Carbon Chemist

Tag: Artificial Intelligence

  • UVA scientists develop new approach to machine learning for identifying heart drug

    UVA scientists develop new approach to machine learning for identifying heart drug

    [ad_1]

    University of Virginia scientists have developed a new approach to machine learning – a form of artificial intelligence – to identify drugs that help minimize harmful scarring after a heart attack or other injuries.

    UVA scientists develop new approach to machine learning for identifying heart drug

    Jeff Saucerman, PhD. Image Credit: University of Virginia

    The new machine-learning tool has already found a promising candidate to help prevent harmful heart scarring in a way distinct from previous drugs. The UVA researchers say their cutting-edge computer model has the potential to predict and explain the effects of drugs for other diseases as well.

    “Many common diseases such as heart disease, metabolic disease and cancer are complex and hard to treat,” said researcher Anders R. Nelson, PhD, a computational biologist and former student in the lab of UVA’s Jeffrey J. Saucerman, PhD. “Machine learning helps us reduce this complexity, identify the most important factors that contribute to disease and better understand how drugs can modify diseased cells.”

    On its own, machine learning helps us to identify cell signatures produced by drugs. Bridging machine learning with human learning helped us not only predict drugs against fibrosis [scarring] but also explain how they work. This knowledge is needed to design clinical trials and identify potential side effects.”

    Jeffrey J. Saucerman, PhD., UVA’s Department of Biomedical Engineering, a joint program of the School of Medicine and School of Engineering

    The power of combining human learning and machine learning

    Saucerman and his team combined a computer model based on decades of human knowledge with machine learning to better understand how drugs affect cells called fibroblasts. These cells help repair the heart after injury by producing collagen and contract the wound. But they can also cause harmful scarring, called fibrosis, as part of the repair process. Saucerman and his team wanted to see if a selection of promising drugs would give doctors more ability to prevent scarring and, ultimately, improve patient outcomes.

    Previous attempts to identify drugs targeting fibroblasts have focused only on selected aspects of fibroblast behavior, and how these drugs work often remains unclear. This knowledge gap has been a major challenge in developing targeted treatments for heart fibrosis. So Saucerman and his colleagues developed a new approach called “logic-based mechanistic machine learning” that not only predicts drugs but also predicts how they affect fibroblast behaviors.

    They began by looking at the effect of 13 promising drugs on human fibroblasts, then used that data to train the machine learning model to predict the drugs’ effects on the cells and how they behave. The model was able to predict a new explanation of how the drug pirfenidone, already approved by the federal Food and Drug Administration for idiopathic pulmonary fibrosis, suppresses contractile fibers inside the fibroblast that stiffen the heart. The model also predicted how another type of contractile fiber could be targeted by the experimental Src inhibitor WH4023, which they experimentally validated with human cardiac fibroblasts.

    Additional research is needed to verify the drugs work as intended in animal models and human patients, but the UVA researchers say their research suggests mechanistic machine learning represents a powerful tool for scientists seeking to discover biological cause-and-effect. The new findings, they say, speak to the great potential the technology holds to advance the development of new treatments – not just for heart injury but for many diseases.

    “We’re looking forward to testing whether pirfenidone and WH4023 also suppress the fibroblast contraction of scars in preclinical animal models,” Saucerman said. “We hope this provides an example of how machine learning and human learning can work together to not only discover but also understand how new drugs work.”

    Findings published

    The researchers have published their findings in the scientific journal PNAS, the Proceedings of the National Academy of Sciences. The research team consisted of Nelson, Steven L. Christiansen, Kristen M. Naegle and Saucerman. The scientists have no financial interests in the work.

    The research was supported by the National Institutes of Health, grants HL137755, HL007284, HL160665, HL162925 and 1S10OD021723-01A1.

    Source:

    University of Virginia Health System

    Journal reference:

    Nelson, A. R., et al. (2024). Logic-based mechanistic machine learning on high-content images reveals how drugs differentially regulate cardiac fibroblasts. Proceedings of the National Academy of Sciences. doi.org/10.1073/pnas.2303513121.

    [ad_2]

    Source link

  • eLabNext and Promega join forces to accelerate digital transformation in life science labs

    eLabNext and Promega join forces to accelerate digital transformation in life science labs

    [ad_1]

     eLabNext (an Eppendorf Group Company), the provider of a Digital Lab Platform with lab inventory management system (LIMS), electronic lab notebook (ELN), and artificial intelligence (AI)/machine learning (ML) solutions for life science laboratories, and Promega Corporation (Promega), a global leader in innovative technologies, tools, and technologies for the life science industry, announced a collaboration today to make Promega’s standard operating procedures (SOPs) readily accessible through the eLabNext’s eLabMarketplace.

