Artificial intelligence Archives | Page 8 of 657

New Stanford scaling method could make AI training cheaper

Researchers at Stanford University have introduced a new approach to scaling laws that could significantly reduce the computational cost of predicting how large language models will perform as they grow.

Scaling laws are used to estimate how smaller models will behave before developers commit to expensive large-scale training runs. These predictions are central to modern AI development, where training advanced models can require enormous computing resources and financial investment.

A research team led by Sanmi Koyejo and Sang Truong developed a framework called Item Response Scaling Laws, or IRSL, which draws on measurement science and educational testing methods. The approach adapts techniques similar to those used in standardised exams to evaluate model capabilities with far fewer test queries.

According to Stanford HAI, IRSL can reduce computational demand by more than 99% while maintaining or improving predictive accuracy. Instead of running every model through large evaluation sets, the method uses carefully selected questions to estimate capability more efficiently.

Researchers argue that the approach could make AI development more accessible, particularly for academic institutions and smaller research teams that lack the computing budgets of major technology companies. It could also help large commercial developers reduce the cost of experimentation before training larger models.

The method remains a research advance rather than a direct reduction in the full cost of training frontier models. However, by making performance prediction cheaper and more statistically rigorous, it could change how developers plan and evaluate future AI systems.

Why does it matter?

AI development is increasingly shaped by access to computing power, which gives the largest technology companies a major advantage. If methods such as IRSL can make model evaluation and scaling predictions far cheaper, they could lower barriers for researchers, universities and smaller developers, while making AI experimentation faster and less resource-intensive.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our chatbot!

Switzerland advances National Cyberstrategy implementation

Switzerland has reported progress in implementing its National Cyberstrategy, with more than 90 projects underway and new measures addressing the role of AI in cybersecurity.

The Federal Council was informed of the 2025 implementation report. The implementation report was prepared by the National Cyberstrategy Steering Committee together with the National Cyber Security Centre. The report tracks work across five objectives:

Empowering the public
Securing digital services and critical infrastructure
Managing cyberattacks
Combating cybercrime
Strengthening international cooperation

The report identifies AI as an important area influencing both cybersecurity risks and defensive capabilities. The report describes measures related to AI-assisted cyber threats, AI-supported cyberdefence, research projects, and public awareness activities.

The report also refers to regulatory safeguards linked to Switzerland’s ratification of the Council of Europe Convention on AI. The report frames those steps as part of a broader response to the growing importance of AI in cybersecurity.

According to the report, the National Cyber Security Centre has received 222 reports since mandatory reporting requirements for cyberattacks on critical infrastructure entered into force in April 2025. Authorities say the reports improve national cyber situational awareness and support coordinated responses to threats.

The report also highlights developments involving sector-specific cybersecurity centres, information-sharing initiatives, and vulnerability management programmes. Switzerland also continued its federal bug bounty programme and other vulnerability management initiatives.

Capacity-building programmes include the Cyber-Defence Campus Fellowship, the Cyber Startup Challenge, and the national S-U-P-E-R.ch awareness campaign. The report also notes information-sharing work through Cyber-CASE, Cyber-STRAT, and NEDIK to support faster handling of digital crimes.

International activities included participation in cyber diplomacy and capacity-building initiatives linked to Geneva Cyber Week and UN and OSCE processes.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!

Singapore and Google expand partnership on frontier AI deployment

Google and Singapore’s Ministry of Digital Development and Information have announced an expanded National AI Partnership to accelerate the deployment of frontier AI across the economy and public sector.

The partnership builds on earlier collaboration between Google and Singapore’s digital authorities and aims to support healthcare innovation, scientific research, workforce development, enterprise transformation and AI governance. The Ministry said the initiative supports Singapore’s National AI Strategy by deploying AI at scale for economic growth and public good.

A major focus is on healthcare and life sciences. Google DeepMind is exploring collaboration with Singapore’s public health clusters on AI co-clinician research, including systems that could support doctors and patients during care journeys under the clinical authority of physicians.

