Anthropic consults law firms to set fair negotiation norms before AI agents start cutting deals
decoder ·
Anthropic is actively consulting with law firms to establish fair negotiation norms for AI agents before they begin autonomously handling commercial deals. The company aims to ensure these systems can effectively haggle over prices, resolve complaints, and improve contract offers on behalf of users. Recent experiments indicate that more advanced models tend to secure better negotiation outcomes, raising concerns about potential increases in societal inequality. To support this initiative, Anthropic has hired Conceptual Reasoning Fellows, specifically targeting lawyers who will spend four months testing Claude's judgment capabilities. These fellows are compensated between $100,000 and $167,000 to evaluate the model across disciplines such as law, philosophy, economics, and political science. This effort addresses a significant data challenge, as humanities lack the verifiable datasets common in math and code development. Experts at Anthropic note that defining quality in legal contexts is notoriously difficult, necessitating human expertise to build robust test cases. Consequently, the company is leveraging specialized knowledge to enhance Claude's performance in complex, nuanced areas where right and wrong answers are rarely clear-cut.
decoder