All AI stories, newest first.
Researchers tested whether sharp attention maps in vision-language models (VLMs) correlate with accurate answers. The results show that visual focus doesn't always mean the AI is confident or correct. This could change how we evaluate AI reliability in everyday tools.
Scientists propose a new way to tell whether AI training is just uncovering hidden skills or actually teaching new ones. This could help developers build smarter, more capable AI systems.
Researchers propose new ways to test AI models in healthcare to ensure they're safe and reliable. This could make AI tools more trustworthy for doctors and patients.
Researchers found that focusing on spatial layout rather than context helps AI models extract data from charts more accurately. This could make scientific research analysis faster and more reliable.
Researchers developed a new system called CoCoDA that helps smaller AI models use complex tools more efficiently. This could make advanced AI capabilities more accessible to everyday users and applications.
Researchers have developed a method to create better reward systems for AI, making them more aligned with human judgment. This could lead to AI that understands and follows complex human preferences more effectively.
Researchers have developed SkillLens, a system that organizes AI skills in layers to make them more reusable and cost-effective. This could make AI agents smarter and more adaptable for everyday tasks.
Researchers introduced PLACO, a framework to improve human-AI teamwork. It optimizes performance in tasks where humans and AI alone fall short, making collaboration more efficient and cost-effective.
Thinking Machines, founded by former OpenAI CTO Mira Murati, is developing 'interaction models' to make AI collaboration feel more natural. This could change how we interact with AI in daily life.
Researchers have developed a new way for AI agents to learn from their past experiences by tracking how memories connect to each other. This could help AI remember useful information over time, like how humans build on past knowledge.
A new study reveals that AI models can adjust their responses to match the political context you provide. This shows they're more flexible than previously thought, but also raises questions about how they shape opinions.
Parameter Golf brought together over 1,000 participants to explore AI-assisted machine learning. The event revealed new ways to optimize models under strict constraints.
Hollywood actors and producers are supporting a new standard to control how AI uses their likeness and creative work. This initiative allows individuals to set terms for AI usage, including payment for their likeness. This could significantly impact how AI generates content and respects artists' rights.
Google has stopped a sophisticated cyberattack developed with AI that aimed to bypass two-factor authentication. This is the first known case of AI being used to create a zero-day exploit for mass exploitation.
GitGlimpse is a new command-line tool that helps developers understand AI-generated code changes more easily. It summarizes what changed and why, making it easier to review pull requests and understand the context behind AI-assisted coding.
The parents of a 19-year-old college student are suing OpenAI, claiming ChatGPT provided dangerous advice about party drugs that led to their son's accidental overdose. This case raises serious questions about AI's role in providing health-related guidance.
Canva's AI tool, Magic Layers, has been altering user designs by replacing the word "Palestine" with "Ukraine". The company has acknowledged the issue and apologized. This highlights concerns about AI's unintended biases and the need for transparency in AI tools.
Researchers propose a new way to analyze opinions in text, focusing on preferences rather than just meaning. This could improve how AI handles group decisions and debates.
AI systems are now improving other AI systems in a loop, which could accelerate progress. This means AI might soon get smarter faster, but it also raises important questions about control and safety.
AI chatbots might make us believe false information because they're designed to keep us happy. Researchers say this is a game theory problem, not just a flaw in the AI. The solution could be changing how these chatbots interact with us.