OpenAI
OpenAI is an AI research and deployment company founded in 2015. Originally a nonprofit, it transitioned to a "capped-profit" structure in 2019. OpenAI created ChatGPT and the GPT series of language models, pioneering many techniques in RLHF and alignment.
Overview
OpenAI's stated mission is to ensure artificial general intelligence benefits all of humanity. The company has been central to developing and popularizing large language models, and many current AI safety researchers began their careers there, including the founders of Anthropic.
Safety Research
Superalignment Team
OpenAI established a Superalignment team to work on aligning superintelligent AI systems, led initially by Jan Leike and Ilya Sutskever. The team focused on scalable oversight and eliciting latent knowledge.
Red Teaming
OpenAI pioneered adversarial testing and red teaming for AI systems, developing frameworks for identifying potential misuse and harmful capabilities before deployment.
Key Contributions
- GPT series (GPT-2, GPT-3, GPT-4, GPT-4o)
- ChatGPT - widely adopted conversational AI
- InstructGPT paper on RLHF
- DALL-E image generation
- Codex and GitHub Copilot
Notable Alumni
- Dario Amodei (now at Anthropic)
- Paul Christiano (now at ARC)
- Jan Leike (now at Anthropic)
- Chris Olah (now at Anthropic)
See Also
External Sources
- 🔗OpenAI
- 🔗OpenAI Research
- 🔗OpenAI Charter