Google DeepMind will also work with the National Research Foundation to train local researchers on agentic AI tools for science, while Google and A*STAR will collaborate on AI-enabled tools for scientific research and analysis in materials and life sciences. The partnership also includes work on a Gemma-powered running assistant for blind and low-vision athletes, in collaboration with SG Enable.

Education and workforce development are another pillar. Google has enabled advanced AI features in Google Workspace for Education for educators from primary schools to junior colleges, while the Ministry of Education and Google will expand collaboration on teacher training, upskilling and AI-supported teaching and learning.

The partnership also covers enterprise transformation and AI governance. Google Cloud’s Forward Deployed Engineers will support Singapore-based companies working on agentic enterprise transformation, while Singapore agencies and Google are testing how ‘computer use’ AI agents behave in real-world settings through an AI Agents Sandbox.

Singapore and Google will also collaborate on AI safety, including the development of multimodal and multilingual safety benchmarks with IMDA and MLCommons. The work is intended to support responsible AI deployment that reflects local languages, cultures and governance needs.

Why does it matter?

The partnership shows how frontier AI is moving from experimentation into national deployment strategies. Singapore is using public-private collaboration to test AI in healthcare, research, education, enterprise workflows and governance, while also building safeguards around agentic systems and multilingual safety. The initiative could strengthen Singapore’s position as a regional AI hub, yet its impact will depend on how effectively these tools are governed in sensitive areas such as healthcare, education and public-sector services.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!

NASA develops AI system to track harmful algal blooms using satellite data

NASA researchers have developed an AI system designed to combine satellite datasets to improve monitoring of harmful algal blooms.

The system uses self-supervised machine learning to analyse patterns across five satellite missions and instruments, helping researchers identify blooms in regions including western Florida and Southern California. According to researchers, the approach could support environmental monitoring and earlier identification of marine health risks.

Harmful algal blooms can affect ecosystems, wildlife, coastal environments, and public health. In parts of Florida, blooms caused by Karenia brevis have disrupted coastal communities for decades, while toxic blooms along the US West Coast have harmed dolphins, sea lions, and other marine species.

NASA researchers said the system combines information from multiple satellite observation technologies. Instruments such as the PACE satellite and the TROPOMI monitoring instrument help identify algae characteristics, including pigment, fluorescence, and biological activity across ocean surfaces.

The researchers said the self-supervised AI model identifies relationships between datasets without relying heavily on manually labelled data. The system was trained using satellite observations collected during 2018 and 2019 before being tested on later bloom events.

Michelle Gierach of NASA’s Jet Propulsion Laboratory said the system could help environmental agencies identify areas for water sampling earlier during bloom development. Researchers said combining satellite observations with field data may improve coordination between scientific and public health teams.

The project team said the system is being expanded using additional coastal and freshwater datasets.

Why does it matter?

NASA’s development highlights growing use of AI and satellite intelligence for environmental monitoring and climate-related risk management. Harmful algal blooms are becoming an increasing concern for coastal economies, fisheries, tourism, biodiversity, and public health systems worldwide.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Australia’s regulator targets AI-nudify platform over child safety and deepfake risks

Australia’s eSafety Commissioner has begun enforcement action against another AI-powered ‘nudify’ service accused of failing to protect children from exposure to sexually explicit deepfake images.

The regulator issued a formal Direction to Comply to one of the most visited nudify services in Australia, giving the provider 14 days to implement stronger protections preventing children from accessing the platform. eSafety said the service allows users to upload images of real people and generate sexually explicit deepfake content on demand.

The regulator warned that such technologies can facilitate non-consensual exploitation, cyberbullying, sexual extortion, image-based sexual abuse, misogynistic harassment and exploitation of minors. The service had attracted nearly 40,000 Australian visits per month as of March 2026, following a sharp increase in traffic over the previous six months.

