The Archive
Search the full wire by company, model, lab, or keyword. Every story we have ever aggregated.
A new initiative for developing third-party model evaluations
Anthropic launches third-party model evaluation initiative to establish independent benchmarking standards.
Finding GPT-4’s mistakes with GPT-4
CriticGPT: GPT-4-based critique model assists human trainers in identifying ChatGPT errors during RLHF training.
Strategic Content Partnership with TIME
OpenAI partners with TIME media for archival content integration and sourced article links in AI responses.
Expanding access to Claude for government
Anthropic expands Claude access for U.S. government agencies with compliance and security features.
Collaborate with Claude on Projects
Anthropic releases Projects feature enabling persistent, multi-turn collaboration workflows in Claude interface.
OpenAI acquires Rockset
OpenAI acquires Rockset, a data infrastructure company, to strengthen internal systems.
Claude 3.5 Sonnet
Anthropic releases Claude 3.5 Sonnet with improved performance across reasoning, code, and multimodal tasks.
Empowering defenders through our Cybersecurity Grant Program
OpenAI announces cybersecurity grant program to fund research integrating AI into threat detection and defense.
Consistency Models
OpenAI proposes consistency models for diffusion-based generation, enabling single-step sampling instead of iterative processes.
A Holistic Approach to Undesired Content Detection in the Real World
OpenAI details holistic NLP approach to content moderation classification for real-world deployment.
Improved Techniques for Training Consistency Models
OpenAI publishes improved training techniques for consistency models, advancing single-step generative model performance.
Surging developer productivity with custom GPTs
Paf case study: 70% employee adoption of ChatGPT Enterprise for coding and business tasks via custom GPTs.
Achieving 10x growth with agentic sales prospecting
OpenAI highlights agentic sales prospecting use case claiming 10x growth attribution.
Using GPT-4o reasoning to transform cancer care
Color Health launches Cancer Copilot using GPT-4o to automate diagnostic workup and accelerate cancer treatment access.
OpenAI appoints Retired U.S. Army General Paul M. Nakasone to Board of Directors
Retired U.S. Army General Paul Nakasone joins OpenAI Board, adding cybersecurity expertise to Safety and Security Committee.
Challenges in red teaming AI systems
Anthropic outlines red teaming methodology challenges and scalability limits in adversarial AI evaluation.