David Robinson, a former leader on OpenAI's Safety Systems team, resigned from the company due to what he described as 'cognitive dissonance.' Robinson stated that despite internal warnings about the dangers of advanced AI models, the company continued to develop and deploy them at a rapid pace.
The departure of a key safety lead from OpenAI, citing internal contradictions regarding AI development speed and safety, raises questions about the company's commitment to responsible AI deployment and the potential risks associated with advanced AI models.
David Robinson, a former leader on OpenAI's Safety Systems team, has resigned from the artificial intelligence company, citing a sense of 'cognitive dissonance.' Robinson stated in an interview with The New York Times' "The Ezra Klein Show" that he felt the company was saying one thing while doing another.
Robinson explained that despite internal warnings about the capabilities of new AI models and expressed interest in slowing down AI development, OpenAI continued to move rapidly behind the scenes. "Part of what was the fundamental cognitive dissonance for me was we keep publishing these warnings, but ultimately, we're still training and deploying these dangerous models that we're warning about," Robinson said.
This sentiment echoes concerns raised by other OpenAI employees. In July, hundreds of employees signed an open letter titled "Pacing the Frontier," urging the government to impose regulations to slow AI development. In September, OpenAI CEO Sam Altman publicly agreed with Anthropic CEO Dario Amodei's essay advocating for AI companies to coordinate a slowdown among themselves.
However, Robinson described continued "frenetic" activity and employees "running on fumes" within OpenAI's offices. He characterized the internal environment as a "ballet or a kind of dance" with "a lot of adrenaline."
Researchers at OpenAI have previously been candid about the dangers of AI's recursive self-improvement (RSI). Jakub Pachocki, OpenAI's chief scientist, warned in September that "This is a time that calls for extreme caution." OpenAI did not immediately respond to a request for comment. A spokesperson previously stated that the company strengthens its safety and security practices and pauses training when necessary to manage risks.
Pick the topics you care about. Get only what matters, on your cadence.