AI leaders say they want safety, but on whose terms?

AI pioneers are calling for heightened safety measures, but their demands come with a demand for influence in setting those standards. Industry titans such as Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, and Google DeepMind co-founder Demis Hassabis are eager to play a pivotal role in fashioning the safety guidelines they desire, according to experts.

Anthony Aguirre, CEO and president of the Future of Life Institute, noted that these leaders have finally acknowledged the impossibility of fully safeguarding AI systems. However, this newfound openness is contingent upon their desire to retain control over the process. AI firms still aim for artificial general intelligence (AGI), which could surpass human intelligence and potentially disrupt job markets and leadership roles.

Despite recognizing the dangers of AGI, these companies remain committed to its development. They propose self-regulation, transparency, accountability, and assurances against antitrust law as the means to achieve safety standards. Pete Furlong of the Center for Humane Technology highlighted that while companies have expressed a need for regulation, their actions often fall short of that support.

They may seek a closed-door system to govern themselves without public input on risks and solutions. Google's Hassabis suggested the formation of a standards body akin to existing public-private partnerships or self-regulatory organizations, modeled on the Financial Industry Regulatory Authority. Anthropic's Amodei backed sensible AI regulation, particularly on transparency and third-party auditing, while also endorsing voluntary efforts for standard-setting.

However, experts caution that mere independent evaluations may not suffice for effective accountability, and transparency efforts should be more extensive.