing conference is that a useful machine is perhaps only a few years away. This will have enough logical qubits to solve real-world problems. As such devices evolve, it is likely more organisations will draw on quantum computing for certain combinatorial optimisation problems, which will need to integrate with classical computing in the datacentre. As quantum computing becomes more accessible, there will also be a need to bolster cryptography with PQC. Read more about quantum developments Cisco lays out plans for networking in era of quantum computing: The network equipment provider has opened a new lab and developed a prototype chip as it fleshes out its quantum networking strategy. Quantum datacentre deployments: How they are supporting evolving compute projects: Quantum datacentre deployments are emerging worldwide, so what are they and where are the benefits?

Join Fox News for access to this content You have reached your maximum number of articles. Log in or create an account FREE of charge to continue reading. By entering your email and pushing continue, you are agreeing to Fox News' Terms of Use and Privacy Policy, which includes our Notice of Financial Incentive. Please enter a valid email address. Hackers are pretty scary. Amoral, hooded figures with magical computer skills that can break into anything within minutes. At least that is what most of us think of when we hear the term "hacker." It is not exactly a realistic or particularly representative image, but, at the same time, it is not too far off the mark either. What many people do not realize is just how much hackers can learn about you from a data broker file, detailed profiles compiled from your personal information collected and sold by data brokers. This hidden industry fuels much of what hackers use to target individuals today. Illustration of a person’s personal data. (Kurt "CyberGuy" Knutsson)What is a data broker, and why should you care?When most people hear the term "data broker," on the other hand, they draw a blank. Is it a person or company that buys and sells data? Basically, yes. Data brokers collect, sort, analyze, package and sell access to personal information. Whose personal information? Anyone's and everyone's, including yours.THINK YOU CAN DELETE YOUR DATA YOURSELF? HERE’S WHY YOU’RE PROBABLY WRONG Illustration of a person’s personal data. (Kurt "CyberGuy" Knutsson)What kinds of personal information do they deal in?Here is a list of possible data points you, random companies, your worst enemy, your neighbors and, yes, hackers can find in a data broker file:Full namePast legal namesKnown aliasesSocial media handlesGenderSexual preferencesDate of birth or ageSocial Security number (SSN)Current and past home addressesCurrent and past phone numbersCurrent, past and secret email addressesPolitical preferences and affiliationsOccupationCurrent employerEmployment historyBusiness associatesEducationMarital statusMarriage and divorce recordsFamily status (number and ages of children)RelativesProperty informationVehicle registrationAssetsFinancial informationBankruptcies, judgments and liensLicenses (drivers license, firearms permits, etc.)Court recordsCriminal recordsMugshotsSex-offender statusHealth historyLocation dataIP informationDevice information (phone, tablet and computer models)Web browser informationShopping habitsInterests and hobbies.Quite the list, is it not? It is not exhaustive; there is more that data brokers collect and more yet that they and their customers can infer from data points like these. What exactly a given data broker has will depend on which category it falls into.HACKERS CLAIM MASSIVE BREACH OF COMPANY THAT TRACKS AND SELLS AMERICANS' LOCATION DATA Illustration of a person’s personal data. (Kurt "CyberGuy" Knutsson)Types of data brokers: Who's collecting and selling your information?The most visible data brokers are commonly known as people search sites or people finder sites. These are the sites that show up when you Google yourself, or a hacker Googles you. They are just the tip of the iceberg, though. There are other data brokers that do not bother indexing individual profiles with search engines, preferring instead to deal with other companies and even governments directly. These are the other major types of data brokers, in addition to people search sites:Marketing data brokers focus more on your browsing habits, past purchases and interests. They are responsible for "personalized marketing" as well as helping other companies target you with those surprisingly relevant ads you see online.Recruitment data brokers collect and process personal information to offer background screening services to organizations evaluating job candidates or performing background checks before making an offer. Unfortunately, although illegal, there is nothing actually stopping unscrupulous employers from using the much less regulated and reliable people search sites for the same purposes.Risk mitigation brokers aggregate a variety of background, criminal, property and other information to provide assessment reports to various investment and business companies. The information they collect is aimed at helping such companies manage risk in taking on new business.Financial information brokers collect various personal finance and background information for credit companies and banks to calculate your credit score and may influence your eligibility to get loans and lines of credit.Health information data brokers collect information about your general health and sell it to companies in healthcare and related fields. This information can be used to target you with health product ads and even set your insurance premiums.Hackers are most likely to use people search sites, though; they are easily accessible, eminently searchable, relatively cheap (with trial offers for as little as a dollar) and do not ask any questions. Other data brokers may also sell personal information to hackers, but they tend to sell profiles in bulk.Want your data taken off this market?It is perfectly normal to want no part in any of this. Unfortunately, there is no federal law on the books in the U.S. that would either prevent this kind of data trade or give you an easy, legally enforced way to opt out.The good news is that, thanks to an incomplete patchwork of state laws, personal information removal services can approach hundreds of individual data brokers on your behalf with legally binding data-removal requests.While no service can guarantee the complete removal of your data from the internet, a data removal service is really a smart choice. They aren’t cheap — and neither is your privacy. These services do all the work for you by actively monitoring and systematically erasing your personal information from hundreds of websites. It’s what gives me peace of mind and has proven to be the most effective way to erase your personal