Fortune has released a special report after interviewing insiders at OpenAI, following the resignation of several high-level executives, including Mira Murati, the CEO. In the past few months, several executives, such as Ilya Sutskever, a co-founder, and most recently Durk Kingma, a founding team member, have also announced their departure. This raises questions about what is happening at OpenAI.
According to the report, tension at OpenAI began before the unveiling of the GPT-4o model in May, which significantly increased capabilities and was introduced on the same day Google I/O was held. The rushed testing of the model prompted teams to work over 20 hours a day to meet the launch deadline. Murati, caught between the security and business teams, faced the challenge of negotiating a delay in the launch, but the business side refused.
The report indicates that the GPT-4o rollout deviated significantly from OpenAI’s standards. The security team had only 9 days to assess various aspects, working tirelessly to meet the launch deadline. Despite the model being deemed safe for release, vulnerabilities related to AI persuasion were identified, posing risks above standard levels. Nevertheless, OpenAI categorized the risk as “Medium,” not reaching the levels of “High” or “Critical” that would prevent model publication.
Fortune noticed that the conflict may have stemmed from the absence of CEO Sam Altman at the GPT-4o unveiling, with Mira Murati being the one to present the key features. A similar issue occurred with the o1 model launch last month, focusing on sequential thinking abilities. The rushed launch did not sit well with some researchers, as stability issues and unsatisfactory security assessments persisted.
Despite these challenges, Altman pushed for the swift release of the o1 model to maintain investor confidence in OpenAI’s AI leadership. As the company sought increased funding, adjustments to the o1 model became necessary to ensure a timely release, even at the cost of compromise.
Fortune’s report also addresses the rapid growth of OpenAI, especially after Altman’s departure. Within 10 months, the workforce expanded from 800 to over 1,800 employees, leading to significant organizational changes. The shift towards commercial products within a short span has altered the company culture, prioritizing product launches over research discussions. This structural shift has emphasized the For-Profit aspect, causing a notable change in dynamics at OpenAI.
Currently, OpenAI retains only 2 of its 11 founding members, Sam Altman and Wojciech Zaremba, who leads the OpenAI Codex team.
TLDR: OpenAI faces internal turmoil following multiple executive resignations, rushed model launches, security concerns, and organizational shifts towards commercialization, raising questions about its future direction and stability.
Leave a Comment