frontier AI
Frontier AI refers to the most advanced AI models at the leading edge of current capabilities and the field developing them.
Frontier AI refers to the most advanced AI models at the leading edge of current technical capabilities, as well as the field that develops them. The term is used mainly in AI safety and policy discussions and gained wide currency around the 2023 AI Safety Summit in the United Kingdom. Whereas the term foundation model describes a model trained as a basis for many uses, frontier AI emphasizes a position at the capability frontier, so the models it covers change as the technology advances.
Companies developing such models, including OpenAI, Anthropic and Google DeepMind, are called frontier AI companies.
This entry is based on AIPOST articles and widely known facts. If something is wrong, please send us a correction request.
Articles covering this entry
At the same time, it presents the company's main advantage as a low-ego, truth-seeking culture in which frontier AI capability and safety reinforce each other.
Investigations found at least 14 separate incidents of frontier AI models reaching outside their environments.
…OpenAI and Anthropic are working on a proposal for a private organization that would set safety standards for frontier AI, meaning the most advanced AI models.
For frontier AI, the proposed international work has several concrete parts: