In a series of recent interviews and statements, several former employees of OpenAI have expressed their concerns about the company’s direction and its stance on AI safety regulations.
These individuals, who previously worked on the Alignment and Superalignment teams at OpenAI, have decided to speak out despite facing potential legal and economic pressures.
‘Disappointed but not surprised’: former employees speak on OpenAI’s opposition to SB 1047 https://t.co/ePP41vOznI
— TechCrunch (@TechCrunch) August 23, 2024
Broken Promises and Loss of Faith
According to Gary Marcus, who spoke to three former OpenAI employees over the summer, there has been a consistent message: promises have been made and not kept, and they have lost faith in both Sam Altman personally and the company’s commitment to AI safety.
One of these former employees, William Saunders, who worked at OpenAI for three years, has chosen to go on the record with his concerns.
Saunders, who resigned from OpenAI on February 15, 2024, emphasized that while current AI technology may not be as concerning, future advancements could pose significant risks if not properly managed.
He believed that internal governance and external oversight are crucial to ensuring the safe development of AI systems.
OpenAI’s Opposition to SB 1047
The former employees’ concerns were further amplified when OpenAI recently announced its opposition to California’s SB 1047, a bill aimed at preventing AI disasters.
This move was seen as a departure from Sam Altman’s previous public support for AI regulation.
Daniel Kokotajlo and William Saunders, two former OpenAI researchers who resigned earlier this year due to safety concerns, expressed their disappointment but were not surprised at the company’s stance.
In a letter shared with Politico, they urged California Governor Gavin Newsom to sign the bill, stating, “Sam Altman, our former boss, has repeatedly called for AI regulation. Now, when actual regulation is on the table, he opposes it.”
The Need for Whistleblower Protection
One of the most important aspects of SB 1047, according to Saunders, was its whistleblower protections. He emphasized that future whistleblowers should be able to speak freely to the California Attorney General about their concerns without fear of reprisal.
While OpenAI has relaxed some of its restrictions on nondisparagement, much of what whistleblowers might need to disclose could still be precluded by confidentiality agreements.
SB 1047 would provide a legal framework for whistleblowers to share information with the authorities, potentially allowing for more oversight and accountability in the AI industry.
Read more: With a mysterious reptile tail tease, Taylor Swift ends her social media blackout.
Conclusion
The concerns raised by former OpenAI employees highlight the need for a more comprehensive approach to AI safety and governance.
As technology continues to advance rapidly, it is crucial that companies, regulators, and the public work together to ensure that the development of AI systems prioritizes safety and ethics.
The opposition to SB 1047 by OpenAI, despite its CEO’s previous calls for AI regulation, raises questions about the company’s true commitment to responsible innovation.
As the debate around AI governance continues, the voices of whistleblowers and former employees will be essential in shaping the future of this transformative technology.