Dr. T's AI brief
dtau...@gmail.com
AI Tool Promises New Era of Earth Intelligence from Space
An open-source AI foundation model developed by a team led by researchers at the University of Cambridge in the U.K. transforms vast archives of satellite imagery into easy-to-use data for environmental analysis. Using imagery from European Space Agency Sentinel satellites, the Tessera model learns patterns over time, fills gaps caused by cloud cover, and generates compact digital representations of Earth’s surface. The resulting GeoTessera database allows researchers to study environmental changes using ordinary computers rather than requiring supercomputers.
[ » Read full article ]
University of Cambridge Department of Computer Science and Technology (U.K.); Constantino Panagopulos (June 4, 2026)
Tech Leaders Sign Letter to Prevent AI-Developed Biological Weapons
Google DeepMind’s Demis Hassabis, OpenAI’s Sam Altman, Anthropic’s Dario Amodei, and Microsoft AI’s Mustafa Suleyman are among the signatories of a letter urging the U.S. Congress to require synthetic DNA and RNA providers to screen customers and orders for potential misuse. The signatories warn that rapidly advancing AI could lower the knowledge barriers that have historically prevented bad actors from developing biological weapons.
[ » Read full article ]
Wired; Emily Mullin (June 3, 2026)
Anthropic Files Preliminary IPO Paperwork
Anthropic has confidentially filed preliminary paperwork for an initial public offering (IPO) with the U.S. Securities and Exchange Commission, the first step toward becoming a publicly traded company. The company said the offering remains subject to regulatory review, market conditions, and other factors, with share numbers and pricing not yet determined.
[ » Read full article ]
NPR; John Ruwitch (June 1, 2026)
U.N. Calculates Nation-Sized Environmental Footprints for AI, Datacenters
A United Nations University report warns that datacenters already consume enormous resources and could double their environmental impact by 2030. In 2025, global datacenters used 448 trillion watt-hours of electricity—more than all but 10 countries—producing carbon emissions comparable to Argentina and consuming about 1.2 trillion gallons of water. By 2030, datacenters could account for nearly 3% of global electricity use, with AI responsible for 40% of demand.
[ » Read full article ]
Associated Press; Seth Borenstein (June 3, 2026)
Florida AG Sues OpenAI, Seeking to Hold CEO Altman Personally Liable for Alleged Harms
Florida Attorney General James Uthmeier has filed a lawsuit against OpenAI and its CEO Sam Altman, alleging ChatGPT was knowingly released despite risks it could cause harm. The complaint claims the chatbot has contributed to violence, suicide, addiction among minors, and reduced critical thinking, and seeks to hold Altman personally liable while requiring the company to comply with Florida consumer protection laws.
[ » Read full article ]
CNBC; Ashley Capoot (June 2, 2026)
Nvidia Launches ‘Superchip’ Putting AI Power into Laptops, PCs
Nvidia has launched a "superchip" for laptops and desktop PCs that, the company says, can run AI agents locally instead of relying on cloud computing. Developed with Taiwan's MediaTek and capable of being integrated into Windows devices, the RTX Spark chip could allow AI assistants to navigate computers autonomously, reducing reliance on traditional mouse and keyboard inputs. Nvidia CEO Jensen Huang described it as a reinvention of the PC for the AI era.
[ » Read full article ]
The Guardian (U.K.); Julia Kollewe (June 1, 2026)
Researchers Demonstrate AI Worm Could Target Any Online Device
Researchers at Canada’s University of Toronto (U of T) demonstrated an AI-powered computer worm that can adapt its attack strategy as it spreads across networks, potentially targeting any Internet-connected device. Built using freely available open-weight AI models, the prototype can analyze each device, exploit known vulnerabilities, gather information, and use compromised machines’ computing power to fuel further attacks. U of T’s Nicolas Papernot said his lab is working to develop countermeasures.
