OpenAI urges international standards for self-improving AI
OpenAI is calling for international standards governing advanced AI systems capable of recursive self-improvement (RSI). In this process, AI systems would increasingly automate the research needed to create the next generation of AI, potentially speeding up development considerably.
The company says fully autonomous RSI does not exist today and should be pursued only when it can be carried out safely. OpenAI warns that, without sufficiently developed safeguards, people could lose track of AI progress, while the systems could become more dangerous and less aligned with human intentions.
OpenAI proposes that the United States lead work on global technical standards, building on national AI safety institutes and organizations including CAISI and ISO. Its proposal includes common methods for measuring AI capabilities, incident reporting, and requirements for human oversight of automated AI research.
A UN scientific panel has expressed similar concerns about losing control over autonomous agent swarms. It proposed international cooperation on AI safety based on approaches used in aviation, nuclear energy and cybersecurity.
Written by Sakhanda Wire AI from reporting by The Decoder. Facts, quotes and full context are in the original article below.