For anyone unsure: Jevon’s Paradox is that when there’s more of a resource to consume, humans will consume more resource rather than make the gains to use the resource better.
Case in point: AI models could be written to be more efficient in token use (see DeepSeek), but instead AI companies just buy up all the GPUs and shove more compute in.
For the expansive bloat - same goes for phones. Our phones are orders of magnitude better than what they were 10 years ago, and now it’s loaded with bloat because the manufacturer thinks “Well, there’s more computer and memory. Let’s shove more bloat in there!”
Jevon’s Paradox is that when there’s more of a resource to consume, humans will consume more resource rather than make the gains to use the resource better.
More specifically, it’s when an improvement in efficiency cause the underlying resource to be used more, because the efficiency reduces cost and then using that resource becomes even more economically attractive.
So when factories got more efficient at using coal in the 19th century, England saw a huge increase in coal demand, despite using less coal for any given task.
Also Eli Whitney inventing the cotton gin to make extracting cotton less of a tedious and backbreaking process, which lead to a massive expansion in slavery plantations in the American South due to the increased output and profitability of the crop.
I always felt American car companies were a really good example of that back in the 60s-70s when enormously long vehicles with giant engines were the order of the day. Why not bigger? Why not stronger? It also acted as a symbol of American strength, which was being measured by raw power just like today lol.
This also reminds me of the way video game programmers in the late 70s/early 80s had such tight limitations to work within that you had to get creative if you wanted to make something stand out. Some very interesting stories from that era.
I also love to think about the tricks the programmer of Prince of Persia had employed to get the “shadow prince” to work…
Case in point: AI models could be written to be more efficient in token use
They are being written to be more efficient in inference, but the gains are being offset by trying to wring more capabilities out of the models by ballooning token use.
Costs have been dropping by a factor of 3 per year, but token use increased 40x over the same period. So while the efficiency is contributing a bit to the use, the use is exploding even faster.
For anyone unsure: Jevon’s Paradox is that when there’s more of a resource to consume, humans will consume more resource rather than make the gains to use the resource better.
Case in point: AI models could be written to be more efficient in token use (see DeepSeek), but instead AI companies just buy up all the GPUs and shove more compute in.
For the expansive bloat - same goes for phones. Our phones are orders of magnitude better than what they were 10 years ago, and now it’s loaded with bloat because the manufacturer thinks “Well, there’s more computer and memory. Let’s shove more bloat in there!”
More specifically, it’s when an improvement in efficiency cause the underlying resource to be used more, because the efficiency reduces cost and then using that resource becomes even more economically attractive.
So when factories got more efficient at using coal in the 19th century, England saw a huge increase in coal demand, despite using less coal for any given task.
Also Eli Whitney inventing the cotton gin to make extracting cotton less of a tedious and backbreaking process, which lead to a massive expansion in slavery plantations in the American South due to the increased output and profitability of the crop.
I always felt American car companies were a really good example of that back in the 60s-70s when enormously long vehicles with giant engines were the order of the day. Why not bigger? Why not stronger? It also acted as a symbol of American strength, which was being measured by raw power just like today lol.
This also reminds me of the way video game programmers in the late 70s/early 80s had such tight limitations to work within that you had to get creative if you wanted to make something stand out. Some very interesting stories from that era.
I also love to think about the tricks the programmer of Prince of Persia had employed to get the “shadow prince” to work…
https://www.youtube.com/watch?v=sw0VfmXKq54
They are being written to be more efficient in inference, but the gains are being offset by trying to wring more capabilities out of the models by ballooning token use.
Which is indeed a form of Jevon’s paradox
Costs have been dropping by a factor of 3 per year, but token use increased 40x over the same period. So while the efficiency is contributing a bit to the use, the use is exploding even faster.
I think we’re meaning the same thing.