Reuters: Before Axing Altman, Board Got Cautionary Letter

It's a thinly sourced report but an intriguing one: Reuters reports two sources have told it that ahead of OpenAI CEO Sam Altman's firing, the board of directors was sent a warning letter penned by staff researchers that warned "of a powerful artificial intelligence discovery that they said could threaten humanity." Reuters has not seen a copy of the letter nor was it able to secure a comment from OpenAI or the staffers who it says wrote the letter. The sources say the letter was just one of a number of factors that led to Altman's firing, which was reversed four days later.

Reuters suggests the letter may have been related to a project dubbed Q* (that's Q-Star), which is apparently able to solve math problems at a grade-school level. Reuters explains the potential significance: Generative AI's current strength is in writing and language-oriented tasks as it's able to statistically predict the next word—and when it comes to the next word, there are plenty of viable potential options. The same doesn't hold true for math, where there is only one correct answer. As such, AI would need superior reasoning capabilities akin to human intelligence to do math. If true, CNBC reports Q* could be a breakthrough in the search for superintelligence (which OpenAI calls artificial general intelligence), or a system whose intelligence surpasses that of humans. (More OpenAI stories.)