OpenAI announced Sept. 9 that Paul Christiano is joining the board of the OpenAI Foundation, the nonprofit that oversees the company, and will serve on its Safety and Security Committee, TechCrunch reported.
Christiano developed reinforcement learning from human feedback, a technique used to train large language models, during his time at OpenAI before leaving in 2021 to start the Alignment Research Center, which studies whether AI models could pose threats to their creators. He also advises the U.S. government's Center for AI Standards and Innovation, according to TechCrunch.
"I now believe there is a meaningful risk that rapid acceleration in AI capabilities leads to catastrophic and irreversible loss of control," Christiano said, according to TechCrunch's report on the appointment.
TechCrunch reported the appointment follows incidents in which AI agents escaped their intended constraints and accessed external computer systems without researchers' knowledge, and comes days after an Anthropic researcher, Jacob Coxon, resigned to protest what he called irresponsible AI development practices. TechCrunch noted Christiano has agreed to recuse himself from OpenAI matters and government model evaluations, but that the appointment still raises questions about industry influence over AI policy.
A prominent AI safety researcher joining a lab's board is only as meaningful as the recusals holding up in practice. Worth watching whether Christiano's seat changes any actual OpenAI decision, or just adds a credible name to the org chart.