Interesting AI news

Google DeepMind researchers propose that truly open-ended AI systems—those that generate endless surprising yet comprehensible discoveries—may be essential for achieving superintelligence, not just bigger models and more data. View source: https://sifted.eu/articles/recursive-superintelligence-500m

New research reveals a troubling paradox: as CEOs gain AI capabilities, they communicate less with boards, forcing companies to reduce board independence and entrench executives.

New research from the University of Chicago reveals how AI agents are moving beyond assistance to complete task takeover, transforming the nature of professional work across industries.

New research reveals AI systems are far more inconsistent than we realize, with models barely outperforming random guessing under strict standards. These findings challenge how we measure and deploy AI.

As autonomous AI agents increasingly navigate the web, they face a novel challenge: the information environment itself. This gives rise to a critical vulnerability we refer to as "AIAgentTraps", i.e. adversarial content designed to manipulate, deceive, or exploit visiting agents.

Morgan Stanley's latest CIO survey reveals the gap between AI hype and actual corporate spending. IT budgets grow modestly while companies navigate the complex reality of AI adoption.

Anthropic conducted the largest qualitative AI study ever, interviewing users across 159 countries. The results challenge our assumptions about human hopes and fears around artificial intelligence. View source: https://www.anthropic.com/features/81k-interviews