Top 10 AI News
Guided learning lets “untrainable” neural networks realize their potential
MIT News – December 18, 2025
Researchers at MIT’s CSAIL have developed a guidance method that enables even “untrainable” neural networks to learn effectively. By using one network’s built-in biases to steer another network’s training, previously untrainable models can now achieve successful learning outcomes. This technique could expand the potential of limited or unusual neural network architectures.
A new way to increase the capabilities of large language models
MIT News – December 17, 2025
Researchers at the MIT-IBM Watson AI Lab have designed a new neural architecture that enhances state tracking and sequential reasoning in large language models over long texts. This expressive architecture improves how LLMs manage context and logic, potentially boosting the performance of AI systems in tasks like extended dialogue or complex document understanding.
A “scientific sandbox” lets researchers explore the evolution of vision systems
MIT News – December 17, 2025
MIT researchers have created an AI-powered simulation environment that allows scientists to study how visual systems evolve. By simulating virtual organisms and their vision sensors, scientists can explore new camera and sensor designs inspired by biological evolution. This “scientific sandbox” could inform the development of better visual systems for robots and autonomous vehicles.
“Robot, make me a chair”
MIT News – December 16, 2025
An AI-driven design tool developed at MIT lets users create and build multicomponent objects simply by describing them in natural language. For example, a user can tell the system to “make me a chair,” and the AI will generate a design composed of interlocking parts that can be fabricated and assembled. This approach simplifies the design process for custom furniture and objects.
3 Questions: Using computation to study the world’s best single-celled chemists
MIT News – December 15, 2025
Assistant Professor Yunha Hwang of MIT uses machine learning and microbial genomics to study methane-producing microbes, some of nature’s most efficient “single-celled chemists.” By applying computational tools to biology, Hwang’s work seeks to uncover principles of these organisms’ chemistry and metabolism. Her research exemplifies the intersection of AI and genetics at MIT.
Deep-learning model predicts how fruit flies form, cell by cell
MIT News – December 15, 2025
MIT scientists have developed a deep-learning model that predicts the cell-level development of fruit fly embryos. The model can simulate how individual cells move and divide as the embryo forms. This breakthrough could be extended to study more complex tissues and organs, potentially helping researchers identify early signals of developmental disorders or diseases.
Enabling small language models to solve complex reasoning tasks
MIT News – December 12, 2025
Researchers at MIT have introduced a “self-steering” system called DisCIPL that coordinates multiple small AI models to collaborate on complicated tasks. For example, small models can work together on itinerary planning and budgeting problems by following constraints provided by the system. This method demonstrates how smaller, specialized AI agents can team up to tackle problems that typically require large models.
New MIT program to train military leaders for the AI age
MIT News – December 12, 2025
MIT has launched a new certificate program designed for U.S. military officers, particularly naval leaders, to develop skills in artificial intelligence and related technologies. The goal is to equip military decision-makers with AI knowledge to address national security challenges. This initiative reflects the growing integration of AI education in national defense training.
New method improves the reliability of statistical estimations
MIT News – December 12, 2025
MIT engineers have developed a statistical technique that helps researchers determine when to trust the results of experiments. This method can show whether observed results might be due to random error or genuine effects, improving confidence in studies across fields like economics and public health. The approach provides clearer guidance on the reliability of experimental findings.
New materials could boost the energy efficiency of microelectronics
MIT News – December 11, 2025
Researchers demonstrated a new approach to reduce energy waste in computer chips by stacking multiple active components made from novel materials on the chip’s back end. This vertical stacking design allows more computations in the same area, cutting down on wasted power. It points toward future microchips that use significantly less energy.
MIT researchers “speak objects into existence” using AI and robotics
MIT News – December 5, 2025
An MIT team developed a “speech-to-reality” platform that turns spoken descriptions into physical objects. The system uses a 3D generative AI model to design an object from speech and then employs a robotic assembly line to build it from modular parts. Users can simply describe an object, and the system will create the object on demand.
Robots that spare warehouse workers the heavy lifting
MIT News – December 5, 2025
Pickle Robot Company, founded by MIT alumni, has created autonomous robotic machines for warehouse environments. These robots can load and unload trucks by themselves using suction systems, reducing the need for manual heavy lifting by workers. The technology is being deployed in logistics centers to improve efficiency and reduce physical strain on people.
Deepfake cyberbullying poses a growing problem for schools
The Associated Press – December 22, 2025
Schools across the U.S. are facing incidents of cyberbullying enabled by deepfake technology. In one case, AI was used to produce a fake explicit video of a middle school girl, leading to abuse on social media. Educators and experts warn that as deepfake tools become widely accessible, schools must prepare strategies to address this new form of harassment and protect students.
Boys at her school shared AI-generated, nude images of her. After a fight, she was the one expelled.
The Associated Press – December 22, 2025
In Louisiana, a middle-school girl was expelled after defending herself in a fight, even though she was a victim of deepfake pornography. Boys at school had used AI to create fake nude images of her and circulated them online. The incident has sparked debate about how schools should handle new issues of AI-driven harassment and the policies needed to protect students.
