OpenAI is allegedly self-censoring its research on the negative impact of AI, and it’s even led to the departure of at least two employees.
According to a new report from WIRED, OpenAI has become “more guarded” about publishing the negative findings of its economic research team, like data on all the jobs that AI might replace.
Employees are allegedly quitting over this, including data scientist Tom Cunningham, who now works as a researcher at METR, a nonprofit that develops evaluations to test AI models against public safety threats. According to the report, Cunningham wrote in an internal message at the time of his recent departure that the economic research team was pretty much functioning as OpenAI’s advocacy arm.
OpenAI began as a research lab, but has since gone through quite an evolution as the company shifted focus towards its commercial products that generate billions of dollars in revenue.
The company’s economic research operations are reportedly being managed by OpenAI’s first chief economist, Aaron Chatterji, who was hired late last year. Under Chatterji, the team recently shared its findings that AI use could save the average worker 40 to 60 minutes a day.
According to the WIRED report, Chatterji reports to OpenAI’s chief global affairs officer, Chris Lehane, who has earned himself the reputation of “master of disaster” with his work for former President Bill Clinton (and years later for Airbnb and Coinbase), and is largely considered the expert on damage control.
This isn’t the first time OpenAI has been accused of favoring product over safety research. Just last month, a New York Times report accused OpenAI of being well aware of the inherent mental health risks of addictive AI chatbot design and still choosing to pursue it.
It’s also not the first time a former employee has deemed OpenAI’s research review to be too harsh. Last year, the company’s former head of policy research, Miles Brundage, shared that he was leaving because the publishing constraints had “become too much.”
“OpenAI is now so high-profile, and its outputs reviewed from so many different angles, that it’s hard for me to publish on all the topics that are important to me,” Brundage shared in a Substack post.
Not only is artificial intelligence changing every aspect of modern-day society, but it is also already proving to have a colossal impact on the economy. AI spend is probably propping up the entire American economy right now, according to some reports. And while the jury’s still out on just how effectively and to what extent AI can take over jobs, early research says that AI is already crushing the early career job market. Even Fed chair Jerome Powell has admitted that AI is “probably a factor” in current unemployment rates.
At the core of this outsized impact of AI is OpenAI. The company is at the heart of a tangled web of multibillion-dollar dealmaking, and ChatGPT is such a central product that it has become almost synonymous with the word “AI chatbot.”
OpenAI is also the centerpiece of Stargate, the Trump administration’s mysterious but massive AI data center buildout plan. Trump and his officials have stood squarely behind the positive potential of AI, while casting aside concerns echoed by competitors like Anthropic, such as fear-mongering or doomerism.
OpenAI executives have also been caught up in an industry-wide divide over AI safety playing out on Capitol Hill. OpenAI President Greg Brockman is one of the top backers of “Leading the Future,” a super-PAC that views most AI safety regulation as an obstacle to innovation.
Read the full article here
