Four leading AI players are uniting to create the Frontier Model Forum, a new industry body aimed at ensuring the “safe and responsible development” of “frontier AI” models. OpenAI, Microsoft, Google, and Anthropic will collaborate to develop technical evaluations, and benchmarks, and advocate for best practices and standards in response to the increasing demands for regulatory oversight.
New Frontier
The forum’s primary objective revolves around what OpenAI refers to as “frontier AI” – highly advanced AI and machine learning models considered to have significant risks to public safety. They emphasize the unique regulatory challenge posed by such models, as their “dangerous capabilities can arise unexpectedly,” making it challenging to prevent potential misuse.
Central to the forum’s mission is the concept of “frontier AI,” a term coined by OpenAI, referring to advanced AI and machine learning models considered to have the potential to pose “severe risks to public safety.” The forum acknowledges that these models present a distinctive regulatory challenge since their “dangerous capabilities can arise unexpectedly,” making it challenging to prevent potential misappropriation.
The new forum has outlined the following self-stated objectives:
i) Advancing AI safety research to foster responsible development of frontier models, minimizing risks and enabling independent, standardized evaluations of capabilities and safety.
ii) Recognizing and promoting best practices for the responsible development and deployment of frontier models, while fostering public understanding of the technology’s nature, capabilities, limitations, and impact.
iii) Engaging in collaborations with policymakers, academics, civil society, and companies to exchange insights on trust and safety risks.
iv) Providing support for the development of applications that tackle society’s significant challenges, including climate change mitigation and adaptation, early cancer detection and prevention, and the fight against cyber threats.
While currently consisting of four members, the Frontier Model Forum welcomes new participants. Organizations seeking membership must actively participate in the development and deployment of frontier AI models while showcasing a steadfast commitment to frontier model safety.
The founding members have outlined initial steps, which include forming an advisory board to guide the forum’s strategy, and creating a charter, governance, and funding structure.
In the upcoming weeks, the companies plan to engage with civil society and governments to seek input on designing the forum and exploring meaningful collaborative opportunities, as stated in their joint statement.
Regulation
The Frontier Model Forum is a reflection of the AI industry’s dedication to addressing safety concerns seriously. However, it also sheds light on Big Tech’s inclination to ward off potential regulations through voluntary initiatives, allowing them to have some say in setting their own rules.
Notably, this news coincides with Europe’s progress towards establishing the first comprehensive AI rulebook, aiming to prioritize safety, privacy, transparency, and anti-discrimination in AI development practices.
Recently, President Biden held discussions with seven AI companies, including the four founding members of the Frontier Model Forum, to reach voluntary agreements on AI safeguards. Critics argue that the commitments made were somewhat ambiguous, while Biden suggested that future regulatory oversight might be necessary.
“Realizing the potential of AI while managing risks will likely necessitate new laws,” as mentioned by President Biden.