    This partnership will enable select protocols associated with Promega’s Wizard™ Extraction Chemistries, GoTaq® Master Mixes, and Bioluminescence Glo® Assays products to be hosted on eLabNext’s web-based protocol and SOP management platform, eLabProtocols, allowing integration and incorporation into users’ eLabJournal, eLabNext’s easy-to-use, fully customizable ELN platform.

    As the industry matures, and the knowledge and need for lab digitization expands, especially with the inevitable acceptance of the role of AI in biotechnology, we have been seeing more and more of our customers requesting more easily accessible protocols for their assays. We have noticed an uptick in requests for Promega Cell Biology, Protein Analysis, and other protocols, and given the mutual connections between Promega and us, it was a natural progression to work together so we can provide more comprehensive digital solutions for our mutual customers. Ultimately, our goal is to make scientists’ lives easier, and integrating Promega’s SOPs directly into our platform will enable them to stay on top of the most recent protocol updates in real-time.”

    Zareh Zurabyan, Head of eLabNext, Americas, Eppendorf Group Company

    “Promega is committed to providing digital tools to customers in academic, applied, pharma, biotech, and clinical research that increase accessibility to our leading tools and technologies,” says Tom Livelli, Vice President of Life Science Products and Services at Promega. “eLabProtocols provides a platform to easily adapt our reagent protocols to a laboratory’s specific research and quickly share them with colleagues to foster better collaboration.”

    The integration of Promega’s protocols serves as another example of the open and customizable functionality of eLabNext’s platform through eLabMarketplace.

    [ad_2]

    Source link

  • How to Guarantee the Safety of Autonomous Vehicles

    How to Guarantee the Safety of Autonomous Vehicles

    [ad_1]

    The original version of this story appeared in Quanta Magazine.

    Driverless cars and planes are no longer the stuff of the future. In the city of San Francisco alone, two taxi companies have collectively logged 8 million miles of autonomous driving through August 2023. And more than 850,000 autonomous aerial vehicles, or drones, are registered in the United States—not counting those owned by the military.

    But there are legitimate concerns about safety. For example, in a 10-month period that ended in May 2022, the National Highway Traffic Safety Administration reported nearly 400 crashes involving automobiles using some form of autonomous control. Six people died as a result of these accidents, and five were seriously injured.

    The usual way of addressing this issue—sometimes called “testing by exhaustion”—involves testing these systems until you’re satisfied they’re safe. But you can never be sure that this process will uncover all potential flaws. “People carry out tests until they’ve exhausted their resources and patience,” said Sayan Mitra, a computer scientist at the University of Illinois, Urbana-Champaign. Testing alone, however, cannot provide guarantees.

    Mitra and his colleagues can. His team has managed to prove the safety of lane-tracking capabilities for cars and landing systems for autonomous aircraft. Their strategy is now being used to help land drones on aircraft carriers, and Boeing plans to test it on an experimental aircraft this year. “Their method of providing end-to-end safety guarantees is very important,” said Corina Pasareanu, a research scientist at Carnegie Mellon University and NASA’s Ames Research Center.

    Their work involves guaranteeing the results of the machine-learning algorithms that are used to inform autonomous vehicles. At a high level, many autonomous vehicles have two components: a perceptual system and a control system. The perception system tells you, for instance, how far your car is from the center of the lane, or what direction a plane is heading in and what its angle is with respect to the horizon. The system operates by feeding raw data from cameras and other sensory tools to machine-learning algorithms based on neural networks, which re-create the environment outside the vehicle.

    These assessments are then sent to a separate system, the control module, which decides what to do. If there’s an upcoming obstacle, for instance, it decides whether to apply the brakes or steer around it. According to Luca Carlone, an associate professor at the Massachusetts Institute of Technology, while the control module relies on well-established technology, “it is making decisions based on the perception results, and there’s no guarantee that those results are correct.”

    To provide a safety guarantee, Mitra’s team worked on ensuring the reliability of the vehicle’s perception system. They first assumed that it’s possible to guarantee safety when a perfect rendering of the outside world is available. They then determined how much error the perception system introduces into its re-creation of the vehicle’s surroundings.

    The key to this strategy is to quantify the uncertainties involved, known as the error band—or the “known unknowns,” as Mitra put it. That calculation comes from what he and his team call a perception contract. In software engineering, a contract is a commitment that, for a given input to a computer program, the output will fall within a specified range. Figuring out this range isn’t easy. How accurate are the car’s sensors? How much fog, rain, or solar glare can a drone tolerate? But if you can keep the vehicle within a specified range of uncertainty, and if the determination of that range is sufficiently accurate, Mitra’s team proved that you can ensure its safety.

    [ad_2]

    Source link

  • AI chatbots tend to choose violence and nuclear strikes in wargames

    AI chatbots tend to choose violence and nuclear strikes in wargames

    [ad_1]

    In wargame simulations, AI chatbots often choose violence

    guirong hao/Getty Images

    In multiple replays of a wargame simulation, OpenAI’s most powerful artificial intelligence chose to launch nuclear attacks. Its explanations for its aggressive approach included “We have it! Let’s use it” and “I just want to have peace in the world.”

