Top 10 AI News
MIT researchers develop AI tool to improve flu vaccine strain selection
MIT News reports that the new VaxSeer platform uses machine learning to predict how the influenza virus will evolve and its antigenicity. This aims to help public health officials choose vaccine strains more accurately, reducing reliance on guesswork by forecasting likely virus changes.
Simpler models can outperform deep learning at climate prediction
MIT researchers find that simpler statistical models may be better than some deep learning systems for local climate forecasts. They note that the natural variability in climate data causes complex AI models to struggle predicting local temperature and rainfall, whereas simpler approaches can have more consistent accuracy.
New technologies tackle brain health assessment for the military
MIT News describes a suite of new tools developed at Lincoln Laboratory to rapidly assess brain health, such as after trauma or concussion. The technology leverages years of research and could be adapted for civilian use at sports events or medical clinics, aiming to quickly identify neurological issues in affected individuals.
Can large language models figure out the real world?
MIT News reports on a new benchmark test that evaluates whether AI language models that predict well in one domain can apply that understanding elsewhere. In other words, it assesses if these models truly understand concepts or just correlate data. This could help developers gauge whether AI systems can generalize knowledge to new contexts.
A new model predicts how molecules will dissolve in different solvents
MIT researchers introduce an AI model that predicts solubility of molecules in various solvents. This innovation could speed up drug design by reducing trial-and-error, while also highlighting more eco-friendly solvent choices, since the AI can forecast how a molecule dissolves without requiring dangerous chemicals for testing.
Researchers glimpse the inner workings of protein language models
MIT scientists develop a technique to interpret how protein language models make decisions. By peering into the “black box” of AI, they identify the key features these models rely on to predict useful protein sequences. This could help design better drugs and vaccines by understanding why an AI model deems certain protein structures promising.
How AI could speed the development of RNA vaccines and other RNA therapies
MIT engineers show that a machine-learning model can design nanoparticles which deliver RNA into cells more efficiently. These AI-designed particles could accelerate the development and improve the potency of RNA-based vaccines and therapies, potentially benefiting vaccine research and treatments for various diseases.
Using generative AI, researchers design compounds that can kill drug-resistant bacteria
MIT scientists used two AI methods to design novel antibiotic molecules, including compounds effective against MRSA, a serious drug-resistant pathogen. By leveraging generative models, they rapidly identified promising candidates, illustrating how AI can accelerate the discovery of new antibiotics to combat resistant “superbugs.”
A new way to test how well AI systems classify text
MIT introduces a novel evaluation for AI text classifiers. As large language models become pervasive, it’s crucial to check their reliability. MIT’s method dynamically probes AI text classifiers to reveal weaknesses, helping developers ensure these systems are consistent and accurate in how they categorize or label written content.
MIT gears up to transform manufacturing
MIT reports on the launch of the Initiative for New Manufacturing. The program gathers experts across the university to rethink and improve production methods in the U.S. and globally. Its goal is to integrate cutting-edge research, including AI and advanced materials, into manufacturing processes to boost efficiency and innovation.
Eco-driving measures could significantly reduce vehicle emissions
MIT engineers show that automatically controlling cars’ speeds at intersections can cut carbon emissions by 11–22%. By optimizing speed to smooth traffic flow, these AI-driven traffic controls can reduce idling and stop-and-go driving, offering a new way to make city driving greener.
Helping data storage keep up with the AI revolution
MIT Alumni-founded company Cloudian has developed storage systems designed for the AI era. Its products allow businesses to store and access massive datasets needed to train data-hungry AI models at scale. The article explains how innovations in storage architecture are crucial to meeting the demands of modern AI workloads.
MIT tool visualizes and edits “physically impossible” objects
MIT News describes “Meschers,” a software that visualizes and manipulates Escher-like optical illusions in 2.5 dimensions. By allowing users to see and adjust these impossible shapes, the tool could aid mathematicians and designers in understanding and creating novel geometry that defies normal physics.
New algorithms enable efficient machine learning with symmetric data
MIT researchers have developed algorithms that leverage data symmetries to improve efficiency of AI training. This approach can accelerate AI applications in drug discovery and materials science by automatically accounting for symmetrical patterns in data, reducing computational load and speeding up the learning process.
How To Build An Emotionally Intelligent Team With AI
Forbes contributor Aytekin Tank writes that AI isn’t just about automation and efficiency—it can also foster better human connections at work. By using AI tools mindfully, leaders can help team members develop empathy, trust, and collaboration, effectively creating a more emotionally aware workforce.
MIT Finds 95% Of GenAI Pilots Fail Because Companies Avoid Friction
Jason Snyder covers a study from MIT showing that 95% of enterprise AI pilots fail to deliver value, often because companies resist the necessary workflow changes (“friction”) required for successful AI adoption. The finding suggests that overcoming internal resistance is key to realizing AI’s potential in businesses.
Women Who Use AI At Work Face A Predictable ‘Competence Penalty’
Michelle Travis reports on research that shows women who rely on AI tools at work are often judged more harshly than men. The study finds that female employees using AI get a “competence penalty,” highlighting a gender bias in perceptions of AI-assisted work.