[ » Read full article ]
U of T News (Canada); Adina Bresge (June 2, 2026)
Leiden Declaration: AI Is Challenging the Core Values of Mathematics
The Leiden Declaration on Artificial Intelligence and Mathematics, released by a group of 16 researchers from 15 universities and endorsed by the International Mathematical Union, warns that AI is challenging core values of mathematics. The Declaration highlights risks such as AI-generated proofs containing hidden errors, inadequate credit for original work, dependence on costly proprietary systems, exaggerated claims about AI capabilities, and commercial influence over research priorities. Rather than rejecting AI, the declaration calls for its responsible use within clear community norms.
[ » Read full article ]
Leiden University (The Netherlands) (June 2, 2026)
Trump AI Executive Order Asks Companies for Early Access
U.S. President Donald Trump signed an executive order requesting AI companies voluntarily provide advanced models to the federal government up to 30 days before public release for security evaluation. The order establishes a benchmarking process to assess cyber capabilities and determine whether systems qualify as “covered frontier models,” while explicitly stating it does not create mandatory licensing or pre-approval requirements. The order also directs the U.S. Department of Defense to strengthen cybersecurity efforts.
[ » Read full article ]
CNBC; Ashley Capoot (June 2, 2026)
Japan to Join U.S. 'Genesis Mission'
Japan will become the first international partner in the U.S. government’s Genesis Mission, an AI-driven research initiative to accelerate scientific discovery and technological innovation. Aiming to gain an advantage in the technological supremacy race with China, Japan and the U.S. each will contribute $500 million toward the Mission over the next five years. The project will combine AI, supercomputing, and vast scientific datasets to speed research across 26 fields, including semiconductors, critical minerals, quantum technology, nuclear fusion, and biotechnology.
[ » Read full article ]
The Japan News; Keiichi Nakane; Hirotaka Kuriyama (June 1, 2026)
The Internet Is Being Rebuilt for Machines
Tech companies are redesigning Internet infrastructure for AI agents, whose unpredictable, machine-driven traffic differs sharply from traditional human Web activity. AWS launched a new version of OpenSearch Serverless that instantly scales computing resources up or down for agentic workloads, allowing enterprises to handle bursts of AI activity without paying for idle infrastructure. Industry leaders including Cloudflare, Microsoft, Databricks, and Snowflake are making similar changes as machine-generated traffic rapidly increases.
[ » Read full article ]
TechCrunch; Rebecca Bellan (May 28, 2026)
Erin Brockovich Targets AI Industry with New Datacenter Map
Activist Erin Brockovich has launched a crowdsourced mapping project aimed at tracking and highlighting community concerns about the rapid expansion of AI datacenters across the U.S. The initiative documents where major hyperscale AI facilities are being built, proposed, or already operating, while also collecting public reports on local impacts such as water use, energy demand, and environmental or health concerns. Brockovich’s map has already gathered thousands of submissions, with Texas accounting for the largest share.
[ » Read full article ]
Gizmodo; Bruce Gil (May 26, 2026)
U.S. Tech Layoffs Hit Two-Year Monthly High in May
U.S. tech companies announced 38,242 job cuts in May, the highest monthly total in nearly two years and more than any other sector, according to data published by outplacement firm Challenger, Gray & Christmas. The layoffs pushed the tech industry’s 2026 total to over 123,000, even as major firms simultaneously expanded hiring plans and sharply increased AI-related capital spending. AI was the most frequently cited reason for workforce reductions for a third straight month.
[ » Read full article ]
Tom's Hardware; Luke James (June 4, 2026)
How to Fight AI Brain Rot at School? For One Country, It’s with Free ChatGPT
Estonia has launched a nationwide experiment to address AI’s impact on learning by providing free ChatGPT accounts to nearly 20,000 high school students. Rather than banning AI, the country is integrating it into education through a customized “Socratic” chatbot that guides students with questions rather than completing assignments for them. Estonia hopes responsible AI use can enhance education without undermining independent thinking.
