OpenAI is seeking to build an "automated AI researcher" that works to make its models safer, but it warns that doing so without sufficient caution could lead to humans losing control over AI development.
The company says it wants to use this automated researcher to address the "alignment problem"—ensuring that AI behaves in accordance with human ethics. It explains that this is becoming "increasingly important" as AI systems grow in capability and autonomy, and take on more of the research and development of AI itself.
These statements come in the context of concerns about "recurrent self-improvement" (RSI), the point at which artificial intelligence becomes capable of training itself and improving its performance without human supervision.
The company and others have hinted that this could lead to a massive acceleration in the pace of development, with OpenAI saying that artificial intelligence "will directly improve people's lives" and "can lower the cost of advanced intelligence so that people around the world can benefit."
The company believes that automated research can also help combat dangerous AI, as an "automated AI researcher can also be an automated AI safety researcher." However, it warns that this development is dangerous and should not happen now or even become safe to do.
She explains: “If iterative self-improvement is carried out without proper care and caution, it could lead to humans losing practical control over the development of artificial intelligence and being unable to oversee search processes they no longer understand. Hence, AI could become more dangerous, less appropriate, and generally more of a threat to people.”
To prevent this from happening, the company is calling for a set of safeguards, most notably new national and international standards for AI safety and safeguards, and new protocols for measuring and reporting incidents that allow for better visibility of risks.
what is your opinion on this news ?do let us know in the comments on tha comment box
if you liked this news article please share on friends & family & don't forget to follow on website & social media
