What does it actually mean to be ‘Robot-Proof’? For author Dr. Vivienne Ming, the answer doesn’t lie in competing with machines, but in mastering the uniquely human skills they cannot replicate.
Joining FCAT’s VP of Research John Dalton, Ming discusses the concept of ‘hybrid intelligence’ and how we can use AI to magnify human reasoning.
John Dalton: There are so many books and “takes” on AI right now, what motivated you to throw your hat into the ring?
Vivienne Ming: I’ve been working on AI for 30 years, and I thought, “wouldn’t it be nice if someone who’s been in the field that long contributed to the conversation?” And it’s worth noting that the core issues really haven’t changed much in that time. Yes, everyone can play with AI now, but the fundamentals of what I was doing building models in education, science, and workforce have held. I built my first model in 1999. It just turns out that building models is an astonishing way to learn more about people, our brains, and societies. And I’m passionate about what it means to be human, what our powers are versus what machines can do.
John Dalton: While reading your book, I was struck by the tone. It’s clearly written from a place of deep concern. Given where we are today, what are your biggest concerns?
Vivienne Ming: I remember hearing Sam Altman appearing before Congress and he asked the audience to imagine a world in which every child has their own AI tutor. Well, I don’t have to imagine that world. AI in education is one of the oldest fields of applied research. For 50 years, we’ve been building tutors on computers for kids to work with. And if there’s one golden rule we’ve learned about AI tutors is that if the computer gives the kid the answer, the kid doesn’t learn anything. And yes, this finding has been strongly replicated with LLMs. So, here’s the man who is sort of like the face of AI getting up in front of Congress, seemingly unaware of what a complicated and yet well researched thing he was pitching to the world. What do you think that says about our understanding of what AI will do as co-workers, assistants, or co-pilots? Don’t misunderstand me. I’m all in on AI, but I’m really tired of the naïve utopian and dystopian commentary.
John Dalton: In your research, you’ve looked closely at what kinds of people use AI most effectively. What have you learned?
Vivienne Ming: There are things that humans are uniquely good at, and things that machines are uniquely good at, and there’s a way that they fit together that makes both of us better. The thing about Altman’s comments and the general sales pitch is that AI is just a bunch of robotic minions that will carry out our every whim. But if you outsource your mind to a machine, you’re going to struggle to do your job, much less getting better at your job. What it boils down to, at least today, is that these LLMs are much better than us at answering well-posed problems. We humans are good at exploring and reasoning through ill-posed problems. We’re actually terrible at it, but still better than anything else. There is a sweet spot there where we can combine AI’s as magnifiers for mapping known domains and our ability to explore uncertainty. It’s a form of hybrid intelligence where you get the best of both together. In my research, we had a group of humans compete with a group of human/AI cyborgs to predict the price of oil. It should come as no surprise that the cyborgs outperformed the humans. When you put the computer and the human together in this way, they become the smartest thing on the planet. That gives me hope.
John Dalton: What would you tell a CEO struggling to balance the need for greater efficiency (which usually involves more automation) and keeping the human touch in their business?
Vivienne Ming: My takeaway from all this is that we can build better people, people who are well positioned to thrive with hybrid intelligence. We focus too much on making models that are better; we need models that make us better. I go into this in great detail in the book, but the research tells us that the best predictor of the kind of people who are best prepared to excel as explorers have fluid intelligence. They are humble and curious. They don’t just accept answers from an AI, they push further. They are highly adaptive; if the AI is right, they accept it and then take that lesson and do something new with it. And I love this: people who score higher in perspective-taking, the ability to understand other people, it was a good predictor of their hybrid intelligence. Who thinks about social skills when we talk about AI? But we should. What an amazing thing – the things that make us better suited to working with AI are deeply human things: humility, curiosity, perspective-taking. That’s what we need to build in our children and ourselves.