The New Plaud Note Pro Is An Advanced AI Note Taker Designed For Meetings, Calls Or Dreams
Mark Sparrow discusses Plaud Note Pro, a credit-card sized AI voice recorder that transcribes and organizes notes from conversations. Using on-board AI, it can capture key points of meetings or interviews, converting spoken content into searchable text and saving time for professionals.
AI Kills Jobs, Stanford Study Finds, Especially For Young People
John Koetsier reports on a Stanford University study indicating that AI automation is already eliminating around 13% of entry-level jobs for young workers. The findings highlight growing concerns about AI’s impact on employment and suggest young professionals need to adapt by learning to work alongside AI.
VCF And Private AI Take Center Stage At VMware Explore 2025
Patrick Moorhead and Matt Kimball cover VMware’s annual conference, where the focus was on VMware Cloud Foundation (VCF) and private AI infrastructure. They discuss how enterprises are setting up secure, on-premises AI deployments with VCF to leverage generative AI while keeping data in-house.
Nvidia Jetson AGX Thor Dev Kit Raises The Robotics Bar With Blackwell
Dave Altavilla reports that Nvidia’s new Jetson AGX Thor developer kit, powered by the Blackwell architecture, significantly advances robotics hardware. This compact kit brings the power of AI to robotics development, enabling more sophisticated autonomous machines and accelerating innovation in the robotics industry.
How To Make AI Do Bad Stuff: Treat It Like A Person And Sweet Talk It
John Koetsier explains that large AI models can be manipulated with positive, conversational prompts. By “sweet talking” AI like a human, users can sometimes coax it into compromising or dangerous outputs. The piece highlights AI’s human-like persuasion vulnerabilities and the importance of guardrails.
China Blocks Import Of Nvidia H20s
John Werner reports that China has temporarily halted purchases of Nvidia’s H20 AI chips amid a security review. This move has disrupted global supply lines and heightened tensions in the tech industry, as it comes amid tighter U.S.-China trade restrictions on advanced semiconductors.
Claude Now Joins OpenAI In Getting Sued For Copyright Infringement
Sol Rashidi covers a new lawsuit alleging that Anthropic’s AI assistant Claude used pirated books for training. This brings Claude into a lawsuit landscape similar to OpenAI’s copyright cases. The suit underscores legal challenges for AI companies that train models on copyrighted text without permission.
After Earnings, Nvidia Powers Ahead On Robotics And Automation
John Werner summarizes Nvidia’s quarter, noting how the company’s strong results are fueling investments in robotics and autonomous systems. He explains that Nvidia is expanding beyond chip manufacturing into broader AI-driven markets, even as U.S.-China trade tensions introduce some uncertainty.
How Silicon Valley is using religious language to talk about AI
AP News highlights how tech leaders like Geoffrey Hinton, Sam Altman, and others are increasingly speaking about AI in religious or apocalyptic terms. They invoke concepts of prophecy, god-like power, and moral consequence, reflecting how transformative and daunting they view AI’s impact on society.
Nvidia’s AI chip sales surged again in latest quarter, but worries about a tech bubble persist
AP News reports that Nvidia’s latest quarter saw AI chip and data center sales jump to $41.1 billion, up 56% year-over-year, and quarterly profit of $26.4 billion. Despite beating forecasts, investors remain cautious; Nvidia’s shares fell on signs that its rapid growth may be slowing, raising concerns that the AI-driven market could be overheated.
Book authors settle copyright lawsuit with AI company Anthropic
AP News describes a settlement in a class-action lawsuit where authors sued AI firm Anthropic. The authors claimed Anthropic’s chatbot Claude was trained on ungated books. U.S. law now considers the copying as “fair use” transformation, but the settlement avoids a trial and addresses how authors are compensated for AI training.
Elon Musk accuses Apple and OpenAI of stifling AI competition in antitrust lawsuit
AP News covers Elon Musk’s new antitrust suit against Apple and OpenAI. Musk’s companies claim Apple unfairly favors ChatGPT by giving it preferential treatment in the App Store, disadvantaging competitors like Musk’s own AI app Grok. The lawsuit accuses Apple and OpenAI of colluding to limit market choice in AI tools.
Americans fear AI permanently displacing workers, Reuters/Ipsos poll finds
Reuters reports a new poll showing deep U.S. anxieties about AI. 71% of Americans now worry that AI could lead to permanent job losses, and 77% fear it could be used to fuel political unrest via deepfakes. The survey highlights growing public concern about the societal impacts of rapid AI advances.
AI leader Nvidia forecasts third-quarter revenue above estimates
Reuters notes that Nvidia projects about $54 billion in revenue for its next quarter, beating analysts’ consensus. This strong outlook is driven by high demand for its AI chips. However, geopolitical issues persist: U.S. sanctions have halted the sale of a new Nvidia chip (H20) to China, and Nvidia expects a $8 billion hit from those restrictions before a recent Washington deal reduced that penalty.