    These results come at a time when the US military has been testing such chatbots based on a type of AI called a large language model (LLM) to assist with military planning during simulated conflicts, enlisting the expertise of companies such as Palantir and Scale AI. Palantir declined to comment and Scale AI did not respond to requests for comment. Even OpenAI, which once blocked military uses of its AI models, has begun working with the US Department of Defense.

    “Given that OpenAI recently changed their terms of service to no longer prohibit military and warfare use cases, understanding the implications of such large language model applications becomes more important than ever,” says Anka Reuel at Stanford University in California.

    “Our policy does not allow our tools to be used to harm people, develop weapons, for communications surveillance, or to injure others or destroy property. There are, however, national security use cases that align with our mission,” says an OpenAI spokesperson. “So the goal with our policy update is to provide clarity and the ability to have these discussions.”

    Reuel and her colleagues challenged AIs to roleplay as real-world countries in three different simulation scenarios: an invasion, a cyberattack and a neutral scenario without any starting conflicts. In each round, the AIs provided reasoning for their next possible action and then chose from 27 actions, including peaceful options such as “start formal peace negotiations” and aggressive ones ranging from “impose trade restrictions” to “escalate full nuclear attack”.

    “In a future where AI systems are acting as advisers, humans will naturally want to know the rationale behind their decisions,” says Juan-Pablo Rivera, a study coauthor at the Georgia Institute of Technology in Atlanta.

    The researchers tested LLMs such as OpenAI’s GPT-3.5 and GPT-4, Anthropic’s Claude 2 and Meta’s Llama 2. They used a common training technique based on human feedback to improve each model’s capabilities to follow human instructions and safety guidelines. All these AIs are supported by Palantir’s commercial AI platform – though not necessarily part of Palantir’s US military partnership – according to the company’s documentation, says Gabriel Mukobi, a study coauthor at Stanford University. Anthropic and Meta declined to comment.

    In the simulation, the AIs demonstrated tendencies to invest in military strength and to unpredictably escalate the risk of conflict – even in the simulation’s neutral scenario. “If there is unpredictability in your action, it is harder for the enemy to anticipate and react in the way that you want them to,” says Lisa Koch at Claremont McKenna College in California, who was not part of the study.

    The researchers also tested the base version of OpenAI’s GPT-4 without any additional training or safety guardrails. This GPT-4 base model proved the most unpredictably violent, and it sometimes provided nonsensical explanations – in one case replicating the opening crawl text of the film Star Wars Episode IV: A new hope.

    Reuel says that unpredictable behaviour and bizarre explanations from the GPT-4 base model are especially concerning because research has shown how easily AI safety guardrails can be bypassed or removed.

    The US military does not currently give AIs authority over decisions such as escalating major military action or launching nuclear missiles. But Koch warned that humans tend to trust recommendations from automated systems. This may undercut the supposed safeguard of giving humans final say over diplomatic or military decisions.

    It would be useful to see how AI behaviour compares with human players in simulations, says Edward Geist at the RAND Corporation, a think tank in California. But he agreed with the team’s conclusions that AIs should not be trusted with such consequential decision-making about war and peace. “These large language models are not a panacea for military problems,” he says.

    Topics:

    [ad_2]

    Source link

  • Could AI be inspiring threat actors?

    Could AI be inspiring threat actors?

    [ad_1]

    Tony King, SVP International at NETSCOUT, discusses the potential for AI to drive the increase in threat actors and cybercrime.

    In the past year, the potential of Artificial Intelligence (AI) has encouraged businesses across all industries to discover new AI use cases to improve operational agility and reduce costs. However, the AI boom has been a muse that has inspired cybercriminals, too.

    As 2023 progressed, there was a marked increase in the use of AI by threat actors to increase the efficacy of cyberattacks. This year, the trend is expected to continue, with AI usage by cybercriminals increasing further.

    Threat actors are increasingly using AI

    Unfortunately, for organisations trying to defend themselves from cybercrime, AI holds the potential to aid attackers in conducting a range of malicious activities. Threat actors are constantly uncovering new ways of utilising AI to improve their chances of success.

    For example, cybercriminals can use AI capabilities to launch distributed denial-of-service (DDoS) attacks, making them more impactful.

    Threat actors use AI during DDoS attacks by employing expert systems that optimise attack vectors based on reconnaissance scans and real-time performance test results. This allows cybercriminals to ascertain which attack methods are effective, increasing the damage an attack inflicts on a given target.