[ » Read full article *May Require Paid Registration ]
The Wall Street Journal; Sam Schechner (June 1, 2026)
Researchers Develop AI Framework For Brain Fluid Mapping
News Medical (5/28, Paharia) reported that researchers have developed a “new artificial intelligence (AI)-based imaging framework to map fluid movement in the brain in unprecedented detail using magnetic resonance imaging (MRI) scans.” The magnetic resonance artificial intelligence velocimetry (MR-AIV) framework “also provides model-inferred information on the pressure generated during fluid movement and how readily the fluid passes through different brain regions.” In this study, researchers “developed MR-AIV, an AI system designed to reconstruct 3D brain fluid flow” from dynamic contrast-enhanced MRI (DCE-MRI) data. The team built “four specialized neural networks” that estimate tracer movement, fluid permeability, pressure variation, and noise. The system was validated with synthetic datasets and in vivo DCE-MRI scans from five healthy mice. The study notes that the brain “regulates fluid movement mainly by how easily fluid can pass through different structures, rather than by strong pressure differences.”
Neuroscience News (5/27) reported the researchers from the University of Rochester, Brown University, and the University of Copenhagen used the AI system to target the “hidden mechanics of the glymphatic system, the fluid network that washes away metabolic wastes like the amyloid-beta proteins linked to Alzheimer’s disease.” The AI models revealed a “dual-speed drainage blueprint, demonstrating that protective fluid moves 50 times faster across the brain’s outer surfaces than it does when trickling through deep brain tissue.” This discovery was made by training neural networks on MRI videos of dye tracking, allowing the deduction of precise fluid flow velocities. The authors of the study, which is led by URochester professor Douglas Kelley, “hope to be able to compare the fluid flow in healthy and sick brains as well as young and old brains, with aspirations to eventually study circulation in humans.”
Siemens Develops AI Data Center Architecture
Data Centre News (6/1, Peck) reports that Siemens, in collaboration with NVIDIA and Fluence, has developed a reference architecture for AI data centers. This design, which features a 136 MW facility with a 100 MW IT load, incorporates Fluence’s battery energy storage technology to enhance operational flexibility and resilience. Jeff Monday, Chief Growth Officer at Fluence, stated, “Our Smartstack platform is central to this new architecture, transforming the grid into an accelerator for compute.” The architecture aims to support high-density AI deployments and maintain compatibility with future IT platforms.
House Committee To Hold Frontier AI Security Hearing
Inside Cybersecurity (6/1, Beard) reports behind a paywall that the House Homeland Security cyber subcommittee is “holding a hearing this week to examine how frontier artificial intelligence models are shaping the cybersecurity threat landscape, while House Armed Services is set to mark up their version of the fiscal 2027 National Defense Authorization Act.” The AI-focused hearing on Thursday “will explore consider the role of frontier AI models, agentic AI systems and AI-enabled coding tools in US cyber defense and critical infrastructure resilience.”
New Maryland Law Sets Guidelines For AI Use In Education
CBS News Baltimore (6/1, Thompson) reports that the A.I. Ready Schools Act “mandates the State Department of Education to issue artificial intelligence guidelines and best practices to help local public school systems.” This law will “incorporate A.I. literacy into workforce preparation standards and computer science standards for kindergarten through grade 12 by June 1, 2027; provide professional development in A.I. to educators and school leaders; and staff and support the new Maryland AI Education Collaborative on Artificial Intelligence in K-12 Education.” Local school districts are tasked with creating policies, acquiring AI tools, and appointing an AI coordinator. The article also highlights the Future Think Edge summer program at City Springs Elementary in Baltimore, Maryland, which featured a new AI software that “caters to the personal needs of every student to teach them subjects like math, science and coding in a game-like setting,” helping them “solve problems and think critically” and maximizing their learning potential.
