Media: OpenAI board warned of a ‘breakthrough’ ahead of Sam Altman’s ouster

Shortly before the dismissal of Sam Altman, OpenAI researchers warned the board of a major AI breakthrough that ‘could threaten humanity’. Reuters reports, citing anonymous sources.

Sources described the employees’ letter as one item in a long list of board grievances that led to Altman’s removal as CEO. In particular, the leadership feared commercialising the technology before its risks were defined.

In an internal memo, the OpenAI board acknowledged the so-called project Q* (pronounced Q-Star). According to many employees, the development could be a breakthrough toward creating artificial general intelligence (AGI).

OpenAI and Altman define AGI as autonomous systems that exceed humans in most routine and computational tasks.

Q* currently can only solve simple math problems suitable for primary-school students, albeit using a vast amount of resources. Yet the company is confident that the technology holds substantial promise.

According to OpenAI experts, mathematics is the frontier in the development of generative AI. Currently, neural networks are able to predict the answer to a question and attempt to fit it with some probability of error.

Mathematical tasks require exact answers, so progress in this direction implies improved AI reasoning abilities.

Unlike a calculator, which performs a limited number of operations, AGI can generalise, learn, and understand, according to OpenAI.

According to one source, the company also had a dedicated group formed from the Code Gen and Math Gen teams. It allegedly studied optimisation and the possibilities of AI-model ‘consciousness’.

Earlier, four days after Altman’s ouster, OpenAI announced his return to the CEO post and a reshuffle of the board.