The AI revolution might end up being much more consequential than the industrial revolution, but there are parallels that one can draw between the two.
Speaking on the rapid acceleration of AI development, Demis Hassabis, CEO of Google DeepMind, offered a compelling analogy, comparing the internet’s role in powering AI to that of coal and oil in the Industrial Revolution. His insights reveal a nuanced perspective on the trajectory of AI, balancing a natural inclination for cautious, scientific exploration with the undeniable momentum of technological progress.
Hassabis articulates a common concern within the AI community: “I actually have some sympathy with that view,” he states, referring to those who advocate for slowing down AI development. He reveals his initial vision for AI’s evolution: “My original plan for this was it would be a scientific endeavor and stay a scientific endeavor for a very long time—this kind of CERN idea.” He further elaborates on an ideal scenario: “If I could wave a magic wand and create the ideal setup, everyone would be working on more narrow AI tools like AlphaFold, which are not general intelligences, but they use that same technology. And we would be doing a lot of very beneficial things like curing cancer and helping with new materials and some of the scientific stuff, either while we are learning about what these general systems are. And then we would take that next step.”
However, Hassabis acknowledges that reality has diverged from this ideal: “But it just, the technology hasn’t turned out that way. It’s turned out because the internet was available.” He then draws his powerful historical parallel: “I sometimes think of it as the Industrial Revolution. This is how I describe it sometimes. What’s happened is we’ve had—we’re about to have an industrial equivalent of, or maybe it’s going to be bigger than the Industrial Revolution—but imagine the Industrial Revolution happened, but there was no coal or oil. We just happened to have dead dinosaurs and dead trees in the ground. But it didn’t have to be like that. Think about that. That’s sort of completely accidental that’s the case. And so if you weren’t able to power the Industrial Revolution by just digging up some coal or surface oil, imagine you had to jump straight to solar or nuclear. How much harder that would have been.”
This analogy underscores a critical point: the availability of vast datasets, readily accessible through the internet, has acted as an unforeseen accelerant for AI. Just as readily available fossil fuels fueled an industrial boom, the internet’s digital “fuel” has propelled AI development at an unprecedented pace. This abundance of data has, in a sense, allowed AI to bypass a more deliberate, research-focused phase, pushing it directly into widespread application and the pursuit of more general intelligence.
The implications of this unplanned acceleration are profound. While the potential for transformative breakthroughs in areas like healthcare and scientific discovery remains immense, as exemplified by tools like AlphaFold, the rapid advancement of general AI systems raises complex ethical and societal questions. The current race in AI development, with companies like Google DeepMind, OpenAI, and Anthropic at the forefront, reflects this dynamic. The challenge now lies in navigating this accelerated path responsibly, ensuring that the incredible power of AI is harnessed for the benefit of humanity, even if its foundational fuel source emerged from an unforeseen convergence of technology and readily available data.