Nvidia, TSMC Expand AI Partnership To Boost Chip Manufacturing
Quartz (6/1, Cabili) reported that Nvidia and TSMC are integrating Nvidia’s AI tools and accelerated computing across TSMC’s chip fabrication facilities to enhance yield and reduce defects. Nvidia founder and CEO Jensen Huang said, “TSMC is bringing Nvidia AI and accelerated computing into the fab itself, tackling some of the world’s most complex design and manufacturing challenges.” The companies highlighted that cuLitho provides a “performance advantage at 20% to 50% over CPU-based workflows,” while cuEST accelerates chemistry simulations “up to 50 times faster.” TSMC CEO C.C. Wei emphasized the goal is to “strengthen TSMC’s manufacturing capabilities in support of its customers’ future products.” The collaboration, announced at Nvidia GTC Taipei, leverages Nvidia’s cuML for process control and the Metropolis platform for enhanced defect detection. Additionally, TSMC is exploring Nvidia’s Omniverse libraries to create a virtual environment, FabTwin, for simulating chip factory layouts.
AI Frameworks Enhance Tumor Immunity Insights
LabMedica (5/29) reports researchers at The University of Texas MD Anderson Cancer Center have developed AI frameworks to create a pan-cancer atlas that characterizes tertiary lymphoid structures (TLSs) within tumors, offering a composite score linked to patient outcomes. This approach, detailed in Science, surpasses traditional TLS measures by integrating TLS maturation, spatial context, and composition, and shows promise for improving prognosis and treatment response prediction across various cancers, pending validation in clinical trials.
Workers Say They Turn To AI For Advice Because It’s Less Judgmental Than Colleagues
HR Dive (6/2, Ewen) reports Workday’s recent research highlights a growing “connection deficit” in workplaces, with 33% of employees primarily engaging in task-related conversations and 16% attributing a decrease in patience for small talk to AI. While AI is praised for reducing stress and burnout, with 62% of employees reporting decreased stress levels, Workday warns that reliance on AI for advice and companionship could erode essential human interactions that foster trust and resilience. The study also reveals a significant concern among 43% of employees about reduced human interaction due to AI, prompting Workday to recommend intentional design in collaboration and mentorship to maintain workplace social connections.
Human Skills Protect Jobs From AI Automation
Forbes (6/4) reported that a Click Finder study identifies human skills that are resistant to automation, emphasizing their importance in the face of AI’s growing influence in the job market. The study analyzed 84 occupations using O*NET data, highlighting skills like crisis intervention, complex case diagnosis, and patient relationship cultivation as having low automation risk. For instance, crisis intervention and emergency decision-making, prevalent in roles like paramedics and firefighters, carry an average automation risk of just 9.8%. Rebecca Hubbard, Technical Director at Click Finder, noted that these skills thrive in unpredictable environments, unlike AI, which excels in controlled settings. The report underscores that interpersonal skills, such as empathy and communication, are crucial as they appear across more occupations than technical skills, offering a competitive edge against automation.
dtau...@gmail.com
Anthropic Releases ‘Safe’ Version of Claude Mythos AI Model to Public
Anthropic has released Fable 5, a public version of its advanced Mythos AI model, while keeping the full-powered Mythos 5 model restricted to approved organizations due to cybersecurity concerns. Fable 5 is designed for tasks such as coding, research, and image analysis, but sensitive requests involving cybersecurity, biology, or chemistry are automatically routed to a lower-tier, less-capable model.
[ » Read full article ]
The Guardian (U.K.) (June 9, 2026)
OpenAI announced that it has filed a confidential S-1 for a potential initial public offering (IPO), though the company emphasized that no timeline has been set for such an offering. The filing comes one week after Anthropic submitted its own IPO paperwork, highlighting the rapid evolution of the AI sector. OpenAI was valued at roughly $852 billion following a major funding round in March.
