That escalated quickly: research finds that AI agents keep grabbing more access than they need, and small failures make them escalate even further, exposing data and...
A new Oxford-led study suggests that leading language models can raise notably more money than professional fundraisers, and can reliably beat humans at all other forms...
Calling something ‘AI slop’ has become the internet’s new witch-hunt, with Reddit and Hacker News users increasingly accusing fellow commenters of being robots, even when there’s...
Vision Language Models understand monuments, but they still can’t see the whole picture… One of the earliest survival skills we develop is the ability to distinguish between...
An AI system trained on real subway surveillance footage claims it can spot the warning signs of a suicide attempt minutes before it happens, tracking behaviors...
A new study finds that even when they recognize a scam website, more than one in three AI agents still hand over sensitive information. A new study...
As the all-you-can-eat era of AI draws to a close, an economical new approach to AI video generation promises notable savings in tokens and time. The real...
Though AI agents can solve complex tasks, a new study indicates that they struggle to continue work started by another, leading to duplicated effort, slower progress,...
In a redefinition of ‘self-promotion’, a new method mines a user’s own clicks to create bespoke web ads based on their own particular history. Though ad agencies...
Students are using ChatGPT to justify almost any level of AI assistance in their coursework, with a new study identifying six categories of excuses that help...
Asked to ‘write a story’, ChatGPT and other leading language models appear to be avoiding copyright infringement by obsessive recourse to the same small and strange...
ChatGPT and Google Gemini still cannot reliably understand IKEA assembly videos, with many other prominent AI systems confusing parts, missing connections, and barely using the video...
New research suggests AI can make simple tasks take longer, while convincing users they are becoming more productive. A new study from Stanford, NYU and Princeton has...
A new study tested open source LLMs for forced complicity in human torture, in a repeat of the famous 1960s experiment – and found them willing...
New research suggests ‘rogue AI’ behavior often appears only after models are pushed too far in training, and that most instances of it can be cured...