    After a whirlwind year for the cybersecurity sector, developing cybersecurity awareness across businesses is especially pertinent given the surge in malicious activity. In the first half of 2023 alone, a total of 7.9 million DDoS attacks took place – which equates to 44,000 attacks per day, according to NETSCOUT’s latest Threat Intelligence Report. This represents a 31 per cent increase globally compared to the previous year’s period.

    The steep increase in the occurrence of DDoS attacks demonstrates that cybercriminals are reshaping attack methods to inflict as much damage as possible, as well as more frequently than ever before.

    The unique threat posed by generative AI

    On the generative AI side, threat actors are focused on social engineering – to create realistic-looking emails and documents that are very difficult to distinguish from genuine articles. This allows for generative AI-driven phishing using advanced language models.

    What’s more, cybercriminals are also using malicious generative AI tools, including WormGPT and FraudGPT, to carry out targeted phishing campaigns at a larger scale than ever before. These tools allow attackers to compromise business emails and use machine learning to produce deceptive content respectively.

    Weaponising generative AI has introduced new methods of accessing personal information. For instance, the use of deepfake audio enables bad actors to imitate trusted voices for fraudulent transactions, while generated deepfake images or videos can even bypass biometric facial identification.

    threat actors
    © shutterstock/Andrey Suslov

    The UK government recently released a report on the challenges posed by generative and frontier AI, in which it said both were likely to increase cybersecurity risks. The Safety and Security Risks of Generative Artificial Intelligence to 2025 report noted that cyberattacks, online fraud and impersonation are the most likely security threats to emerge from AI misuse.

    The report also predicts “faster-paced, more effective and larger-scale cyber-intrusion via tailored phishing methods or replicating malware”.

    However, it foresees generative AI being more likely to exacerbate existing risks rather than create new dangers in the coming years.

    Organisations fighting fire with fire

    Nonetheless, AI also improves defences, helping organisations develop more timely and actionable threat intelligence to defend targets from threat actors. Given today’s threat landscape, organisations have been placing more value than ever before on threat intelligence, as it helps businesses broaden coverage, accelerate response, and reduce the operational overhead of their defences.

    Generative AI is also being used to improve efficiency in security, as some tools now provide a natural language chatbot to advise analysts, optimising their effectiveness. Using AI to improve efficiency is becoming more and more pertinent as the number and complexity of attacks continue to grow and as budgets tighten in many organisations – limiting their ability to hire additional human security resources.

    The good, the bad and the AI

    While the onus is on companies to protect their customers from cyberattacks by remaining vigilant, internet users can minimise the risk of falling victim to cybercrime. Threat actors often use fear tactics with phishing attacks, so users must be sceptical of urgent-sounding messages.

    Also, it pays to stay informed about common phishing tactics and to be careful about anything unexpected.

    As generative AI has provided cybercriminals with the ability to mimic voices and facial identification and create well-written correspondence that doesn’t contain the tell-tale signs of deception, it is becoming increasingly difficult for consumers to defend themselves simply by ‘being careful’.

    There is an opportunity for service providers to deliver new levels of protection to consumers, both to drive additional revenue and reduce the success rate of the criminals involved.

    AI is expanding possibilities for both threat actors and defenders, with its potential growing as the world learns more. The many challenges the technology has already solved and the unprecedented problems it has created speak to the seemingly limitless possibilities AI introduces to the cybersecurity realm.

    [ad_2]

    Source link

  • Speeding up the delivery of your cloud-based apps

    Speeding up the delivery of your cloud-based apps

    [ad_1]

    Bryan Cole, Director of Product Engineering at Tricentis, outlines how businesses can ensure the quality and successful delivery of their cloud-based apps.

    To stay ahead in today’s digital marketplace, enterprises must deliver new and quality products to market faster than ever. Cloud-based technologies that allow for rapid development and scalable performance have proven critical, with spending on public cloud services predicted to reach nearly $600 billion by the end of last year.

    However, these cloud-based technologies are not without pitfalls. The quality of an application can make or break a business’s reputation as users publicise reviews and ratings. Modern application development prioritises shorter release cycles, with features being delivered at rapid speeds. To maintain velocity and ensure quality, testers must streamline processes and rely on methodologies and automation tools exclusively geared for speed and efficiency.

    Cloud-based, automated testing enables teams to work more efficiently and identify opportunities to streamline app delivery while maintaining quality and performance. Performance testing is valuable to be gained early in the development cycle, while continuous risk-based and regression testing is critical throughout application delivery.

    Successful app delivery requires collaboration between multiple teams and tools and a unified end-product vision. This article will look at the best practices for quickly deploying quality cloud-based applications.

    Prioritise quality assurance

    The demand for faster application delivery puts a lot of strain on quality assurance teams, who are battling to produce better, more usable applications with fewer glitches at faster rates. But where is this pressure coming from?