[ » Read full article ]
NBC News; Steve Kopack (June 8, 2026)
ACM Technology Policy Council TechBrief Addresses Agentic AI Systems
According to a new TechBrief from ACM’s Technology Policy Council, AI systems increasingly browse the Web, execute code, manage files, and send messages without step-by-step human approval, raising new risks. The TechBrief identifies policy dimensions where current government frameworks fall short. Said Simson Garfinkel, chair of the ACM TechBriefs Committee, “When something goes wrong with a system that takes actions on a user’s behalf, it can be genuinely difficult to determine who is responsible. Existing law simply doesn’t answer this question.”
[ » Read full article ]
ACM Media Center (June 11, 2026)
CISA Rewrites Federal Patching Requirements for AI Threat Era
The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has updated its federal vulnerability management directive to prioritize patching based on risk, requiring the most critical vulnerabilities to be fixed within three days. The updated framework uses factors such as exploitability, Internet exposure, automation risk, and potential system impact to classify severity. The directive is designed to address rising concerns about AI-enabled cyber threats.
[ » Read full article ]
Dark Reading; Jai Vijayan (June 10, 2026)
Humanoid Robot to Run Convenience Store in Hong Kong
Hong Kong will launch its first 24/7 convenience store operated by a humanoid robot, as part of a broader effort to promote public adoption of AI. The robot-run capsule store, developed by a mainland Chinese embodied-AI company, will provide multilingual customer service and sell a variety of products, serving as a real-world demonstration of embodied AI technology.
[ » Read full article ]
Interesting Engineering; Atharva Gosavi (June 8, 2026)
U.K. Commits £750M to National AI Supercomputer in £1.1B Hardware Strategy
The U.K. unveiled a £1.1 billion (U.S.$1.48 billion) AI Hardware Plan to strengthen domestic AI and semiconductor capabilities, including £750 million (U.S.$1 billion) for a national AI supercomputer expected by 2030. The strategy supports British chip companies through funding, procurement opportunities, and a new AI Hardware Innovation Program, while expanding skills training for engineers and chip designers.
[ » Read full article ]
HPCwire (June 8, 2026)
Amazon Unveils Warehouse Robot as Tech Giants Continue Layoffs
Amazon has unveiled an AI-powered version of its warehouse robot Proteus that can understand conversational language and follow natural-language instructions from workers. The robot is part of Amazon’s broader automation push, and its introduction comes as Amazon and other tech companies continue workforce reductions attributed to AI-driven efficiencies.
[ » Read full article ]
CNBC; Sawdah Bhaimiya (June 5, 2026)
Tech Jobs Grew in May Despite AI Layoffs
The U.S. tech job market continued to grow in May despite a surge in AI-related layoffs. Employers added 69,000 technology jobs, according to a CompTIA review of U.S Bureau of Labor Statistics data, marking a second straight month of gains, while unemployment among tech workers fell to 3.1% from 3.5% in April. Hiring was strongest in cloud infrastructure, IT services, cybersecurity, and software development.
[ » Read full article ]
CIO Dive; Roberto Torres (June 5, 2026)
Spelman College Names AI Pioneer Howard its Next President
Spelman College has named AI and robotics pioneer Ayanna Howard, an ACM Athena Lecturer Award laureate, its next president. A co-founder of Zyrobotics, which develops educational and therapeutic technologies for children with special needs, and Black in Robotics, an organization focused on expanding representation in the robotics field, Howard is recognized for advancing AI, robotics, and inclusive technology. Spelman leaders said her vision will help prepare students for a rapidly evolving world.
[ » Read full article ]
CBS News; Christopher Harris (June 5, 2026)
Bot Web Traffic Overtakes Human Web Traffic
According to data from Internet hosting service provider Cloudflare, automated bots and AI agents now generate more Web traffic than humans, accounting for 57.4% of Internet requests. Cloudflare CEO Matthew Prince said the shift happened much sooner than expected, driven by rapid growth in AI agents that can autonomously browse thousands of websites while performing tasks.
