Security Week
New physics-based research suggests large language models could predict when their own answers are about to go wrong — a potential game changer for trust, risk, and security in AI-driven systems.
Media coverage related to the team’s work
Security Week
New physics-based research suggests large language models could predict when their own answers are about to go wrong — a potential game changer for trust, risk, and security in AI-driven systems.
Newswise
Newswise — Health Secretary Robert F. Kennedy Jr.’s push to overhaul the Vaccine Injury Compensation Program is stirring debate over vaccine safety and trust in public health. Experts warn his rhetoric could fuel misinformation and undermine confidence in vaccines at a time when hesitancy is already at historic highs.
Metro UK
‘I wonder how much money OpenAI has lost in electricity costs from people saying “please” and “thank you” to their models.’
Daily Papers AI
Today’s paper: Multispin Physics of AI Tipping Points and Hallucinations
Security Week
The promise of agentic AI is compelling: increased operational speed, increased automation, and lower operational costs. But have we ever paused to seriously ask the question: can we trust this thing?
Beauty Matter
A broad spectrum of “sunscreen truthers” on platforms like TikTok, Instagram, and Facebook have peddled the trope that sunscreen causes cancer for at least the last decade. These types of conspiracy theories have reached a fever pitch on social media since the pandemic. From vegan anti-vaxxers and bro-biohackers to MAHA and QAnon supporters, they all have two things in common: a case of chemophobia and a belief that sunscreen is the enemy.
Security Week
No-one really understands how AI works or when and why it doesn’t. But the application of first-principle physics theory to the working of AI’s Attention mechanism is providing new insights.
Newswise
The Chicago Sun-Times and the Philadelphia Inquirer recently published stories with unidentifiable quotes from fake experts and imaginary book titles created by AI.
Gadget
A new study explores when and why the output of large language models goes awry and becomes a threat.
NeuroEdge
Researchers at George Washington University have developed a groundbreaking mathematical formula that predicts exactly when artificial intelligence systems like ChatGPT will suddenly shift from helpful to harmful responses – a phenomenon they’ve dubbed the “Jekyll-and-Hyde tipping point.” The new research may finally answer why AI sometimes abruptly goes off the rails.