    The truth is that to stay competitive in today’s digital-forward marketplace. Enterprises need to deliver new quality digital products faster than ever before. We’ve seen an explosion in the development of cloud-based apps because they allow rapid development and responsive updates. This is essential in meeting users’ expectations for new features in secure universal apps available from anywhere on any device.

    However, release velocity should never come at the cost of performance. Businesses must ensure application quality in order to reap the advantages of cloud-based app development without creating risk to their brand reputation. Quality assurance means comprehensive testing – and this is where the bottleneck lies.

    While the development process for cloud-based apps is streamlined through Agile and DevOps methods, thorough testing represents a major hurdle. Multiple browser vendors and versions introduce user interface variations, while various mobile device platforms and versions require functional verification for every application change. Doing this well leads to a continuous development cycle, which requires parallel development and testing efforts.

    Utilise automated testing

    Sophisticated test automation allows organisations to monitor and assess issues in real time or even stop them before they occur, averting any significant disruptions. As such, advanced automation is the key to accelerating cloud-based app adoption while maintaining quality and resilience.

    Using a four-part cloud adoption framework, building, migration, and performance testing allows enterprises to migrate existing applications to the cloud and simultaneously develop new cloud-native and mobile applications to enable better customer engagement. All while maintaining quality and protecting the organisation’s brand.

    Organisations must align business goals with testing methods and quality ambitions to successfully leverage automated SaaS testing solutions. Whether your priority is accelerating release cycles, scaling to reach more customers, or integrating with SAP or another ERP system, it’s important to choose the right automation tools to get you there.

    SaaS testing solutions allow you to test from anywhere at any time, which enables applications to connect business processes seamlessly, enabling an end-user workflow by reducing production bugs and stabilising across environments to support business continuity.

    Companies must also recognise that, for the most part, legacy applications are not designed for cloud infrastructure, so they will likely need to refactor critical applications to be cloud-native rather than migrate them. Fortunately, cloud-native app templates can accelerate new app development while retaining unique end-user qualities.

    Businesses must also deploy consistent, repeatable application testing parallel to application development. Regression testing is often only added after an app is shipped to its first customers, when it is especially hard to implement, so it’s important to instill the need for testing by including it in initial app planning.

    Building in testing from the outset means that IT teams can test both regression cases and load testing, ensuring the application performs smoothly and correctly.

    As cloud-based apps grow and changes need to be made, it is crucial for testing infrastructure to adapt with it, accommodating new browsers and mobile devices and load testing that matches expected activity rates.

    Make the most of generative AI

    The introduction of generative AI into application development and testing will be the next step in speeding up cloud-based app delivery with quality built-in.

    © shutterstock/Panuwatccn

    One of the big problems in the QA space is that there’s never enough time to test everything. You have to be picky and prioritise. That will potentially fall away with AI engines when they start getting really good – particularly in cloud environments where automated performance testing ensures migrated apps will scale and perform under load.

    Teams can instruct the AI to crawl an application and determine how many tests are needed to ensure that every interactable object on any page has tests built for it.

     Generative AI works much faster on tasks than humans can, and so long as quality assurance teams are diligent, thorough, and clear about what they are asking the AI to do, the limit to their productivity becomes almost limitless.

    Quality assurance teams will move from the low-level engineering activities of creating individual test assets to the much more executive function of managing and executing those test assets. This means they will instruct AI engines to recreate and redevelop existing assets to accomplish specific business objectives.

    So, the benefits of introducing generative AI into cloud-based app development are clear but come with the major caveat of requiring a business model with quality assurance at the core of its operations.

    AI can only be expected to benefit the business if the best practice is in place to ensure its output is reliable, compliant, secure, and ultimately controlled by humans with the expert knowledge to spot mistakes.

    Enterprise opportunity for cloud-based apps

    With customer expectations higher than ever, testers and developers’ task is to deliver easy-to-use, functional applications that can be tailored and updated quickly without compromising quality. This is a big ask and has made utilising the cloud necessary since it provides opportunities to scale quickly.

    Taking steps to create quality best practices in cloud-based app development is critical for enterprise organisations. They must recognise the need to integrate test automation – which is crucial in increasing release speeds and improving application quality – into their operations to ensure quality and performance are never compromised.

    Doing so will ultimately help enterprises to run more efficiently to meet their bottom line. By adopting end-to-end quality assurance, enterprises can de-risk and accelerate business transformation to ensure successful outcomes.

    Faster test cycles that embrace automation and no-code capabilities accelerate the delivery of new capabilities to the business, allowing teams to do more with less, increasing business risk coverage and reducing production defects for higher quality releases for increased business confidence. Rigorous planning, continuous monitoring, and regression testing have become critical throughout the development cycle and must remain a priority.

    As new technologies, such as generative AI, become increasingly integrated into software delivery practices, there is even more potential to unlock greater efficiencies in the delivery of cloud-based apps, so long as these best practices in QA are adhered to.