[ » Read full article ]
NBC News; Samantha Elkins (June 4, 2026)
Canada Bids to Lead Middle Powers in AI Sovereignty Race
Canadian Prime Minister Mark Carney unveiled the “AI for All” strategy, aiming to make Canada a leader among middle powers developing sovereign AI capabilities. The plan treats AI as critical infrastructure, seeks to reduce dependence on U.S. tech firms, and promotes partnerships with allies in Europe and the Indo-Pacific. Key priorities include expanding Canadian-owned datacenters, cloud infrastructure, semiconductor production, and a national supercomputer. Carney stressed that AI must serve the common good and protect human dignity.
[ » Read full article ]
Politico; Mickey Djuric (June 4, 2026)
Team Builds First Silicon Spintronic Chip for Smart Computers
A team of researchers from Japan's Tohoku University and the U.S. National Institute of Standards and Technology developed a silicon-based spintronic probabilistic bit (p-bit), a key building block for probabilistic computing. Unlike conventional computer bits that are fixed at 0 or 1, p-bits randomly fluctuate between states, making them well-suited for AI, machine learning, and optimization problems. Built using standard semiconductor manufacturing processes, the chip combines silicon transistors with spintronic devices that exploit electron spin to generate controlled randomness.
[ » Read full article ]
Interesting Engineering; Georgina Jedikovska (June 2, 2026)
AI degree programs are rapidly expanding across U.S. colleges and universities. According to Northeastern University researchers, the number of AI majors has grown from just five programs in 2021 to at least 74 majors and 89 minors today, with more schools preparing to launch related programs this year. Many of the programs overlap heavily with computer science curricula, raising questions about whether some programs represent substantive innovation or simply rebranding.
[ » Read full article *May Require Paid Registration ]
The New York Times; Alan Blinder (June 8, 2026)
Phoenix Is a Datacenter Mecca, and a Test Case for AI Power
Phoenix has become a major hub for AI-driven datacenters, triggering a dispute over who should pay for the massive power infrastructure needed to support them. The state's largest utility, Arizona Public Service, has proposed a 45% rate increase for large energy users such as datacenters and a 14.5% increase for residential customers, arguing that “growth should pay for growth.” Tech companies say they already fund many grid upgrades and want more flexibility to build their own power generation.
[ » Read full article *May Require Paid Registration ]
The Wall Street Journal; Jennifer Hiller (June 4, 2026)
Martin Scorsese Is Embracing AI
Martin Scorsese has publicly backed Black Forest Labs, an AI start-up that specializes in image generation. The acclaimed filmmaker joined the company in 2025 as a partner and adviser and said he has used its technology during pre-production to create and share storyboards more efficiently. Scorsese said AI can support creativity and help filmmakers communicate visual ideas without replacing artistic craftsmanship.
[ » Read full article *May Require Paid Registration ]
The New York Times; Brooks Barnes (June 2, 2026)
NIST Expands Goals for Renamed AI Consortium
The U.S. National Institute of Standards and Technology (NIST) has rebranded its AI Safety Institute Consortium as the “Artificial Intelligence Consortium,” expanding its mission beyond safety to focus on AI measurement, innovation, and adoption. The change reflects a policy shift under the White House, with the group now focusing on developing standards, evaluation tools, and “measurement science” to assess AI systems across sectors.
[ » Read full article *May Require Paid Registration ]
Federal News Network; Justin Doubleday (June 1, 2026)
Google Trains Teachers In AI Utilization
NBC News (6/9) reported that Google hosted a two-day AI training program for 70 K-12 educators and school technology directors at its Mountain View campus on June 2, 2026. The training, part of Google’s Educator Series, emphasized the integration of AI tools like Gemini and NotebookLM in classrooms to enhance teaching efficiency and address educators’ “pain points.” With growing public skepticism towards AI, the training aimed to equip teachers to advocate for AI’s benefits in education. Jennie Magiera, Google’s global head of education impact, highlighted the necessity of AI in future classrooms, noting its potential to save time in lesson planning and grading. Despite resistance from some parents and educators, Google continues to promote AI as a tool for personalized learning and adaptive content. The training also introduced free online modules for educators, with plans for monthly updates starting September. Educators expressed optimism about the training, though acknowledged challenges in promoting AI adoption in schools.