The enforcement action was taken under Australia’s Age-Restricted Material Codes, which came into force in March 2026. The codes are designed to prevent children from accessing or being exposed to age-restricted material, including pornography, high-impact violence, self-harm, suicide or disordered eating content.

eSafety said the Argentina-based provider failed to respond to earlier engagement after the codes took effect and had not committed to improving protections for children. The regulator chose not to name the service to avoid inadvertently promoting it.

If the service does not meet the requirements within the 14-day timeframe, eSafety may pursue further action, including civil penalties of up to AU$49.5 million and delisting notices to search engine providers that help facilitate access to the site.

The action follows earlier enforcement in late 2025 that led three widely used nudify services, which had reportedly been used to generate child sexual exploitation material in schools, to withdraw from Australia. Those services have since relaunched under new ownership with additional safety measures, including mandatory age assurance.

Why does it matter?

The case shows how online safety regulators are beginning to apply age-assurance and child protection rules directly to generative AI services. Nudify platforms are treated as high-risk because they can enable non-consensual sexualised deepfakes, image-based abuse and exploitation involving minors at scale. Australia’s enforcement approach also signals that regulators may target foreign-based AI services when they are accessible to local users and fail to implement safeguards.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Meta reportedly cuts 8,000 jobs as AI investment and restructuring accelerate globally

Meta is reportedly cutting about 8,000 jobs globally as part of a restructuring aimed at reducing costs while increasing spending on AI infrastructure and products.

According to media reports, the cuts represent about 10% of Meta’s workforce and are intended, in part, to offset the cost of the company’s expanding AI investments. The reductions are expected to affect engineering and product teams in particular, with employees in several regions notified as the restructuring begins.

Reports also indicate that around 7,000 employees are being reassigned to new AI-focused teams, while thousands of open roles have been closed. The restructuring reflects Meta’s effort to redirect resources towards AI products, infrastructure and agent-based tools across its platforms.

In Ireland, reports said around 350 jobs were affected, representing a significant share of Meta’s local workforce. The company has not publicly confirmed all regional figures, but said affected employees and authorities had been notified.

The cuts come as Meta prepares for a major increase in AI-related capital expenditure. Reports say the company expects spending to rise sharply in 2026 as it builds infrastructure for AI models, personalised assistants and other AI-powered features across Facebook, Instagram, WhatsApp and its wider product ecosystem.

Staff concerns have also emerged around the pace of restructuring, internal communication and workplace monitoring linked to AI development. Reports cited employee unease over plans to monitor computer activity as part of AI training practices.

Why does it matter?

Meta’s restructuring shows how major technology companies are reallocating labour and capital around AI. The reported job cuts are not only a cost-saving exercise, but part of a wider shift in which companies are redirecting resources towards AI infrastructure, automation and agentic systems. The development also highlights a growing tension in the tech sector: AI is being presented as a long-term growth engine, while workers face uncertainty over how that transition will reshape roles, teams and investment priorities.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Japan backs blockchain and AI-based financial infrastructure proposal

Japan has approved a policy proposal focused on blockchain technology and AI within future financial infrastructure development. The proposal reflects broader efforts to integrate digital technologies into financial systems and economic operations.

According to the proposal, backed by the ruling Liberal Democratic Party and endorsed by the government’s Policy Council, the initiative envisions expanded use of automated and continuously operating digital financial systems.

The proposal, titled the ‘Next-generation AI & Onchain Finance Concept’, envisions a system that enables 24/7 digital commerce through blockchain networks, including those supporting cryptocurrencies such as Bitcoin. The proposal describes blockchain technology as a potential foundation for future financial infrastructure because of its verification and record-keeping features.

The strategy includes consideration of tokenised financial instruments, including tokenised stocks and yen-denominated stablecoins. The proposal also discusses possible tokenisation models linked to the Bank of Japan’s current account deposits.

The Financial Services Agency has been tasked with developing a five-year roadmap to encourage both public and private sector investment in the initiative. Policymakers said the initiative is intended to support financial innovation and the development of programmable financial services.

Why does it matter?