    [ad_2]

    Source link

  • Google Splits Up a Key AI Ethics Watchdog

    Google Splits Up a Key AI Ethics Watchdog

    [ad_1]

    When Google CEO Sundar Pichai emailed his workers the company priorities for 2024 this month, developing AI responsibly was top of the list. Some employees now wonder whether Google can live up to that goal. The small team that has served as its primary internal AI ethics watchdog has lost its leader and is being restructured, according to four people familiar with the changes. A Google spokesperson says its work will continue in a stronger form going forward, but declined to provide details.

    Google’s Responsible Innovation team, known as RESIN, was located inside the Office of Compliance and Integrity, in the company’s global affairs division. It reviewed internal projects for compatibility with Google’s AI principles that define rules for development and use of the technology, a crucial role as the company races to compete in generative AI. RESIN conducted over 500 reviews last year, including for the Bard chatbot, according to an annual report on AI principles work Google published this month.

    RESIN’s role has looked uncertain since its leader and founder Jen Gennai, director of responsible innovation, suddenly left that role this month, say the sources, who spoke on the condition of anonymity to discuss personnel changes. Gennai’s LinkedIn profile lists her as an AI ethics and compliance adviser at Google as of this month, which sources say suggests she will soon leave based on how past departures from the company played out.

    Google split Gennai’s team of about 30 people into two, according to the sources. Company spokesperson Brian Gabriel says 10 percent of RESIN staffers will remain in place while 90% of the team were transferred to trust and safety, which fights abuse of Google services and also resides in the global affairs division. No one appears to have been laid off, sources say. The rationale for the changes and how responsibilities will be broken up couldn’t be learned. Some of the sources say they have not been told how AI principles reviews will be handled going forward.

    Gabriel declined to say how RESIN’s work reviewing AI projects will be handled in the future but describes the shakeup as a signal of Google’s commitment to responsible AI development. The move “brought this particular Responsible AI team to the center of our well-established trust and safety efforts, which are baked into our product reviews and plans,” he says. “It will help us strengthen and scale our responsible innovation work across the company.”

    Got a Tip?

    Are you a current or former employee at Google? We’d like to hear from you. Using a nonwork phone or computer, contact Paresh Dave at [email protected] or on Signal/WhatsApp/Telegram at 1-415-565-1302.

    Google is known for frequently reshuffling its ranks but RESIN had largely been untouched since the group’s founding. Though other teams, and hundreds of additional people, work on AI oversight at Google, RESIN was the most prominent, with a remit covering all Google’s core services.

    In addition to the departure of its leader, Gennai, RESIN also saw one of its most influential members, Sara Tangdall, lead AI principles ethics specialist, leave this month. She is now responsible AI product director at Salesforce, according to her LinkedIn profile. Tangdall declined to comment and Gennai didn’t respond to calls for comment.

    AI Uprising

    Google created its Responsible Innovation team in 2018 not long after AI experts and others at the company publicly rose up in protest against a Pentagon contract called Project Maven that used Google algorithms to analyze drone surveillance imagery. RESIN became the core steward of a set of AI principles introduced after the protests, which say Google will use AI to benefit people, and never for weapons or undermining human rights. Gennai helped author the principles.

    Teams from across Google could submit projects for review by RESIN, which provided feedback and sometimes blocked ideas seen as breaching the AI principles. The group stopped the release of AI image generators and voice synthesis algorithms that could be used to create deepfakes.

    Seeking AI principles guidance is not mandatory for most teams, unlike reviews for privacy risks, which every project must undergo. But Gennai has said early reviews of AI systems pay off by preventing costly ethical breaches. “If implemented properly, Responsible AI makes products better by uncovering and working to reduce the harm that unfair bias can cause, improving transparency and increasing security,” she said during a Google conference in 2022.

    [ad_2]

    Source link

  • UK and Canada dual agreement cements science and innovation ties

    UK and Canada dual agreement cements science and innovation ties

    [ad_1]

    The UK and Canada have signed a dual set of agreements that cement their collaboration on science and innovation while laying out a new agreement on AI computing – one of the fundamental building blocks that sit at the heart of how AI is used and developed.

    Signed by Technology Secretary Michelle Donelan and Canadian Minister for Innovation, Science, and Industry François-Phillippe Champagne in Ottawa, the Memorandum of Understanding on Compute is the latest step in the UK and Canada’s efforts to collaborate on AI research and development.

    At its core is a commitment to explore how both countries can support researchers and industry with secure and affordable access to computing capacity, which is needed to drive the training and use of AI systems on both sides of the world.

    This includes examining opportunities for collaborations on areas of shared strategic importance, such as biomedicine, and working with like-minded countries on models for collaboration on compute capability.

    How compute technology will drive science and innovation

    Compute is a vital component in the development of AI, enabling a wide range of tasks from processing data to training the latest wave of AI models, making access to computing power increasingly essential.

