Researchers have tested a deepfake detector that uses light to process several videos at the same time. The laboratory results are striking, although the technology is a screening experiment rather
Read articleResearchers have tested a deepfake detector that uses light to process several videos at the same time. The laboratory results are striking, although the technology is a screening experiment rather than a tool that can
Read article
New AI browsing assistants can work inside websites where you are already signed in. That could save time
What if a map could combine flood forecasts, satellite pictures and population data to show where help is
An AI assistant improved the documents patent lawyers produced, but a study found lasting gains in independent judgement
People were shown the same responsive behaviour from humans and AI, yet were less willing to call the
Ask a chatbot to invent a story and a familiar lighthouse keeper may turn up again. Researchers found striking repetition across thousands of AI stories, raising questions about what originality means when machines borrow familiar
Read article
A new compact humanoid costs $2,490 and can walk or roll, but its invitation-only first release is aimed
Anthropic says it wants to dedicate a stake in the company to a public benefit fund as AI
Samsung has revealed how it presents AI appliances differently around the world, putting cooking convenience first in America
Google has introduced a Gemini hub for small businesses, promising help with branding, spreadsheets and connected business apps.
Researchers have used X-ray scans and machine learning to read writing inside a Roman-era scroll that was too fragile to open. The achievement concerns the surviving portion of one scroll, with other manuscripts still requiring
Read articleMIT researchers have shown an AI-controlled robot assembling simple furniture from reusable parts after receiving an ordinary-language request.
BirdNET uses recordings of bird calls to suggest which species may be nearby, even when the bird is
Researchers using satellite images and machine learning found that most industrial fishing vessels in their global study were
Automated recruitment tools can rank applicants long before a human offers an interview. Regulators have found problems with
Researchers have built headphones that can focus on the person you choose to listen to while reducing competing voices. It is a laboratory prototype rather than an everyday feature already available in all wireless earbuds.
Read article
Microsoft is warning that a company’s own AI agents can create security risks when they are given broad
Government science advisers have warned that Britain could become dependent on powerful foreign AI suppliers without more domestic
Amazon has joined a plan to explore building twelve advanced nuclear reactors at a former coal site in
Google is making links to websites easier to find inside its AI-generated search answers. The updates could help
MIT researchers have presented a system that pairs the three-dimensional shape of a crystal with descriptions of its properties. The result offers a more useful way to search for promising materials, while stopping well short
Read article
A new OECD policy brief looks at two places AI gathers personal information: conversations people hand over to
Researchers have given robots an artificial skin able to detect fine contact and react to potentially harmful pressure.
Amazon has unveiled two Bee devices that can turn conversations into reminders and shopping suggestions. One sits on
A team including MIT researchers has improved forecasts for the difficult stretch between the daily weather outlook and
Apple says its newer AI features can use powerful remote servers without leaving personal requests readable to Apple. The important question is not whether a cloud is involved, but what protections surround the data sent
Read articleSome websites now offer a selfie instead of an identity document to check whether someone is old enough.
A camera that claims to know whether you feel stressed, bored or pleased at work sounds futuristic. In
An American lender can use a complicated algorithm to assess your application, but it cannot use that complexity
AI can write a glowing product review without ever trying the product. American regulators have banned several kinds
OpenAI is offering eligible US university laboratories sponsored access to its specialist science model for up to a year. The plan could lower software costs for researchers, but access is vetted and the announcement is
Read article
Eight US states filed separate lawsuits accusing OpenAI of misleading consumers about ChatGPT's safety, with particular concern for
Meta has released a powerful open-weight AI model, and Apple has shown how it can run across Mac
Apple says its new folding iPhone can keep Siri visible next to another app and use on-screen information
A Cornell-linked startup is developing machines that assess apples at harvest and prepare potatoes for catering kitchens. Its
MIT researchers have tested a way for robots to plan their next movement while completing the present one. The approach sped up several laboratory tasks, offering a possible route towards less hesitant robotic assistants rather
Read articleA controlled experiment found that more conversational chatbots made false information seem more credible. Giving users a way
A classroom tool in India turns textbook material into lesson plans in English and Kannada, with experienced teachers
Researchers examined 250,000 anonymised Claude conversations and found that people frequently challenge, refine and question AI responses. The
A Banque de France study finds permanent hiring has slowed more for young people in occupations exposed to
A study of UK university policies found many were difficult to locate and argued that rules presented as educational often centred on monitoring and discipline. The findings reflect the researcher’s analysis, not a verdict on
Read articleA UK student survey finds that generative AI has become routine in higher education, while many students want
Google has demonstrated ways its Gemini 4 Argon model can reason about physical spaces and help with robot
A US workforce survey found many employees admit overstating their experience with AI. The results point to a
An OpenAI analysis suggests smaller European businesses are increasingly using tools that carry out tasks rather than simply
Researchers tested how sure leading AI models sound compared with how often they are right. They found a persistent gap, especially on harder questions, which helps explain why a polished answer is not always a
Read article
Hesai has launched a camera-and-cloud system intended to capture real places as detailed 3D training material for robots.
A new Gallup survey finds most US school educators used AI last year, and many users say it
Anthropic researchers say small AI models could eventually spread themselves between poorly protected connected devices. Their experiments reveal
Singapore and Google DeepMind have developed AI that looks for manipulation within scam conversations directly on a phone.