AI security expert resigns from Anthropic, says the world is in danger
Mrinank Sharma led an AI safety team at Anthropic and now wants to dedicate himself to poetry. He is the next former employee of AI companies with concerns.
(Image: Andrey_Popov/Shutterstock.com)
A senior AI safety researcher has left the AI company Anthropic because he sees the world in danger. Mrinank Sharma justifies this cryptic warning not only with the development of AI, but also with biological weapons and current global crises. In a farewell letter to colleagues after his resignation, he also mentions the pressure he felt within Anthropic, which would push aside what really matters. Sharma now wants to go underground for a while and dedicate himself to poetry.
Sharma's resignation occurred just one day before an AI researcher left OpenAI. Zoe Hitzig, however, justified her departure from the Anthropic competitor with the fact that ChatGPT is introducing advertising in the USA. "OpenAI has the most detailed record of private human thoughts ever compiled," she writes on X. "Can we trust them to resist the temptations to misuse these records?"
Warning about dependence on AI
Hitzig pointed to "early warning signs" indicating that dependence on AI tools is "worrying." Similar to social networks, this could negatively impact users' mental health and even lead to delusions. "It's really dangerous to create an economic engine that profits from fostering these new relationships before we understand them," she said in conversation with the BBC.
Videos by heise
However, OpenAI dismissed the concerns, telling the BBC: "Our mission is to ensure that AGI (Artificial General Intelligence, editor's note) benefits all of humanity; our advertising commitment always serves to support this mission and the goal of making AI more accessible." The AI company further stated: "We treat your conversations with ChatGPT confidentially and never sell your data to advertisers."
AI companies question their own values
Although the former employees of the high-ranking AI companies justify their resignations differently, Hitzig and Sharma have similar concerns. The ex-employee of Anthropic also admits to having "repeatedly seen how difficult it is to let our values truly guide our actions." This is because the AI offered by the startups can not only be used for useful purposes; Anthropic's Claude was also misused by cybercriminals, as the AI company itself states. However, there are also doubts about this autonomous AI cyberattack.
In any case, according to Mrinank Sharma, it is "time to move on," even though he enjoyed his time at Anthropic. Because "the world is in danger," he writes on X. "And not just from AI or bioweapons, but from a whole host of interconnected crises unfolding right now." Sharma states that he will return to Great Britain and "become invisible for a while." Instead of dealing with AI, he hopes to "begin a study of poetry."
(fds)
