Anthropic's proposed AI watchdog METR has deep ties to Effective Altruism (nypost.com)

Anthropic has proposed the establishment of an independent oversight group called Model Evaluation and Threat Research (METR), aimed at enhancing AI safety through embedded third-party evaluators. This initiative, highlighted in a recent essay by CEO Dario Amodei, seeks to ensure that advanced AI models are rigorously monitored to prevent potential catastrophic outcomes. The proposal is significant as it represents a proactive step by a leading AI firm to address safety concerns amidst rising skepticism about the risks associated with AI technologies. However, METR's connections to the Effective Altruism (EA) movement raise concerns about potential conflicts of interest. Critics point out that many of METR's founders and staff have deep ties to EA, which promotes using wealth for altruistic causes, often leading to accusations of bias. Despite assurances of independence and funding safeguards, skepticism remains regarding the group’s ability to objectively evaluate AI firms, especially given its financial backing from EA proponents. The initiative underscores a broader industry debate on self-regulation versus external oversight, as industry leaders like Sam Altman of OpenAI express support for independent evaluators while voicing concerns over the potential for AI regulation to be influenced by vested interests.