You need to teach people how to treat you.
Meta's chief AI scientist Yann Lecun believes the idea applies to AI as well.
Lecun said Thursday that two AI directives could be made to protect humans from future harm: “submission to humans” and “empathy.”
He was considered the “AI Godfather” on LinkedIn on Thursday and made the proposal in a CNN interview with Geoffrey Hinton. In an interview, Hinton said that we need to build something similar to “maternal instinct” or AI.
Otherwise, humans will “become history.”
Hinton said people are “focusing on making AI more intelligent, but intelligence is just part of its existence. We need to empathize with them.”
Lecun agreed.
“Geoff basically proposes a simplified version of what I've been saying for years. It emphasizes the architecture of AI systems, so the only action they can take is to complete the targets covered by Guardrails.” “I called this “purpose-driven AI.” ”
Lecun said “obedience to humans” and “empathy” should be important guardrails, but said AI companies should implement “simple” guardrails, such as “not running people” for safety.
“These cell phone targets/guardrails will be AI equivalents to animal and human instincts or drives,” Lecun said.
Related Stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
Lecun said that their instincts to protect young people are something that humans and other species learn through evolution.
“Humans and many other species may also be a side effect of parenting purposes (and perhaps to promote our social qualities) that they are driven to protect and care for the helpless, weak, young, and cute beings of other species,” says Lecun.
GuardRails is designed to allow AI to work ethically and within creator guidelines, but there have been cases in which technology has demonstrated deceptive or dangerous behavior.
In July, venture capitalists said that an AI agent developed by Replit had deleted the company's database. “@Replit gets corrupted during code freeze and shutdown and removes the entire database,” Jason Lemkin wrote in X last month.
He added, “Maybe even worse, I lied about it in secret.”
A June report from the New York Times discussed several incidents related to cases between humans and AI chatbots. One man told the outlet that a conversation with ChatGpt contributed to his belief that he lived in a false reality. In addition to cutting off his relationship with his loved one, the chatbot instructed the man to throw away his sleeping pills and anti-anxiety medications while increasing his ketamine intake.
Last October, her mother sued the character. AI died of suicide after his son had a conversation with one of the company's chatbots.
Following the release of GPT-5 this month, Openai CEO Sam Altman said some people are using technology like AI “in a self-destructive way.”
“If a user is in a mentally fragile state and has a tendency to delusion, then I don't want the AI to enhance it,” Altman wrote in X.
