Godfather of AI’ says chatbots need ‘maternal instincts’ – but what they really need is to understand humanity

Geoffrey Hinton, scientist, former Google employee, and widely recognized ‘Godfather of AI,’ has made a late-stage career of criticizing his godchildren. And now he’s taken it all a step further, insisting we need “AI Mothers,” not AI Assistants.

Speaking at the AI4 Conference in Las Vegas this week, and as first reported by Forbes, Hinton again sounded the alarm on the impending advent of Artificial General Intelligence, which he now believes will arrive in a few years, a notion that syncs with recent comments from OpenAI CEO Sam Altman.

That acceleration from what was once thought to be decades to a few orbits around the sun is, perhaps, what prompted Hinton to argue that we need something other than AI Assistants.

“We need AI mothers rather than AI assistants,” Hinton said, according to Forbes. The idea, Hinton posits, is that AI’s with “maternal instincts” are a sort of protection system. After all, mothers generally don’t harm and usually protect their children.

If AI systems like ChatGPT, Claude AI, and Gemini truly become smarter than us in a matter of years, having them in some way feel as if it’s their job to look out for us might prevent them from harming us or society.

Hinton, who recently won a Nobel Prize and helped develop the technological foundation that arguably made all this AI possible, left Google in 2023 and immediately started warning people about a dire AI future. Imagine a parent disowning their child, and you get the idea.

I don’t think Hinton is turned off from AI. After all, he can’t stop talking about it, and appears to recognize its potential, but it’s also clear it scares him.

He previously told The New York Times in 2023 that

Increased competition is leading to a less cautious approach
He presciently warned about the flood of fake online content
He warned about AI taking on all the jobs we don’t want to do (it may be taking others, as well)
He worried about AI that can both program and then run programming (a very dangerous closed loop)
And he was talking almost immediately about AI outsmarting us

So, sure that day is now fast approaching, but is a motherly AI what we want or need? I don’t think so.

The minute we start training “Mom Instincts” into AI, it will start to act like a mother and slip into that creepy, uncanny valley where you can no longer tell if you’re talking to a program or a person. Motherly instincts imply warmth, compassion, caring, understanding, and love. I don’t want those things from an AI.

What I think we need, though, is for AI assistants to understand what it means to be human. Put another way, if AI chatbots can at least understand humanity, they can serve us better. They can also recognize our propensity for trust and perhaps finally stop presenting us with false narratives and fake friendliness and interest.

We shouldn’t want companionship out of our super-intelligent AI systems. Instead, we need utility and trust, an ability to carry out our wishes in a way that best serves our interests.

The last thing we need is an AI full of maternal instincts, which then makes its own choices and, when things go awry, insists, “Well, dear, mother knows best.”