data from the internet. By limiting the information available, you reduce the risk of scammers cross-referencing data from breaches with information they might find on the dark web, making it harder for them to target you. Check out my top picks for data removal services here. Kurt's key takeawaysIt is easy to feel overwhelmed when you realize just how much of your personal information is out there, and how many different players are collecting, selling and using it. But knowing how data brokers operate is the first step to taking back control. Whether it is people search sites or the less visible brokers working behind the scenes, your data is valuable, and you deserve to know who has it and what they are doing with it. The good news is, there are tools and services out there that can help you clean up your digital footprint and protect your privacy. So, do not just sit back and hope for the best — take action and make your data work for you, not against you.CLICK HERE TO GET THE FOX NEWS APPIn your opinion, what should be done to give people more control over their data? Let us know by writing us at Cyberguy.com/ContactFor more of my tech tips and security alerts, subscribe to my free CyberGuy Report Newsletter by heading to Cyberguy.com/NewsletterAsk Kurt a question or let us know what stories you'd like us to cover.Follow Kurt on his social channels:Answers to the most-asked CyberGuy questions:New from Kurt:Copyright 2025 CyberGuy.com. All rights reserved.  

May 15, 20253 min readNew Google AI Chatbot Tackles Complex Math and ScienceA Google DeepMind system improves chip designs and addresses unsolved math problems but has not been rolled out to researchers outside the companyBy Elizabeth Gibney & Nature magazine DeepMind says that AlphaEvolve has helped to improve the design of AI chips. MF3d/Getty ImagesGoogle DeepMind has used chatbot models to come up with solutions to major problems in mathematics and computer science.The system, called AlphaEvolve, combines the creativity of a large language model (LLM) with algorithms that can scrutinize the model’s suggestions to filter and improve solutions. It was described in a white paper released by the company on 14 May.“The paper is quite spectacular,” says Mario Krenn, who leads the Artificial Scientist Lab at the Max Planck Institute for the Science of Light in Erlangen, Germany. “I think AlphaEvolve is the first successful demonstration of new discoveries based on general-purpose LLMs.”On supporting science journalismIf you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.As well as using the system to discover solutions to open maths problems, DeepMind has already applied the artificial intelligence (AI) technique to its own practical challenges, says Pushmeet Kohli, head of science at the firm in London.AlphaEvolve has helped to improve the design of the company’s next generation of tensor processing units — computing chips developed specially for AI — and has found a way to more efficiently exploit Google’s worldwide computing capacity, saving 0.7% of total resources. “It has had substantial impact,” says Kohli.General-purpose AIMost of the successful applications of AI in science so far — including the protein-designing tool AlphaFold — have involved a learning algorithm that was hand-crafted for its task, says Krenn. But AlphaEvolve is general-purpose, tapping the abilities of LLMs to generate code to solve problems in a wide range of domains.DeepMind describes AlphaEvolve as an ‘agent’, because it involves using interacting AI models. But it targets a different point in the scientific process from many other ‘agentic’ AI science systems, which have been used to review the literature and suggest hypotheses.AlphaEvolve is based on the firm’s Gemini family of LLMs. Each task starts with the user inputting a question, criteria for evaluation and a suggested solution, for which the LLM proposes hundreds or thousands of modifications. An ‘evaluator’ algorithm then assesses the modifications against the metrics for a good solution (for example, in the task of assigning Google’s computing jobs, researchers want to waste fewer resources).On the basis of which solutions are judged to be the best, the LLM suggests fresh ideas and over time the system evolves a population of stronger algorithms, says Matej Balog, an AI scientist at DeepMind who co-led the research. “We explore this diverse set of possibilities of how the problem can be solved,” he says.AlphaEvolve builds on the firm’s FunSearch system, which in 2023 was shown to use a similar evolutionary approach to outdo humans in unsolved problems in maths. Compared with FunSearch, AlphaEvolve can handle much larger pieces of code and tackle more complex algorithms across a wide range of scientific domains, says Balog.DeepMind says that AlphaEvolve has come up with a way to perform a calculation, known as matrix multiplication, that in some cases is faster than the fastest-known method, which was developed by German mathematician Volker Strassen in 1969. Such calculations involve multiplying numbers in grids and are used to train neural networks. Despite being general-purpose, AlphaEvolve outperformed AlphaTensor, an AI tool described by the firm in 2022 and designed specifically for matrix mechanics.The approach could be used to tackle optimization problems, says Krenn, or anywhere in science where there are concrete metrics, or simulations, to evaluate what makes a good solution. This could include designing new microscopes, telescope or even materials, he adds.Narrow applicationsIn mathematics, AlphaEvolve seems to allow significant speed-ups in tackling some problems, says Simon Frieder, a mathematician and AI researcher at the University of Oxford, UK. But it will probably be applied only to the “narrow slice” of tasks that can be presented as problems to be solved through code, he says.Other researchers are reserving judgement about the tool’s usefulness until has been trialled outside DeepMind. “Until the systems have been tested by a broader community, I would stay sceptical and take the reported results with a grain of salt,” says Huan Sun, an AI researcher at the Ohio State University in Columbus. Frieder says he will wait until an open-source version is recreated by researchers, rather than a rely on DeepMind’s proprietary system, which could be withdrawn or changed.Although AlphaEvolve requires less computing power to run than AlphaTensor, it is still too resource-intensive to be made freely available on DeepMind’s servers, says Kohli.But the company hopes that announcing the system will encourage researchers to suggest areas of science in which to apply AlphaEvolve. “We are definitely committed to make sure that the most people in the scientific community get access to it,” says Kohli.This article is reproduced with permission and was first published on May 14, 2025.