Survey Highlights Gap In AI Code Understanding Among Junior Engineers
Insider (6/11) reported that BairesDev released its Q2 2026 Dev Barometer, revealing insights into software developers’ readiness in the AI era. The survey, conducted with 1,569 developers across 77 countries, highlighted a significant gap in AI-generated code comprehension between junior and senior engineers. Only 16% of senior developers believe junior engineers fully understand AI-generated code, while 57% think they understand it “to some extent.” This gap underscores concerns about the future development of technical leaders. Nacho De Marco, CEO of BairesDev, emphasized the need for juniors to bridge the gap between generating and understanding code to ensure future leadership in the field. An independent analysis by Professor Francisco Anello supported these findings, stressing the importance of critical thinking and deep code comprehension over AI tool proficiency. The survey also found that 85% of juniors believe AI tools have enhanced their development understanding, yet 70% of seniors prioritize real-world experience as a job-readiness indicator. The report calls for evolving hiring and upskilling strategies to focus on practical experience and robust technical fundamentals in AI-assisted development.
dtau...@gmail.com
Anthropic Suspends New AI Models After U.S. Directive
The U.S. government ordered Anthropic to restrict foreign nationals from accessing its newest AI models, Fable 5 and Mythos 5, due to national security concerns about a potential method for bypassing certain safeguards. To ensure compliance, Anthropic temporarily disabled access to both models for all users. The company maintains the action stems from “a misunderstanding” and hopes to restore access soon.
[ » Read full article ]
NBC News; Jared Perlo (June 12, 2026)
Tool Detects Hallucinations in Machine Vision Models
A tool called the Prelim Attention Score (PAS), developed by Los Alamos National Laboratory researchers, can detect when vision-language AI models are “hallucinating,” or generating outputs not grounded in an input image. Most commonly used vision-language models generate new tokens, or words, partly by relying on the words they have already produced. PAS analyzes how much a model relies on its own previously generated text versus actual visual data while producing each token in a response.
[ » Read full article ]
Interesting Engineering; Atharva Gosavi (June 15, 2026)
Timing Trick Cuts Energy Used in LLM Training
Researchers at the University of Twente in the Netherlands demonstrated a way to reduce the energy used to train large language models by up to 14% without meaningfully affecting performance. Their approach uses dynamic voltage-frequency scaling, which adjusts GPU clock speeds in real time based on the specific computation being performed. The researchers optimized settings at the level of individual GPU kernels, allowing the GPU to consume less power while maintaining nearly the same training speed, with only a 0.6% slowdown observed.
[ » Read full article ]
IEEE Spectrum; Dina Genkina (June 10, 2026)
AI Helps Track the World’s Shrinking Glaciers
Researchers at the Friedrich-Alexander University of Erlangen–Nuremberg in Germany developed an AI-based method that improves the monitoring of glacier retreat using satellite imagery. By combining a deep learning model with just one manually labeled image per glacier, seasonal reference images, and maps of underlying rock formations, the team reduced boundary-detection errors from more than 1 kilometer to less than 70 meters—roughly comparable to the accuracy of human analysts.
[ » Read full article ]
IEEE Spectrum; Edd Gent (June 10, 2026)
People Around the World See China as the AI Superpower
A global survey by U.K.-based research firm Public First of more than 18,000 respondents in 15 countries found that while Americans and people in a few Asian countries see the U.S. as the dominant AI leader, majorities in most surveyed nations—including Canada, France, and the U.K.—now view China as leading the field. The findings also reveal growing skepticism toward AI, especially in the U.S. Young Americans have become particularly pessimistic, citing concerns about misinformation, deepfakes, job displacement, and the broader social impact of AI.