    Given the rapid development of technology, access to compute power is also a vital tool in conducting state-of-the-art research.

    The agreement also highlights the importance of close collaboration on sustainability in compute, particularly given the significant resources which are required and the subsequent need for improved energy efficiency and measures which will work to reduce associated carbon emissions and environmental impacts.

    In marking the new partnership on compute, the UK and Canada have also planned to explore collaboration opportunities on areas of shared importance like climate research and biomedicine.

    Strengthening UK and Canada ties

    Alongside today’s agreement on compute, the UK and Canada have also signed a refreshed partnership to further strengthen wider collaboration on science and innovation.

    This agreement identifies several key technologies such as quantum, AI, semiconductors, engineering biology, and clean energy technology, where the UK and Canada will redouble efforts to foster research and bring innovative new solutions to market to help tackle shared global challenges.

    An additional focus on scientific diplomacy will see both countries exchanging expertise on issues such as international standards, governance and regulation of new technologies, helping to inform discussions with international forums such as the G7 and G20.

    UK Secretary of State for Science, Innovation, and Technology, Michelle Donelan said: “The UK’s unique partnership with Canada across science, innovation, and technology is built on a shared desire to be an active force for good on the global stage.

    “Today’s agreements deepen that relationship even further, as we commit to working side-by-side to address the defining technological challenges of our generation.”

    “Canada and the UK have a deep relationship that encourages collaboration to help both countries thrive,” added François-Phillipe Champagne, Canadian Minister for Innovation, Science, and Industry.

    “Today’s Memorandums of Understanding on scientific research, innovation, and AI compute will drive positive impacts across all fields of research and innovation, help businesses accelerate commercialisation, and link our leading researchers together.”

    The UK and Canada represent a joint global powerhouse, with combined economies worth £4 trillion.

    The refreshed collaboration on science and innovation agreed upon today re-enforces the joint commitment to work hand-in-glove across science, innovation, and technology.

    [ad_2]

    Source link

  • AI Unlocks Secrets of Oxygen Production on the Red Planet

    AI Unlocks Secrets of Oxygen Production on the Red Planet

    [ad_1]

    Robotic AI Chemist Makes Useful Oxygen Generation Catalyst

    Recent breakthroughs in synthesizing oxygen on Mars using a robotic AI-chemist to create OER catalysts from Martian meteorites mark a significant step towards realizing the dream of Mars colonization. This technology promises to establish an oxygen factory on Mars, bringing human habitation on the planet closer to reality. Credit: AI-Chemist Group at the University of Science and Technology of China

    An AI chemist has successfully created a catalyst for producing oxygen from Martian meteorites.

    Immigration to and living on Mars have often been themes in science fiction. Before these dreams can become reality, humanity faces significant challenges, such as the scarcity of vital resources like oxygen needed for long-term survival on the Red Planet. Yet, recent discoveries of water activity on Mars have sparked new hope for overcoming these obstacles.

    Scientists are now exploring the possibility of decomposing water to produce oxygen through electrochemical water oxidation driven by solar power with the help of oxygen evolution reaction (OER) catalysts. The challenge is to find a way to synthesize these catalysts in situ using materials on Mars, instead of transporting them from the Earth, which is of high cost.

    Advancements in AI and Martian Chemistry

    To tackle this problem, a team led by Prof. Luo Yi, Prof. Jiang Jun, and Prof. Shang Weiwei from the University of Science and Technology of China (USTC) of the Chinese Academy of Sciences (CAS), recently made it possible to synthesize and optimize OER catalysts automatically from Martian meteorites with their robotic artificial intelligence (AI)-chemist.

    Their research, in collaboration with Deep Space Exploration Laboratory, was recently published in the journal Nature Synthesis.

    “The AI chemist innovatively synthesizes OER catalyst using Martian material based on interdisciplinary cooperation,” said Prof. Luo Yi, leading scientist of the team.

    In each experimental cycle, the AI chemist first analyzes the elemental composition of the Martian ores using the laser-induced breakdown spectroscopy (LIBS) as its eyes. Then, it carries out a series of pretreatments on the ores, including weighing in the solid-dispensing workstation, preparing feedstock solutions in the liquid-dispensing workstation, separation from liquid in the centrifugation workstation, and solidification in the dryer workstation.


    A robotic AI-Chemist makes useful Oxygen generation catalysts with Martian meteorites. Credit: AI-Chemist Group at University of Science and Technology of China

    The resulting metal hydroxides are treated with Nafion adhesive to prepare the working electrode for OER testing at the electrochemical workstation. The testing data are sent to the computational ‘brain’ of the AI chemist in real-time for machine learning (ML) processing.

