The departure of Dario and Daniela Amodei from OpenAI was driven not only by safety concerns but by years of escalating personal conflicts with co-founders Sam Altman and Greg Brockman, according to a Wall Street Journal investigation.
The tensions erupted in multiple confrontations over credit, control and trust, as "Hvylya" reported, citing the WSJ account. In one incident, Altman called the Amodei siblings into a conference room and accused them of plotting against him by encouraging colleagues to send negative feedback to the board. When a third executive was brought in and denied any knowledge of the claim, Altman then denied having said it - prompting the Amodeis to begin shouting at him.
The conflicts also involved a bitter fight over OpenAI's landmark language model project. Dario Amodei, then research director, blocked Brockman from working on the project. Daniela, who was co-leading it, told Brockman he could not join and offered to step down rather than allow him on the team. Altman ultimately sided with the Amodeis, but the standoff left lasting scars.
By March 2020, relations had grown so tense that Altman asked executives to write peer reviews of each other. Brockman wrote a lengthy critique of Daniela, accusing her of abusing her power to "create bureaucratic processes to get her way." Altman reviewed the document beforehand and declared it "tough but fair." Daniela delivered a point-by-point rebuttal, and the fight over the dueling feedbacks escalated so badly that Brockman offered to withdraw his critique entirely.
Earlier in their partnership, Dario had felt disturbed when Brockman raised the idea of selling artificial general intelligence to governments - specifically to the nuclear powers on the UN Security Council. Amodei considered the notion tantamount to treason and nearly quit. By late 2020, he, Daniela and about a dozen colleagues had left to found Anthropic.
Earlier, "Hvylya" reported on what OpenAI's top scientist actually expects from AI instead of AGI.