Video / Short
The AI Paradox No One's Talking About
Pete Cooper explains Jevons' paradox, the 1865 observation that more efficient steam engines burned more coal rather than less, and applies it to AI. It is a short, clear answer to why falling AI costs are pushing resource use up instead of down.
Video summary
Pete Cooper introduces Jevons' paradox and why he thinks it matters now. In 1865, he explains, William Stanley Jevons noticed that even though steam engines were becoming more efficient and individually using less coal, they were collectively burning more of it — because as they became more efficient, more use cases opened up, and more steam engines meant more coal.
He maps the same pattern onto AI. AI is getting more efficient, and far more can be done for far fewer dollars, but AI is also doing many more things with much more capability. That expansion consumes more chips, more data-centre warehouses, more electricity, more cooling and more resources as more and more uses are found.
He closes by addressing the name: it is not really a paradox so much as a counter-intuitive result. You would expect greater efficiency to reduce consumption of the underlying resource, but in practice the opposite happens, because you end up using it more.
Video transcript
Read the spoken content without loading the YouTube player.
Good day. Jevons' paradox, what is it, and why is it relevant today? Well, in 1865, William Stanley Jevons noticed that even though steam engines were becoming more efficient, and in themselves using less coal, they were actually using more coal. That's because as they become more efficient, they found more use cases, more efficient use cases. More steam engines meant more coal was being burnt. Why is that relevant today? Well, we see the use of AI increasing. Now, AI is getting more efficient. We can do so much more with AI for so many less dollars. But, of course, it's obvious that AI is doing so many more things, so much more capability.
And of course, that uses more chips, more data center warehouses, more electricity, more cooling, more resources, as we find more and more uses for AI. Now, why isn't Jevons' paradox a paradox? Well, it's really just a counter-intuitive thing. You would think that as something became more efficient, you would use less of the resource that it depended on. Well, in actual fact, the opposite is true, because you end up using it more, just as Jevons outlined in his paradox.