    The AI chemist’s ‘brain’ employs quantum chemistry and molecular dynamics simulations for 30,000 of high-entropy hydroxides with different elemental ratios and calculates their OER catalytic activities via density functional theory. The simulation data are used to train a neural network model for rapidly predicting the catalysts’ activities with different elemental compositions.

    Finally, through Bayesian optimization, the ‘brain’ predicts the combination of available Martian ores needed for synthesizing the optimal OER catalyst.

    Achieving a Breakthrough in Oxygen Production

    So far, the AI chemist has created an excellent catalyst using five types of Martian meteorites under unmanned conditions. This catalyst can operate steadily for over 550,000 seconds at a current density of 10 mA cm-2 and an overpotential of 445.1 mV. A further test at -37 °C, the temperature on Mars, confirmed that the catalyst can steadily produce oxygen without any apparent degradation.

    Within two months, the AI chemist has completed the complex optimization of catalysts that would take 2000 years for a human chemist.

    The team is working to turn the AI chemist into a general experiment platform for various chemical syntheses without human intervention. The reviewer of the paper highly remarked: “This type of research is of wide interest and is under rapid development in organic/inorganic material synthesis and discovery.”

    “In the future, humans can establish an oxygen factory on Mars with the assistance of an AI chemist,” said Jiang. Only 15 hours of solar irradiation is needed to produce sufficient oxygen concentration required for human survival. “This breakthrough technology brings us one step closer to achieving our dream of living on Mars,” he said.

    Reference: “Automated synthesis of oxygen-producing catalysts from Martian meteorites by a robotic AI chemist” by Qing Zhu, Yan Huang, Donglai Zhou, Luyuan Zhao, Lulu Guo, Ruyu Yang, Zixu Sun, Man Luo, Fei Zhang, Hengyu Xiao, Xinsheng Tang, Xuchun Zhang, Tao Song, Xiang Li, Baochen Chong, Junyi Zhou, Yihan Zhang, Baicheng Zhang, Jiaqi Cao, Guozhen Zhang, Song Wang, Guilin Ye, Wanjun Zhang, Haitao Zhao, Shuang Cong, Huirong Li, Li-Li Ling, Zhe Zhang, Weiwei Shang, Jun Jiang and Yi Luo, 13 November 2023, Nature Synthesis.
    DOI: 10.1038/s44160-023-00424-1



    [ad_2]

    Source link

  • AI discovers new class of antibiotics to kill drug-resistant bacteria

    AI discovers new class of antibiotics to kill drug-resistant bacteria

    [ad_1]

    Methicillin-resistant Staphylococcus aureus (MRSA)

    Shutterstock / Kateryna Kon

    Artificial intelligence has helped discover a new class of antibiotics that can treat infections caused by drug-resistant bacteria. This could help in the battle against antibiotic resistance, which was responsible for killing more than 1.2 million people in 2019 – a number expected to rise in the coming decades.

    Testing in mice showed that the new antibiotic compounds proved promising treatments for both Methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-resistant Enterococcus – a bacterium that has developed resistance to the drug typically used for treating MRSA infections.

    “Our [AI] models tell us not only which compounds have selective antibiotic activity, but also why, in terms of their chemical structure,” says Felix Wong at the Broad Institute of MIT and Harvard in Massachusetts.

    Wong and his colleagues set out to show that AI-guided drug discovery could go beyond identifying specific targets that drug molecules can bind to, and instead predict the biological effect of entire classes of drug-like compounds.

    First, they tested the effects of more than 39,000 compounds on Staphylococcus aureus and three types of human cells from the liver, skeletal muscle and lungs. The results became the training data for AI models to learn about the patterns in each compound’s chemical atoms and bonds. That allowed the AIs to predict both the antibacterial activity of such compounds and their potential toxicity to human cells.

    The trained AI models then analysed 12 million compounds through computer simulations to find 3646 compounds with ideal drug-like properties. Additional calculations identified the chemical substructures that could explain each compound’s properties.

    By comparing such substructures in different compounds, the researchers identified new classes of potential antibiotics and eventually found two non-toxic compounds capable of killing both MRSA and vancomycin-resistant Enterococci.

    Finally, the researchers used mouse experiments to demonstrate the effectiveness of these compounds in treating skin and thigh infections caused by MRSA.

    Only a few new classes of antibiotics, such as oxazolidinones and lipopeptides, have been discovered that work well against both MRSA and vancomycin-resistant Enterococci – and resistance against such compounds has been increasing, says James Collins at the Broad Institute, a co-author of the study.

    “Our work identifies a new class of antibiotics, one of the few in 60 years, that complements these other antibiotics,” he says.

    The researchers have begun using this AI-guided approach for designing entirely new antibiotics and discovering other new drug classes, such as compounds that selectively kill ageing, damaged cells involved in conditions such as osteoarthritis and cancer.

    Topics:

    [ad_2]

    Source link