Videos
AI Revolution: Is AI Driving a New Industrial Revolution?
Insights
- AI’s progress follows exponential, not incremental, growth—doubling compute power every two years for six decades.
- Unlike past industrial revolutions that plateaued, AI’s trajectory compounds on itself, leading to explosive acceleration.
- This exponential era is transforming not just technology but how humanity understands and creates knowledge.
Is artificial intelligence the next industrial revolution—or something far greater?
This talk features Professor James Fergusson, Executive Director of Data Intensive Science at University of Cambridge and Director of Infosys-Cambridge AI Centre.
Professor Fergusson examines the exponential nature of AI’s growth and its implications for science and society. Drawing an analogy between the steam engine’s gradual evolution and AI’s explosive trajectory, he shows how compute power—doubling every two years for over half a century—has propelled AI into a new phase of human advancement. Unlike earlier innovations that reached efficiency plateaus, AI compounds on itself, enabling systems that grow smarter, faster, and more capable with each iteration.
From modeling complexity to redefining discovery itself, Fergusson argues that AI represents not merely another industrial revolution, but a fundamental shift in the pace of human progress—an exponential era where innovation accelerates beyond our historical imagination.
This session offers researchers, technologists, and innovators a glimpse into how AI will not just accelerate discovery—but reshape the scientific process itself.
Professor James Ferguson:
Hi, welcome to this short talk on the AI revolution. Is AI driving a new industrial revolution? I'm Professor James Ferguson. I’m the executive director of Data Intensive Science Group here in Cambridge, and I'm also the director of the Infosys Cambridge AI Center. So, often when people talk about AI, they say it's the next industrial revolution. And so I think it's worth looking at that comparison and seeing really is that true and what's true about it and what maybe is misleading about thinking about it as an industrial revolution. So here is a plot of the efficiency of steam engines from around 1700 through to now. And the key thing you can see is the industrial revolution really began around early 1800s when steam engines were made. There were around 40 in London at the time. They had about five to 10 horsepower and they're about 5% efficient. And so what happened is that we've slowly improved our ability to build steam engines. You can see the efficiency going up, the normal S curve of innovation. We get the technology and then we have a rapid period where we improve it and then there's a stable period where we've got to the point where we've built it about as good as we can and it really plateaus. And so we've seen in the last 300 years, we’ve built, we went from that five to 10 horsepower with 5% efficiency to 800,000 horsepower engines, which is what a large gas turbine and a generator would be. And they work about 60% efficiency, which is pretty close to the maximum you'd expect from thermodynamics. And so this is pretty impressive, right? It's a growth of about 160,000 in 300 years. And you can see that clear S curve of innovation. And so let's think about what it looks like in terms of AI. Right, so AI is really driven by the exponential growth in compute. And if we think about computing, you know, it grows according to Moore's law and so it's doubling every two years and has done for 60 years. Here's this beautiful plot of the number of transistors, which is a good proxy for the computing power, on a log plot and it's very, very straight, amazingly straight really. And so the thing about this is its exponential growth. And so in 300 years, we're expected to grow by 1.4 times 10 to the power 29, or this number with lots and lots of zeros x growth. And this is what's really different about the AI revolution, is this exponential growth in compute means that we have much, much more rapid transformation. It's as if we were building steam engines, but the energy density of coal was doubling every two years. And so if we thought about that, what would that look like, to kind of see the analogy, it means that in 1800, right, we started with these little 10 horsepower engines. But if we add Moore's law for the power of engines, well, by 1810, they’d have 310 horsepower. So now, going from these very large, sort of fairly low power engines, which were, you know, very exciting at the time, but quite weak by today's standards, we've now got something that's, you know, quite a powerful car today. Another 10 years, that'd be 10,000 horsepower. So now we've got, you know, the kind of large engines you find in ships. By 1830, we'd have 300,000 horsepower. 1840, we'd be about 10 million. And then by 1850, we'd have about 335 million horsepower. And this is what exponential growth really means. It's something that people don't really appreciate, that actually what we're doing with computers is probably one of the most impressive technological feats humans have ever had which is maintaining exponential growth over a really long period, leads to this really sort of quite scary increase in power. And if you look at that, essentially that 350,000 is double what we use to get to the moon in the Apollo rocket. And so you can see a technology that starts out very, very small in 50 odd years gets to the point where it's incredibly powerful. And that's what we're seeing with AI. And so if we look to the future, the real difference between the AI revolution and the industrial one is that our S-curve innovation is built on this exponential. And this drives really explosive growth. So this is a log plot of the size of AI systems. And you see that's growing exponentially on an exponential. And this means that we expect AI to improve very rapidly, much more rapidly. We're not going to expect this kind of S-curve plateau that we would in other technology kind of things.