Google DeepMind’s AI systems have taken big scientific strides in recent years — from predicting the 3D structures of almost every known protein in the universe to forecasting weather more accurately than ever before.  The UK-based lab today unveiled its latest advancement: AlphaEvolve, an AI coding agent that makes large language models (LLMs) like Gemini better at solving complex computing and mathematical problems.  AlphaEvolve is powered by the same models that it’s trying to improve. Using Gemini, the agent proposes programs — written in code — that try to solve a given problem. It runs each code snippet through automated tests that evaluate how accurate, efficient, or novel it is. AlphaEvolve keeps the top-performing code snippets and uses them as the basis for the next round of generation. Over many cycles, this process “evolves” better and better solutions. In essence, it is a self-evolving AI.   DeepMind has already used AlphaEvolve to tackle data centre energy use, design better chips, and speed up AI training. Here are five of its top feats so far.  1. It discovered new solutions to some of the world’s toughest maths problems AlphaEvolve was put to the test on over 50 open problems in maths, from combinatorics to number theory. In 20% of cases, it improved on the best-known solutions to them. The 💜 of EU techThe latest rumblings from the EU tech scene, a story from our wise ol' founder Boris, and some questionable AI art. It's free, every week, in your inbox. Sign up now! One of those was the 300-year-old kissing number problem. In 11-dimensional space, AlphaEvolve discovered a new lower bound with a configuration of 593 spheres — progress that even expert mathematicians hadn’t reached.  2. It made Google’s data centres more efficient The AI agent devised a way to better manage power scheduling at Google’s data centres. That has allowed the tech giant to improve its data centre energy efficiency by 0.7% over the last year — a significant cost and energy saver given the size of its data centre operation.  3. It helped train Gemini faster AlphaEvolve improved the way matrix multiplications are split into subproblems, a core operation in training AI models like Gemini. That optimisation sped up the process by 23%, reducing Gemini’s total training time by 1%. In the world of generative AI, every percentage point can translate into cost and energy savings.  4. It co-designed part of Google’s next AI chip The agent is also using its code-writing skills to rewire things in the physical world. It rewrote a portion of an arithmetic circuit in Verilog — a language used for chip design — making it more efficient. That same logic is now being used to develop Google’s future TPU (Tensor Processing Unit), an advanced chip for machine learning.   5. It beat a legendary algorithm from 1969 For decades, Strassen’s algorithm was the gold standard for multiplying 4×4 complex matrices. AlphaEvolve found a more efficient solution — using fewer scalar multiplications. This could lead to more advanced LLMs, which rely heavily on matrix multiplication to function. According to DeepMind, these feats are just the tip of the iceberg for AlphaEvolve. The lab envisions the agent solving countless problems, from discovering new materials and drugs to streamlining business operations. AI’s evolution will be a hot topic at TNW Conference, which takes place on June 19-20 in Amsterdam. Tickets for the event are now on sale — use the code TNWXMEDIA2025 at the checkout to get 30% off. Story by Siôn Geschwindt Siôn is a freelance science and technology reporter, specialising in climate and energy. From nuclear fusion breakthroughs to electric vehic (show all) Siôn is a freelance science and technology reporter, specialising in climate and energy. From nuclear fusion breakthroughs to electric vehicles, he's happiest sourcing a scoop, investigating the impact of emerging technologies, and even putting them to the test. He has five years of journalism experience and holds a dual degree in media and environmental science from the University of Cape Town, South Africa. When he's not writing, you can probably find Siôn out hiking, surfing, playing the drums or catering to his moderate caffeine addiction. You can contact him at: sion.geschwindt [at] protonmail [dot] com Get the TNW newsletter Get the most important tech news in your inbox each week. Also tagged with