Japan’s move is a major shift in how a leading economy is attempting to merge traditional monetary systems with blockchain and AI, potentially setting a benchmark for other countries exploring programmable finance and tokenised assets.

It could accelerate competition among jurisdictions to define standards for digital financial infrastructure, influencing how central banks, regulators and markets approach the integration of crypto, tokenisation and automated financial systems.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our chatbot!

Việt Nam highlights AI in national digital transformation strategy

Việt Nam’s Ministry of Science and Technology has highlighted AI as part of the country’s digital transformation and innovation strategy. Officials said AI is being prioritised alongside technologies including big data, cloud computing, blockchain, and the Internet of Things.

The comments were made during a workshop focused on AI products and technology cooperation. Participants said businesses are showing growing interest in AI adoption while facing implementation and investment challenges.

Discussions also addressed data infrastructure, computing capacity, and explainable AI systems for public administration and urban management.

Participants said stronger infrastructure, workforce development, and research support could help expand Việt Nam’s role in the regional AI and digital technology sectors. The workshop took place in Hà Nội, Việt Nam.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Singapore and Google strengthen collaboration on AI innovation and digital governance

Google and Singapore’s Ministry of Digital Development and Information have announced an expanded National AI Partnership designed to accelerate the deployment of frontier AI technologies across the country’s economy and public sector.

The initiative builds on earlier collaboration between Google and Singapore’s digital authorities and aims to support healthcare innovation, scientific research, workforce development, enterprise transformation, and AI governance. Officials said the partnership aligns with Singapore’s National AI Strategy and broader ambitions to position the country as a global AI hub.

A major focus of the collaboration involves healthcare and life sciences. Google DeepMind is exploring AI co-clinician systems with Singapore’s public healthcare sector, examining how AI agents could support doctors and patients throughout medical treatment and decision-making processes.

Google DeepMind will also collaborate with the National Research Foundation to train researchers on agentic AI systems designed to accelerate scientific discovery. Additional partnerships with the Agency for Science, Technology and Research will focus on AI-enabled research and secure cloud-based scientific analysis tools.

The agreement also expands AI deployment in education. Google and Singapore’s Ministry of Education plan to strengthen educator training programmes and integrate AI-powered teaching support tools across schools. Officials said the partnership aims to improve digital learning capabilities while supporting broader AI workforce readiness initiatives.

Singapore and Google additionally announced plans to collaborate on AI safety, governance, and cybersecurity frameworks. A joint initiative involving Cyber Security Agency of Singapore and other agencies is examining how AI agents interact with real-world digital systems and how governance rules should evolve around autonomous AI technologies.

Officials described the partnership as part of a wider effort to deploy frontier AI responsibly while supporting economic growth, public services, and digital transformation.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

MIT researchers explore AI-driven approaches to drug discovery

AI is increasingly being used in drug discovery to analyse large chemical datasets and identify potential therapeutic compounds. Researchers estimate that the number of potentially useful small-molecule compounds is too large for experimental testing alone, increasing reliance on computational screening methods.

Researchers at MIT are developing machine learning models designed to predict molecular behaviour and chemical reaction pathways. The research focuses on identifying promising drug candidates and improving how chemical reactions can be simulated and understood using data-driven methods.

The research incorporates chemical principles such as reaction mechanisms and physical constraints into AI models. The group has developed models including ShEPhERD, which predicts molecular interactions with proteins, and FlowER, which models chemical reaction outcomes.

Research in the group also extends to automated experimentation, structure analysis and experimental design, aiming to build more efficient workflows for drug discovery. According to the researchers, the broader aim is to improve the realism and accuracy of computational predictions in chemistry.

Why does it matter?

AI-driven chemistry significantly reduces the time and cost required to identify viable drug candidates by narrowing down vast chemical search spaces that would otherwise be impossible to evaluate experimentally.

Embedding chemical principles into machine learning models also improves reliability, making computational predictions more useful for real-world pharmaceutical development and potentially accelerating the delivery of new treatments.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our chatbot!