OpenAI proposes global AI safety standards — CNBC
OpenAI has published proposals on safety and security in the development of advanced artificial intelligence systems. They focus on research into aligning AI with human values and recursive self-improvement of systems, or RSI, CNBC Top News reports.
International standards for AI
OpenAI called for international cooperation to create standards in the field of advanced artificial intelligence. The company also recommended building on the work of existing AI safety institutes in countries around the world.
According to the developer of ChatGPT, technical standards should target advanced AI models and their developers. It is also proposed to focus on managing the benefits and risks associated with automated AI researchers, including systems with RSI capabilities.
In a blog post, OpenAI said that alignment research must keep pace with advances in AI system capabilities for the transition to be safe. According to the company, this is necessary to ensure systems remain aligned with human values and under human control.
More current news is available on the UA.News Telegram channel Telegram.
Warnings about autonomous self-improvement
Recursive self-improvement is attracting developers' attention because it could make it possible to create foundation models capable of updating themselves without human involvement. At the same time, the development of such technologies raises concerns among some experts: foundation model developers may lose control of the technology or fail to account for undesirable consequences as AI systems become more complex and spread across the internet.
OpenAI stated that fully autonomous RSI does not currently exist and should not be developed until the safety of such a process is ensured. The company added that without proper caution, people could lose practical control over AI development and would be unable to oversee research processes they no longer understand.
Anthropic's position
Last week, OpenAI competitor Anthropic also presented its own approaches to the safe development of advanced AI models. This came amid warnings from industry researchers about the threats artificial intelligence poses to humanity.
Nearly two weeks ago, former Anthropic and OpenAI employee Jacob Coxon announced his resignation and said that these companies are “gambling with our lives.” According to CNBC, his statement sparked a global discussion about AI safety.