Starlink in the crosshairs: How Russia could attack Elon Musk’s conquering of space
The Associated Press – December 22, 2025
As SpaceX’s Starlink network expands global internet coverage via satellites, defense analysts warn about the risk of adversaries targeting these assets. The report examines possible strategies that nations like Russia could use to disrupt or disable Starlink satellites and ground infrastructure, potentially affecting communications worldwide. The threats include electronic jamming, cyber attacks, and even kinetic actions against satellites.
US stocks rose again in 2025 after overcoming turbulence from tariffs and Trump’s fight with the Fed
The Associated Press – December 23, 2025
The U.S. stock market ended 2025 on a high note, finishing the year with gains despite multiple episodes of volatility. Investor fears about trade tariffs, rising interest rates, and a supposed AI tech bubble were offset by strong underlying economic data and corporate profits. The year saw several historic one-day losses on worries about President Trump’s policies and AI hype, but overall 2025 turned out to be profitable for the market.
Trump signs executive order to block state AI regulations
The Associated Press – December 11, 2025
President Donald Trump has issued an executive order preventing individual U.S. states from enacting their own regulations on artificial intelligence. The order directs federal agencies to pre-empt conflicting state AI laws. The move comes amid a national debate over how to balance innovation and oversight in the rapidly evolving AI sector, and it sets up a confrontation between federal and state authorities on tech policy.
AI Minted More Than 50 New Billionaires In 2025
By Alicia Park, Forbes – December 25, 2025
The boom in artificial intelligence startups and investments this year led to a boom in wealth creation. Over 50 founders and executives in the AI industry saw their net worths climb to billionaire levels in 2025, according to a Forbes analysis. The surge in stock prices and funding for major AI companies translated into unprecedented gains for many entrepreneurs in the field.
As OpenAI Shifts To For-Profit, Its Foundation Controls $130 Billion. Who Benefits?
By Paulo Carvão, Forbes – December 22, 2025
OpenAI’s recent transition to a for-profit model has left its original nonprofit foundation in control of roughly $130 billion in assets. This piece examines how that wealth is planned to be used and who stands to gain. The analysis raises questions about governance and the distribution of the AI windfall, as well as how the funds might shape the future direction of AI research and benefit society.
AI Startup Lovable Just Minted One Of Europe’s Youngest Ever Self-Made Billionaires
By Alicia Park, Forbes – December 19, 2025
Lovable, a Singapore-based generative AI startup that creates personalized avatar companions, recently raised expansion funding at a valuation that turned its young co-founder Zalinder Alexander into one of Europe’s youngest self-made billionaires. This article highlights how Lovable’s user-friendly AI and venture backing fuelled this meteoric rise and discusses the broader wave of youth-led AI companies gaining high valuations.
Amazon’s Robotaxi Unit Launches In San Francisco Without Steering Wheels—Or Fees
By Alan Ohnsman, Forbes – November 18, 2025
Amazon’s self-driving taxi subsidiary revealed its first vehicles on the streets of San Francisco, offering free rides in electric vehicles that lack steering wheels, mirrors, or pedals. The co-founder of Zoox, which Amazon acquired, says regulatory approval for a paid service may be near. The article discusses Zoox’s strategy and the future of robotaxis in urban areas.
Microsoft Confronts Agent Sprawl With Agent 365
By Janakiram MSV, Forbes – November 19, 2025
Microsoft is rolling out “Agent 365,” a new suite of tools designed to manage the growing number of AI agents and assistants within organizations. As businesses deploy more specialized AI bots for different tasks, Agent 365 aims to provide oversight and integration so that corporate networks can handle “agent sprawl” effectively. The article explains how this system works and why Microsoft sees it as necessary.
Gemini 3 Approaches The Uber-Software Point: AI As The New UI
By John Koetsier, Forbes – November 18, 2025
Google’s latest AI model, Gemini 3, is presented as an interface shift where users interact more naturally with AI assistants, akin to the old concept of “Uber for software.” Rather than navigating complex menus, people can use plain language to command applications. The article analyzes how AI-based user interfaces are evolving and what this means for software design in the coming years.
Stop Wasting Money On AI Agents: 5 Rules For Choosing The Right Use Cases
By Bernard Marr, Forbes – December 24, 2025
While AI agents promise to automate many tasks, not all use cases benefit equally from them. In this article, AI expert Bernard Marr outlines five guidelines for selecting the right applications for AI agents. He emphasizes focusing on routine, well-defined tasks with high return on investment, rather than forcing AI onto every process. The rules help businesses avoid investment pitfalls and make AI adoption more effective.
The Robot Olympics Will Have Zero Sports. Here’s Why
By John Koetsier, Forbes – December 24, 2025
Upcoming robot competitions will focus on practical tasks rather than athletic sports. This visionary piece argues that instead of mimicking human sports like basketball or sprinting, robot Olympics should include events that reflect real-world applications—such as assembly challenges, search-and-rescue simulations, or precision coordination tasks. The shift underscores how robotics is becoming integrated into everyday problem solving.
ChatGPT’s App Store Moment Has Arrived
By Ron Schmelzer, Forbes – December 23, 2025
OpenAI has transformed ChatGPT from a standalone conversational AI into a platform by introducing an “App Store” for plugins and extensions. This article explains how this change lets developers and companies distribute specialized AI tools through ChatGPT, akin to how mobile apps work on a phone. It discusses the implications for competition, innovation, and how users can now perform a wide range of tasks within the ChatGPT ecosystem.