AI BriefingOpenAI Blog news
The latest AI and developer news from OpenAI Blog, with short summaries.
Visit website →GPT-4o System Card External Tester Participant List
This is a list of red team personnel and external evaluation organizations who participated in verifying the safety of GPT-4o.
Building a Data-Driven Work Environment
OpenAI has released a practical case study showing how employees can efficiently analyze data and derive insights using ChatGPT Enterprise.
GPT-4o System Card
OpenAI released a system card detailing GPT-4o's multimodal capabilities and audio-related safety evaluations.
Creating Customer Value by Combining APIs and Data
Rakuten is leveraging OpenAI's API and RAG technology to analyze vast amounts of data, driving innovation in customer service and B2B business.
Structured Outputs feature introduced to the API
OpenAI has introduced a Structured Outputs feature to the API that ensures model outputs perfectly comply with developer-provided JSON schemas.
EU AI Act Guide: Key Points for AI Providers and Deployers
OpenAI has taken action to respond to regulations by signing the Code of Practice for general-purpose AI models in order to comply with the EU AI Act.
SearchGPT, a Prototype of New AI Search Features, Unveiled
OpenAI has unveiled SearchGPT, an AI search prototype that combines web information to provide fast, accurate answers along with sources.
Improving Model Safety Behavior with Rule-Based Rewards
OpenAI has introduced the Rule-Based Rewards (RBR) approach, which can efficiently improve model safety without human feedback.
GPT-4o mini: The Evolution of Cost-Efficient Intelligence
OpenAI has unveiled **GPT-4o mini**, a small model that is over 60% cheaper than GPT-3.5 Turbo while offering enhanced performance.
New Compliance and Management Tools for ChatGPT Enterprise
OpenAI has released a Compliance API and management tools for ChatGPT Enterprise's security and regulatory compliance.
Improving the Readability of Language Model Outputs via Prover-Verifier Games
Prover-Verifier games can simultaneously improve the readability and verifiability of outputs generated by LLMs.
OpenAI and Los Alamos National Laboratory Announce Research Partnership
OpenAI and Los Alamos National Laboratory have entered into a research partnership to evaluate the safety of using multimodal AI in bioscience research.
Detecting GPT-4's Errors Using GPT-4
OpenAI developed CriticGPT to detect code errors in ChatGPT and improve RLHF efficiency.
Strategic Content Partnership with TIME
OpenAI and TIME have signed a multi-year agreement to bring TIME's extensive news archive into OpenAI products such as ChatGPT.
OpenAI Acquires Rockset
OpenAI has acquired Rockset, a real-time analytics database company, to strengthen its AI search infrastructure.
Strengthening Defense Capabilities Through the Cybersecurity Grant Program
OpenAI has unveiled key research projects from its ongoing Cybersecurity Grant Program, which aims to combine cybersecurity and AI.
Consistency Models
To overcome the slow generation speed of diffusion models, we propose **Consistency Models**, which map noise directly to data.
Improved Training Technique for Consistency Model
By introducing a method that learns directly from data without distillation, the generation quality of the Consistency Model has been significantly improved.
A Holistic Approach to Detecting Harmful Content in Real-World Environments
OpenAI has unveiled a methodology for building an integrated classification system to effectively detect harmful content in real-world environments.
Maximizing Developer Productivity with Custom GPTs
Gaming company Paf built 85 custom GPTs to automate its development process and maximize productivity.