OpenAI CEO Sam Altman recently admitted he needs extensive assistance from his company’s AI chatbot to get through life.
During an interview last night with late-night talk show host Jimmy Fallon, Altman argued that raising a child without constantly pestering a chatbot for life advice is now impossible — even though humans have managed just fine since the dawn of time without it.
“I cannot imagine having gone through, figuring out how to raise a newborn without ChatGPT,” Altman told Fallon during his late-night debut on Monday. “Clearly, people did it for a long time — no problem. But I have relied on it so much.”
For instance, he purportedly asked ChatGPT why his newborn son kept “dropping pizza on the floor and laughing.” He even admitted to hiding in a bathroom during a social gathering to ask the tool whether it was normal for his child not to be able to walk at just six months.
Considering the tech’s well-documented track record of distorting the truth and giving terrible advice, it’s an alarming admission. While Altman’s storytelling was presumably meant to serve as a relatable punch line during a high-visibility interview, the risks are far higher than he’s letting on, particularly when it comes to raising young children.
Researchers have warned that over-relying on large language models could be dangerous — arguing in a 2024 study, for instance, that there’s a “critical need for expert oversight of ChatGPT” for “safeguarding child healthcare information.”
“ChatGPT and other large language models have the potential to transform the health information landscape online,” they wrote. “However, lack of domain-specific expertise and known errors in large language models raise concerns about the widespread adoption of content generated by these tools for parents making healthcare decisions for their children.”
In an experiment, the research team led by University of Kansas doctoral student Calissa Leslie-Miller found that participating parents struggled to distinguish between real medical advice from experts and unverified advice generated by ChatGPT.
“During the study, some early iterations of the AI output contained incorrect information,” she said at the time. “This is concerning because, as we know, AI tools like ChatGPT are prone to ‘hallucinations’ — errors that occur when the system lacks sufficient context.”
Despite the real risks of being exposed to potentially dangerous information, parents are increasingly turning to ChatGPT for advice.
... continue reading