Accenture Selected by Anthropic as First Embedded Evaluator for AI Development Deceleration Initiative
Anthropic selects Accenture to assist in decelerating AI development. The collaboration is non-exclusive, with Anthropic planning to reveal additional evaluators in the coming weeks.
Accenture has been selected by Anthropic as the company's inaugural embedded evaluator, a move designed to support the deceleration of AI development as outlined in a proposal CEO Dario Amodei presented last week.
On Sept. 12, Amodei released a proposal consisting of three steps aimed at decelerating AI development, enabling the implementation of necessary safeguards in response to concerns about potentially catastrophic consequences resulting from rapid, unregulated advancement.
The proposal received favorable responses from OpenAI CEO Sam Altman and SpaceX CEO Elon Musk, though Nvidia CEO Jensen Huang expressed opposition, contending that such regulatory measures were unnecessary.
Nevertheless, in his proposal, Amodei stated that "AI has been advancing drastically faster, driven primarily by AI's growing ability to build the next generation of AI. This dynamic is called recursive self-improvement," and "left unchecked, it could outrun our ability to understand and control these systems."
According to Amodei's proposal, the initial step involves establishing independent evaluators who possess employee-like access, a commitment to which Anthropic had already pledged unilaterally, as Amodei documented.
On Friday, Anthropic revealed progress toward honoring this commitment through its partnership with Accenture and its AI division Faculty for "evaluating and red-teaming models, conducting alignment assessments and testing model safeguards."
The specific mechanisms for implementing this approach remain under development, given that embedded evaluation represents a novel concept, according to the company's statement.
According to the announcement, both Anthropic and Accenture anticipate investing a minimum of $1 billion in this initiative throughout the next five years.
Anthropic will fund work directly amid urgency
Anthropic pointed out that no established mechanism currently exists for financing independent evaluation, suggesting that long-term funding ought to originate from pooled or governmental resources, although considering the pressing nature of this work, Anthropic will provide direct funding for Accenture's efforts.
The collaboration is non-exclusive in nature, with Anthropic anticipating the announcement of additional evaluators in the upcoming weeks with whom the company will similarly collaborate.
According to Friday's announcement, Accenture's Faculty possesses expertise in testing and evaluating models for several of the world's premier AI laboratories and developing sophisticated AI systems that are designed to be safe and ethical from the ground up.
"Embedded evaluation is an emerging area, and we look forward to partnering with Anthropic to help accelerate the development of embedded evaluators, which we see as an important part of the safety landscape going forward."
Julie Sweet, chair and CEO of Accenture