[ » Read full article ]
Politico; Owen Dahlkamp (June 15, 2026)
AI Deepfakes Getting Harder to Spot in the Midterms
Political campaigns involved in the 2026 U.S. midterm elections increasingly are using AI-generated images, videos, and deepfakes to promote candidates and attack opponents. Examples include altered videos of candidates singing songs, appearing as superheroes, or making controversial statements. Although many states have laws restricting election-related deepfakes, enforcement remains difficult due to free speech concerns.
[ » Read full article *May Require Paid Registration ]
The Wall Street Journal; Amrith Ramkumar (June 14, 2026)
AI Agents Collaborate To Enhance Humor Generation
Tech Xplore (6/10) reported that a research initiative led by Shiwei Hong, a Ph.D. student at George Mason University, is exploring how artificial intelligence (AI) can improve humor generation through a collaborative workshopping approach. The study, presented at the 64th Annual Meeting of the Association for Computational Linguistics, involves a multi-agent system where 35 AI agents simulate a comedy-writing environment. These agents, powered by GPT-4, are assigned distinct roles as performers, critics, or audience members, allowing them to generate, critique, and refine humorous content collectively. The research found that scripts produced in a discussion-based environment were preferred by human evaluators in over 70% of cases, indicating the potential of structured interaction to enhance AI creativity. Hong and her adviser, Assistant Professor Zhicong Lu, plan to incorporate human participants and advanced technologies such as virtual and augmented reality to further refine the system. This work suggests that the future of AI may involve communities of agents learning and creating together.
The College Of New Jersey And Mercer County Technical Schools Initiate AI Pathway
ROI-NJ (NJ) (6/15) reported that The College of New Jersey and Mercer County Technical Schools have initiated a dual enrollment partnership to provide high school students with a pathway into the fields of artificial intelligence and robotics. This initiative is supported by the New Jersey Department of Education’s Expanding Career Pathways in Artificial Intelligence & Robotics Grant, marking one of the first comprehensive AI and robotics career pathways in the state. Mercer County Technical Schools is one of only two districts selected for this grant. Michael Bernstein, president of The College of New Jersey, emphasized the partnership’s role in expanding access to college-level learning in AI and robotics, thereby enhancing workforce readiness and inspiring students to envision careers as innovators and technology leaders. The grant encourages county vocational-technical school districts to collaborate with four-year colleges and industry experts to develop rigorous career and technical education pathways. The initiative also aims to enhance generative AI literacy and foster collaboration among K–12 educators, higher education faculty, and industry professionals to ensure curriculum relevance. Research indicates that students in dual enrollment programs are more likely to enroll in college and complete degree programs, benefiting from exposure to college expectations and support systems within their high school environment.
Loop Engineering Enhances AI Capabilities
Forbes (6/17) reported that loop engineering is an emerging trend in AI that involves instructing AI to perform iterative loops until a specified condition is met. This contrasts with the traditional step-by-step interaction with AI, offering a more autonomous approach. The technique is particularly useful for agentic AI, allowing AI agents to continuously perform tasks like finding better deals on hotel bookings without constant human intervention. However, the article cautions that poorly designed loops can lead to excessive computing costs and undesirable outcomes, such as booking undesirable hotels. Loop engineering requires careful planning, including setting clear goals, establishing assessment mechanisms, and incorporating human feedback checkpoints. Loop engineering is described as a blend of prompt and workflow-centric approaches, highlighting its potential to enhance AI’s utility in tasks requiring ongoing monitoring and adjustments. Despite its benefits, the complexity of loop engineering necessitates attention to detail to prevent AI from going astray. The article suggests that as AI technology advances, loop engineering will become more prevalent in both agentic and conventional generative AI applications, offering new opportunities for efficiency and effectiveness in AI-driven tasks.