OpenAI has educated a brand new committee to oversee “critical” security decisions related to the corporate's projects and activities. But in a move sure to attract the ire of ethicists, OpenAI has decided to staff the committee with company insiders – including OpenAI CEO Sam Altman – somewhat than outside observers.
Altman and the remainder of the safety committee — OpenAI board members Bret Taylor, Adam D'Angelo and Nicole Seligman, in addition to chief scientist Jakub Pachocki, Aleksander Madry (who leads OpenAI's “readiness” team), Lilian Weng (head of security systems), Matt Knight (head of security) and John Schulman (head of “alignment science”) — will likely be chargeable for evaluating OpenAI's security processes and precautions over the subsequent 90 days, in response to a post on the corporate's corporate blog. The committee will then present its findings and suggestions to the total OpenAI board for review, OpenAI says. At that point, an update will likely be provided on any proposals which have been accepted, “in a fashion consistent with safety and security.”
“OpenAI recently began training its next frontier model, and we expect the resulting systems will take us to the subsequent level of capability on our journey to (artificial general intelligence),” OpenAI writes. “We are proud to develop and release models which are industry-leading in each capability and safety, but welcome robust debate at this essential moment.”
OpenAI has in recent months several High-profile departures from the safety side of the technical team – and a few of these former employees expressed concerns about what they perceive as a deliberate downgrade of AI security.
Daniel Kokotajlo, who worked on OpenAI’s governance team, stop in April after losing confidence that OpenAI would “behave responsibly” in releasing increasingly powerful AI, as he wrote in a post on his personal blog. And Ilya Sutskever, OpenAI's co-founder and the corporate's former chief scientist, left the corporate in May after a lengthy dispute with Altman and Altman's allies – reportedly partly due to Altman's rush to bring AI-powered products to market on the expense of security work.
Recently, Jan Leike, a former DeepMind researcher who helped develop ChatGPT and ChatGPT's predecessor InstructGPT at OpenAI, resigned from his position as a security researcher. In a series of posts on X, he said that in his opinion OpenAI was “not on the suitable track” to “properly” address AI security and protection issues. AI policy researcher Gretchen Krueger, who left OpenAI last week, echoed Leike's statements: calls on the corporate to enhance its accountability and transparency, in addition to “the care with which it uses its own technology.”
quartz Remarks that along with Sutskever, Kokotajlo, Leike and Krueger, a minimum of five of OpenAI's most security-conscious employees have either quit or been fired since late last yr, including former OpenAI board members Helen Toner and Tasha McCauley. In a commentary for The Economist published On Sunday, Toner and McCauley wrote that they don’t imagine OpenAI might be relied upon to take responsibility for itself with Altman on the helm.
“Based on our experience, we imagine that self-management cannot reliably withstand the pressures of profit incentives,” said Toner and McCauley.
To Toner and McCauley's point, TechCrunch reported earlier this month that OpenAI's Superalignment team, chargeable for developing methods to guide and control “superintelligent” AI systems, was promised 20% of the corporate's computing resources – but received barely a fraction of that. The Superalignment team has since been disbanded, with much of its work placed under the purview of Schulman and a security advisory group formed by OpenAI in December.
OpenAI has advocated for AI regulation and has also made efforts to shape that regulation. Attitude an in-house lobbyist and lobbyist at a growing variety of law firms, spending a whole bunch of hundreds of dollars on U.S. lobbying within the fourth quarter of 2023 alone. Recently, the U.S. Department of Homeland Security announced that Altman will likely be among the many members of its newly formed Artificial Intelligence Safety and Security Board, which can make recommendations for “protected development and deployment of AI” in U.S. critical infrastructure.
To avoid the looks of playing ethical fig leaf with the executive-dominated security committee, OpenAI has committed to bringing in outside experts in “security, safety and engineering” to help with the committee’s work. These include cybersecurity veteran Rob Joyce and former U.S. Department of Justice official John Carlin. Aside from Joyce and Carlin, nonetheless, the corporate has not disclosed the scale or makeup of this outside expert panel—nor has it make clear the boundaries of the group’s power and influence on the committee.
In a post On X, Bloomberg columnist Parmy Olson notes that corporate oversight bodies just like the Safety and Security Committee, much like Google’s AI oversight bodies just like the Advanced Technology External Advisory Council, “do virtually nothing by way of actual oversight.” Significantly, OpenAI says She wants to make use of the committee to handle “justified criticism” of her work – although “justified criticism” is, in fact, in the attention of the beholder.
Altman once promised that outsiders would play a significant role in OpenAI's governance. In a 2016 New Yorker article, he wrote: said that OpenAI was “planning a strategy to allow large parts of the world to elect representatives to a … board of directors.” That never happened – and it currently seems unlikely that it’ll occur.
