“The model, called Q* – and pronounced as “Q-Star” – was able to solve basic maths problems it had not seen before, according to the tech news site the Information, which added that the pace of development behind the system had alarmed some safety researchers.”
– Dan Milmo
More reports are emerging about events at OpenAI close to the time of the dramatic developments that included the dismissal and then re-hiring of CEO Sam Altman. The Guardian draws from Reuters and other news providers in describing a possible leap in AI capabilities that sounded warning signals inside OpenAI.
Earlier reporting from multiple sources pointed to tensions between “go-fast” and “go-slow” approaches in how AI advances should be pursued and released.
The Guardian recalls a quote from Altman a week before he was dismissed, in which he told delegates to an Asia-Pacific conference, ““Four times now in the history of OpenAI, the most recent time was just in the last couple weeks, I’ve gotten to be in the room, when we sort of push the veil of ignorance back and the frontier of discovery forward, and getting to do that is the professional honour of a lifetime.”
The large language models that have captured so much attention in recent months, like OpenAI’s ChapGPT series, are based on words not mathematics. Computer science researchers say if math functors could be incorporated in their processing, it would be another breakthrough toward high-performing digital intelligence.
OpenAI ‘was working on advanced model so powerful it alarmed staff’ | THE GUARDIAN | November 23, 2023 | by Dan Milmo