In a world where AI is rapidly reshaping everything from how we work to how we live, one truth stands out: education is the cornerstone of both innovation and safety.
On The AI Forecast, host Paul Muller sat down with Peter Norvig, former Director of Research at Google, Stanford Education Fellow, and co-author of the world’s most-used AI textbook—Artificial Intelligence: A Modern Approach. Their discussion spanned decades of AI progress, the deep transformation of education, and what the next evolution of data might look like.
Here are some takeaways from Paul and Peter’s conversation.
AI education must go beyond a static textbook, according to the author of the go-to AI education textbook.
Paul: How would you describe the state of AI education today?
Peter: It’s overwhelming. There’s just too much going on, too fast. When I started, you could keep up with every new development. AI papers came out slowly, and textbooks were relevant for years. Now, there are a dozen breakthrough papers published every week.
I don’t think the “one textbook” model works anymore. What we need is something interactive and personalized. And we need to shift from the idea that you go to college for four years, get a degree, and never have to learn again. That’s not how the world works. AI education should be a lifelong, continually evolving experience.
Paul: What’s your view on how AI is impacting education?
Peter: AI is changing both what we teach and how we teach. Tools like ChatGPT or GitHub Copilot can be great accelerators, but only if you already know what you’re doing. They can lead you astray if you don’t.
The real value of learning to program isn’t memorizing syntax – it’s developing judgment. It’s about knowing how to scope a problem, debug it, recover from failure, and think critically. That’s the mindset we should teach, whether people are coding or using AI tools. The goal is not to produce code. It’s to produce understanding.
AI is all about solving uncertainty.
Paul: What does AI mean to you?
Peter: In our book, we define it as making better decisions to accomplish your goals. But that’s also what economists and software engineers try to do. The real difference lies in the kind of problems AI takes on.
Traditional software is about complexity: writing exact rules to solve exact problems. But AI deals with uncertainty. You're asked to classify an image, decide what a sentence means, or predict how someone might vote. There’s often no ground truth. You're trying to make the best decision based on incomplete, noisy, or ambiguous data. That’s where AI lives.
Can AI give us the most unbiased view of our world?
Paul: What are you most excited about when it comes to the future of AI?
Peter: I’m excited about new ways of gathering data, especially video. We’ve made huge strides in training on text because it’s compact. Image models are catching up. But video? That’s still untapped.
The challenge is scale. Training on all of YouTube isn’t economically feasible today. But give it a few more generations of processing power, and we’ll get there. What excites me is that the video medium captures action. It gives a more accurate view of the world. And, unlike text or photos, it’s less biased.
Everything written is something someone thought was important. Every photo was taken deliberately. But some videos are just a camera running 24/7. That’s about as close as we get to an unbiased record of physical reality. Combine that with what we already know, and we start to connect the physics of the world with the psychology of the world.
We’ve improved processing power by a factor of 1,000 at least three times in my lifetime. We just need to do it one more time.
Catch the full conversation with Peter Norvig on The AI Forecast on Apple Podcasts, Spotify, and YouTube.
