OpenAI Board Was Warned Of AGI Breakthrough Days Before Altman's Firing

Sam Altman may have returned as CEO of OpenAI, but the reasons behind the board's decision to dismiss him, and the lack of an explanation for this action which threatened the company's existence, remain unclear. It has emerged that days before Altman's dismissal, OpenAI researchers sent a letter to the board, warning of a significant AI breakthrough potentially leading towards AGI, which could pose a threat to humanity, according to a Reuters report.

This initiative was spearheaded by OpenAI's Chief Scientist, Ilya Sutskever, who also served on the board. Sutskever's concerns were crucial in the decision to fire Altman. In fact, during a Google Meet call, it was Sutskever who informed Altman of his dismissal. This letter was the final straw, following a longer list of grievances, including concerns about the commercialisation of OpenAI's advancements.

The project in question is Q*, the startup's work on Artificial General Intelligence, which involves autonomous systems capable of surpassing humans in most economically valuable tasks. OpenAI had developed a new model capable of solving certain mathematical problems. While these problems were only at graduate school level, the results made the researchers optimistic about Q*'s future success. The development was undertaken by a team comprising members from the Code Gen and Math Gen teams, focused on optimising existing AI models at OpenAI.

Researchers view mathematics as a frontier in generative AI. Currently, large language models (LLMs) can assimilate vast amounts of text, generating or predicting subsequent words or answers to equations, excelling in writing and translation. However, this area has some ambiguity, as there can be multiple correct answers. In contrast, mathematics offers a unique challenge, as it typically has only one correct answer, indicating that AI could develop greater reasoning capabilities, akin to human intelligence. Such advancements could significantly impact scientific research. It's important to distinguish this from calculators, which solve a limited range of operations. AGI, in contrast, can generalise, learn, and understand context.

Interestingly, a day before his dismissal, Altman hinted at this development at the Asia-Pacific Economic Cooperation Summit. He said, "Four times now in the history of OpenAI, the most recent time being just in the last couple of weeks, I've had the honour of being in the room when we pushed the veil of ignorance back and the frontier of discovery forward. Participating in these moments is the professional honour of a lifetime."

Sutskever turned out to be instrumental in Altman's return. After Altman was dismissed and Emmett Shear appointed as interim CEO, Microsoft announced that Altman and his co-founder, OpenAI President Greg Brockman, were joining Microsoft. This led to a widespread revolt at the company, with over 90% of the staff writing a letter demanding the board's resignation and Altman's reinstatement. Sutskever, a co-signee of this letter, even tweeted his regret over his role in Altman's firing.