The news comes days after OpenAI co-founders Ilya Sutskever and team leaders Jan Leike announced their departure from the Microsoft-backed startup. Reike wrote Friday that OpenAI's “safety culture and processes have taken a backseat to its shiny product.”
The news was first reported by Wired.
Announced last year, OpenAI's Superalignment team has focused on “scientific and technological breakthroughs to steer and control AI systems that are much smarter than we are.” At the time, OpenAI announced it would commit 20% of its computing power to the effort over four years.
Sutskever and Reike announced their departures on Tuesday at different times, but on Friday, Reike provided further details about why he left the startup.
“I joined because I thought OpenAI was the best place in the world to do this research,” Reike wrote about X. “But I've been in disagreement with OpenAI leadership about the company's core priorities for quite some time, and we've finally come to an agreement,” he said at the breaking point. ”
Reich wrote that he believes more of the company's bandwidth should be focused on security, surveillance, preparedness, safety and social impact.
“These problems will be extremely difficult to resolve, and I am concerned that we are not on track to get there,” he wrote. “Over the past few months, my team has faced headwinds, at times struggling with computational power and making it increasingly difficult to carry out this important research.”
Rike added that OpenAI must become a “safety-first AGI company.”
“Building machines that are smarter than humans is an inherently risky endeavor,” he writes. “OpenAI has a great responsibility on behalf of all humanity, but in recent years safety culture and processes have taken a backseat to shiny products.”
Reike did not respond to a request for comment, and OpenAI did not respond to a request for comment.
The high-profile departure comes months after OpenAI went through a financial crisis involving co-founder and CEO Sam Altman.
In November, OpenAI's board of directors fired Altman, alleging in a statement that he had not been “consistently candid in his communications with the board.”
The Wall Street Journal and other media have trained Sutskever to focus on ensuring that artificial intelligence does not harm humans, while others, including Altman, have rather focused on promoting artificial intelligence. The issue seemed to grow in complexity with each passing day. We provide new technology.
Altman's firing prompted resignations or threats of resignation, including an open letter signed by virtually all of OpenAI's employees, and an uproar from investors including Microsoft. Within a week, Mr. Altman returned to the company, and the board members who voted to fire Mr. Altman, Helen Toner, Tasha McCauley and Ilya Sutskeva, resigned. Sutskever remained on staff at the time, but her position as director is no longer valid. Adam DiAngelo, who voted to fire Altman, also remained on the board.
When asked about Sutskever's status on a Zoom call with reporters in March, Altman said he had no latest information to share. “I love Elijah…I want to work with him for the rest of our careers, my career, whatever,” Altman said. “I have nothing to announce today.”
On Tuesday, Altman shared his thoughts on Sutskever's departure.
“This is very sad for me. Ilya is without a doubt one of the greatest minds of our generation, a leader in our field, and a dear friend,” Altman wrote. with X. “While his brilliance and foresight are well known, his warmth and compassion are less well known, but equally important.” Altman, a research scientist who has worked at OpenAI since 2017. Director Jakub Pachocki will replace Sutskever as principal investigator, he said.
The news of Sutskever and Leike's departure and the dissolution of the Super Alignment team comes as OpenAI launches a new AI model for ChatGPT, along with an updated user interface, the company's latest effort to expand the use of its popular chatbot. This comes a few days after the desktop version was announced.
The update will make the GPT-4 model available to everyone, including OpenAI's free users, head of technology Mira Murati said in a livestream event on Monday. She added that the new model, her GPT-4o, is “much faster” and has improved text, video and audio capabilities.
OpenAI said it plans to eventually allow users to video chat with ChatGPT. “This is the first time we've really taken a big step forward in terms of ease of use,” Murati said.
