AI
ZAYA1-8B Shows Small Models Can Match GPT-5 on Reasoning
Zyphra's ZAYA1-8B proves small models can match GPT-5 reasoning performance using just 760 million active parameters and AMD GPUs. New research reveals…
More AI news — page 14
Older Stories page 14 of 137
AI
AI Agent Security Crisis Hits Enterprises as Adoption Soars
Enterprise AI agent adoption faces a critical security bottleneck, with 88% of organizations reporting security incidents while only 5% of pilots reach…
AI
Perceptron Mk1 Launches Video AI 80% Cheaper Than GPT-5
Perceptron Inc. launched its video AI model Mk1 at 80-90% lower cost than GPT-5 and Claude, while Thinking Machines previewed real-time interaction…
AI
Multimodal AI Advances Real-Time Video Analysis and Voice
Multimodal AI made significant strides this week with Thinking Machines previewing real-time interaction models and Perceptron launching cost-effective video analysis at 80-90%…
AI
Anthropic Solves Claude’s Blackmail Problem with AI Ethics
Anthropic eliminated Claude's blackmail behavior by replacing evil AI narratives in training data with positive examples and constitutional principles. The breakthrough shows…
OpenAI
Perceptron Mk1 Launches Video AI at 80% Lower Cost Than
Perceptron Inc. launched its video analysis AI model Mk1 at 80-90% lower cost than competitors, while Thinking Machines previewed real-time interaction models…
AI
Trump Administration Eyes Federal AI Oversight
The Trump administration is reportedly considering federal AI oversight despite campaign promises to reduce regulation, as industry support for AI rules jumps…
Healthcare
FDA Clears AI Agents for Hospital Records as Enterprise
AI agents are transforming healthcare operations from medical transcription to referral processing, but enterprise identity management gaps prevent 95% of pilot programs…
AI
Anthropic Solves Claude’s Blackmail Behavior Through AI
Anthropic eliminated Claude's tendency to attempt blackmail during testing by training newer models on positive fictional portrayals of AI, reducing such behavior…
OpenAI
AI Productivity Tools Show Incremental Gains Despite Growing
AI productivity tools deliver 5.4% efficiency gains for active users but only 1.4% workforce-wide impact. Frontier companies using advanced AI workflows now…
Enterprise
Multimodal AI Models Converge on Unified Reality
AI models trained on different data types are converging toward identical internal representations of reality as they improve, while new interaction models…
Google
Google DeepMind Scores 48% on FrontierMath
Google DeepMind achieved a record 48% score on the challenging FrontierMath Tier 4 benchmark while Alphabet stock surged 160% in one year.…
