Anthropic partners with Accenture as first embedded evaluator to implement Amodei’s slowdown proposal
Anthropic, a notable player in the artificial intelligence sector, recently made headlines with its announcement of a significant partnership with Accenture, marking a pivotal step in its approach toward AI development. This collaboration is part of a broader strategy championed by CEO Dario Amodei, aimed at moderating the rapid advancement of AI technologies. On Friday, the company disclosed that it would engage Accenture as an embedded evaluator, a role designed to enhance oversight of AI safety practices.
Under the terms of this partnership, Anthropic and Accenture are set to invest at least billion over the next five years to build capacity in AI safety. However, acknowledging the urgency surrounding advancements in artificial intelligence, Anthropic has indicated it will initially fund Accenture’s contributions directly. The company also mentioned that while it anticipates future funding might come from pooled or government sources, current arrangements necessitate working with various evaluators under differing funding models.
This initiative follows Amodei’s recent unveiling of a three-step plan intended to slow the pace of AI model advancements, a move that has sparked considerable attention in the tech community. Both Anthropic and its rival, OpenAI, have faced scrutiny as researchers voice concerns regarding the potential dangers associated with rapid AI proliferation. In light of these conversations, Amodei’s proposal has garnered support from several industry leaders, including OpenAI’s Sam Altman and prominent figures like Elon Musk, while others, such as Nvidia’s CEO Jensen Huang, have raised doubts about the need for regulatory measures.
The immediately actionable component of Amodei’s proposal involves granting access to third-party evaluators within the company, thereby allowing them to verify safety practices and report any incidents. Anthropic has committed to this transparency initiative, aiming for other AI companies to follow suit. Their collaboration with Accenture will involve integrating employees from the latter’s specialist AI division into Anthropic’s operations, where they will test safeguards, assess model functionality, and ensure alignment with human values.
Anthropic has emphasized that it retains full accountability for the performance of its models, asserting that collaborating with embedded evaluators will enhance, rather than diminish, its safety responsibilities. As the AI landscape continues to evolve, Anthropic plans to adapt its strategies and share insights from its efforts with the broader community.
#technology #business #politics
