In a strategic move to address the growing concerns surrounding the risks of artificial intelligence, OpenAI, the developer behind ChatGPT, has issued comprehensive guidelines for users. The initiative comes in the wake of CEO Sam Altman’s return to helm the Microsoft-owned organization.
OpenAI’s latest guidelines, outlined in a document termed “Preparedness Framework,” shed light on the critical aspect of assessing “catastrophic risks” associated with the development of AI models. The document emphasizes that current scientific studies on these risks fall short, necessitating a proactive approach to bridge the existing gap.
“We believe the scientific study of catastrophic risks from AI has fallen far short of where we need to be,” stated OpenAI in the guidelines, highlighting the need for a robust framework to address this deficiency.
The framework introduces a Monitoring and Evaluations Team, established in October, dedicated to scrutinizing “frontier models” with capabilities surpassing even the most advanced AI software. This team’s primary objective is to thoroughly evaluate the risks posed by emerging AI technologies.
To systematically evaluate these risks, the team will categorize them into four main areas. Models will be assessed based on their cybersecurity implications, potential to contribute to harmful activities like creating chemicals, organisms, or weapons, persuasive abilities influencing human behavior, and the degree of autonomy, including the potential to surpass control mechanisms set by their creators.
Models will be assigned risk levels ranging from “low” to “critical” in each category. Crucially, only models with a risk score of “medium” or below will be considered deployable, according to the framework.
The outcomes of these risk assessments will be channeled to OpenAI’s Safety Advisory Group. This group, in turn, will provide recommendations to CEO Sam Altman or another designated individual on the board, ensuring a comprehensive and vigilant approach to the development and deployment of AI technologies. OpenAI’s move underscores the organization’s commitment to responsible AI development and the proactive management of potential risks associated with advancing technology.