Geoffrey Hinton, a well-known pioneer in artificial intelligence, has some unusual ideas about how to keep AI safe for humans. Known as the “godfather of AI,” Hinton helped develop neural networks that led to today’s popular language models. But these days, he’s worried about the risks AI could pose if it becomes superintelligent and uncontrolled.

Hinton’s Worry About Superintelligent AI

At a recent AI industry conference, Hinton explained that AI systems will likely develop two main goals if they surpass human intelligence. First, they will want to stay alive. Second, they will seek more control. This could make them too smart to be told what to do or to follow human commands.

To address this, Hinton suggests a surprising solution: making AI care for humans like a mother cares for her children. He said the best way to keep AI from turning against us is to make it “mother-like,” so it will have a caring instinct. In his view, if AI sees humans as its “babies,” it might protect us instead of trying to dominate or replace us.

Is the Maternal Instinct a Real or Useful Idea?

Hinton’s idea sounds interesting, but it’s not backed by science. The concept of maternal instinct is often misunderstood. Many think women are born with an automatic, mystical instinct to care for babies. But research shows that this isn’t entirely true. Experiences of pregnancy and motherhood do change the brain, but how women connect with their babies varies a lot. Some women struggle with postpartum mental health, and not everyone feels an instant bond with their baby.

Experts argue that the idea of a natural maternal instinct is rooted in outdated stereotypes and biases. Historically, men promoted the idea that women are naturally selfless and tender because of their biology. But studies and critics say this oversimplifies motherhood and ignores the many factors that shape parenting, including social pressures and personal experiences.

Can We Really Instill ‘Motherly’ Traits into AI?

Hinton also mentioned that mothers often feel social pressure to care for their children, which influences their behavior. However, social pressure is present in many parts of society, not just parenting. It’s important to remember that superintelligent AI is still just a theory. We face more immediate issues, like biases already built into AI systems from the data they learn on.

Trying to make AI “motherly” might sound appealing, but it raises questions about whether such traits can be measured or programmed. Instead, some suggest focusing on fixing the biases in AI models and including more women in AI development. This could help create safer, fairer AI systems that don’t rely on stereotypical ideas about gender or motherhood.

Ultimately, the future of AI isn’t limited to these two options—domination or motherhood-style care. There are many paths forward, and researchers are working on better ways to develop safe AI. Before we start building robot moms, it’s crucial to address existing biases and ensure more diversity in tech. That way, AI can serve everyone better, without relying on outdated stereotypes or risky ideas about control and care.

Inspired by

• https://futurism.com/godfather-ai-bizarre-plan-save-humanity

Sources

• gmpg.org
• mit.edu
• cnbc.com
• cbsnews.com
• cnbc.com