Steam tractors were very heavy so they used steel wheels which were narrow resulting in high ground pressure making steam tractors useful only in places with hard soils. The solution was the caterpillar continuous track but it appeared at a time when ICE already doomed small steam engines. An interesting what if is what if the caterpillar track was invented in the 1850's. Imagine Civil War pics with tracked steam vehicles pulling heavy artillery and supply wagons.
There's a range of occupations: farming grain versus truck farming; dairy farming, ranching; fishing, aquaculture, floriculture--etc. The tractor impacted grain and cotton production earlier and more greatly than the other areas.
I enjoyed this post, but would disagree that the longer term historical limiting factor for agriculture was power. The limiting factor has been productive or fertile land.
Steam was a hard system to use on small mobile systems. Needed stokers or more complex oil burners, plus a slow startup time. Gas engines and smaller tractors gave the required market share.
I've been binge watching old Perry Mason shows on Pluto. It surprising how many advertisements aimed at programmers are on (IBM Watson Ai and others). The IBM add shows how their Ai can reduce programmer loads (read job numbers). Programming has much more structure than other Ai market areas. It will be the first to see huge hits.
Interesting post -- I think there is also a question about labour costs (relatively low for much of the early period of tractors being available, thus disincentivising adoption) plus the fact that farmers needed time to buy up neighbours' land in order to have fields that were big enough to justify a tractor. All explored here... https://www.economist.com/christmas-specials/2023/12/20/a-short-history-of-tractors-in-english
I wonder if the analysis as applied to generative AI is entirely correct though.
one advantage of AI is freeing talented people up to do more high quality work, by taking away a lot of the drudgery that’s inherent in many “knowledge work” tasks. It’s not easy to imagine what would happen if you freed say the 20% smartest people in the world to do 6x more interesting work. It doesn’t exactly scale the same as farming productivity improvements.
Another big advantage of AI is in translating between medium - for example, taking a software program written in an obsolete language and moving it to a modern language. This is something that wasn’t economically practical before LLMs - it just wasn’t worth the expense of a talented persons time to go back and rewrite old software. But an LLM makes it economical and that potentially allows you to update old systems and move them,eg, to more efficient computer systems.
So you get more and better results from the most-capable people. And you get to improve things that weren’t economically capable of being improved without AI.
Plus knowledge work is a lot different than farming. It has the potential for a geometric sort of improvement, where insights in one domain spur innovation in another, and so on.
And there is no upper bound on what might be achieved with knowledge work - there’s diminishing returns with agriculture at some point, because there’s an upper limit on what you need to produce based on what people can consume. But that’s not the case with knowledge creation - the benefit is basically infinite up to the laws of physics and nature. If you can improve the efficiency of improvement in this area, you can keep going forever.
David, great piece! Steam technology was a “general purpose” technology, as I suspect that generative AI will ultimately prove to be.
Early steam-powered tractors, as you pointed out, were limited by their power density. The power density of internal combustion is simply higher because there was no “medium” (water) and the fuel itself was burned directly.
Certainly, ICE-powered tractors led to a surge in farm productivity after WW1 but don’t discount the influence of nitrogen fixation either: https://www.lianeon.org/p/bombs-and-fertilizer
AI, as you illustrate, has incredible potential, but it may take several complementary inventions to realize that potential.
Steam tractors were very heavy so they used steel wheels which were narrow resulting in high ground pressure making steam tractors useful only in places with hard soils. The solution was the caterpillar continuous track but it appeared at a time when ICE already doomed small steam engines. An interesting what if is what if the caterpillar track was invented in the 1850's. Imagine Civil War pics with tracked steam vehicles pulling heavy artillery and supply wagons.
There's a range of occupations: farming grain versus truck farming; dairy farming, ranching; fishing, aquaculture, floriculture--etc. The tractor impacted grain and cotton production earlier and more greatly than the other areas.
I enjoyed this post, but would disagree that the longer term historical limiting factor for agriculture was power. The limiting factor has been productive or fertile land.
Go Ahead, and Do the History of Tractors.
My late Father restored Antique Tractors and there are still several left on his property.
Steam was a hard system to use on small mobile systems. Needed stokers or more complex oil burners, plus a slow startup time. Gas engines and smaller tractors gave the required market share.
I've been binge watching old Perry Mason shows on Pluto. It surprising how many advertisements aimed at programmers are on (IBM Watson Ai and others). The IBM add shows how their Ai can reduce programmer loads (read job numbers). Programming has much more structure than other Ai market areas. It will be the first to see huge hits.
Interesting post -- I think there is also a question about labour costs (relatively low for much of the early period of tractors being available, thus disincentivising adoption) plus the fact that farmers needed time to buy up neighbours' land in order to have fields that were big enough to justify a tractor. All explored here... https://www.economist.com/christmas-specials/2023/12/20/a-short-history-of-tractors-in-english
Really interesting.
I wonder if the analysis as applied to generative AI is entirely correct though.
one advantage of AI is freeing talented people up to do more high quality work, by taking away a lot of the drudgery that’s inherent in many “knowledge work” tasks. It’s not easy to imagine what would happen if you freed say the 20% smartest people in the world to do 6x more interesting work. It doesn’t exactly scale the same as farming productivity improvements.
Another big advantage of AI is in translating between medium - for example, taking a software program written in an obsolete language and moving it to a modern language. This is something that wasn’t economically practical before LLMs - it just wasn’t worth the expense of a talented persons time to go back and rewrite old software. But an LLM makes it economical and that potentially allows you to update old systems and move them,eg, to more efficient computer systems.
So you get more and better results from the most-capable people. And you get to improve things that weren’t economically capable of being improved without AI.
Plus knowledge work is a lot different than farming. It has the potential for a geometric sort of improvement, where insights in one domain spur innovation in another, and so on.
And there is no upper bound on what might be achieved with knowledge work - there’s diminishing returns with agriculture at some point, because there’s an upper limit on what you need to produce based on what people can consume. But that’s not the case with knowledge creation - the benefit is basically infinite up to the laws of physics and nature. If you can improve the efficiency of improvement in this area, you can keep going forever.
Anyway good post. Very thought provoking!
David, great piece! Steam technology was a “general purpose” technology, as I suspect that generative AI will ultimately prove to be.
Early steam-powered tractors, as you pointed out, were limited by their power density. The power density of internal combustion is simply higher because there was no “medium” (water) and the fuel itself was burned directly.
Certainly, ICE-powered tractors led to a surge in farm productivity after WW1 but don’t discount the influence of nitrogen fixation either: https://www.lianeon.org/p/bombs-and-fertilizer
AI, as you illustrate, has incredible potential, but it may take several complementary inventions to realize that potential.
Interesting, thanks for sharing that link J.K.! Glad to hear this resonates with you.