Last August, I casually said that AI can’t design a car without first seeing the car, referencing Henry Ford’s incredible words: “If I had asked people what they wanted, they would have said faster horses.”
I will not back down from anything. But the history of technology is always richer than we imagine. Daimler and Benz are credited with the first automobile, but we forget that the “steam engine welded to a tricycle” was invented over 100 years earlier, in 1769. The history of the assembly line probably dates back to the 12th century AD. The more you unravel the history, the more interesting it becomes. That’s what I want to do. Unpack it and ask what would have happened if the inventor had access to AI.
Learn faster. Take a deeper dive. Look further.
What would an AI have told Nicolas-Joseph Cugnot, the man who created a device for transporting artillery over roads by welding a steam engine to a giant tricycle? Would you suggest this combination? Maybe not. Maybe you realized that was a bad idea. Ultimately, the prototype car could only travel 2.25 miles per hour for 15 minutes at a time. A team of horses will do a better job. But there was something about this idea that remained, even though it seemed to have disappeared.
At the end of the 19th century, Daimler and Benz created the first machines generally recognized as automobiles: high-speed internal combustion engines, four-stroke engines, two-cylinder engines, two-cylinder engines, pivot steering, differentials, and even transmissions. Some of these innovations have appeared before. Planetary gears date back to the Greek Antikythera mechanism. Double pivot steering (placing a joint on the wheel instead of turning the entire axle) came and went twice in the 19th century. Karl Benz rediscovered it in a trade journal. This difference dates back to at least 1827, but probably also appears in Antikythera. We can learn a lot from this. It’s easy to think in terms of single innovations and innovators, but it’s not that simple. Early Daimler-Benz cars combined many new technologies and repurposed many old technologies in unexpected ways.
Could a hypothetical AI have helped with these inventions? It might have been possible to resurrect dual pivot steering in “winter steering.” It’s something we’ve done before and we can do it again. But for that to happen, Daimler and Mercedes need to get the prompts right. Could AI have invented a primitive transmission if watchmakers had known about planetary gears? Again, sending the message like it is now is probably going to be the hard part. But the important question was not, “How can we build a better steering system?” But “What does it take to build a practical car?” And they would have to present that prompt without “car,” “horseless carriage,” or the German equivalent. Because these words are just emerging.
Now, let’s spend the next 20 years looking at the Model T and Henry Ford’s famous quote, “If I had asked people what they wanted, they would have said faster horses” (whether or not they actually said it). ? And what does it mean? In Ford’s time, the automobile itself already existed. Some of them still looked like horse-drawn carriages with engines. Others looked noticeably modern cars. They were faster than horses. So Ford didn’t invent cars or faster horses. But we all know that.
What did he invent that people didn’t know they wanted? The first Daimler-Benz automobile (still in modified buggy form) predated the Model T by 23 years. The price was $1,000. That was a lot of money in 1885. The Model T appeared in 1908. The price was around $850, with competitors costing much more ($2,000-$3,000). And a few years later (in 1913), when Ford’s assembly lines rolled into production, he was able to lower the price even further, eventually lowering it to $260 in 1925. This is the answer. What people didn’t know they wanted, but what they wanted was a car they could afford. Cars are firmly established as a luxury item. People may have known they wanted it, but they didn’t know they could ask for it. They didn’t know it could be cheap.
This is what Henry Ford invented. It’s economics. Assembly lines did not first appear in the early 12th century, when Venetian armories built ships by lining them up on canals and moving them downstream as each step of manufacturing was completed. The same goes for the automobile assembly line that Olds used (and patented) in 1901. Ford’s innovation was to produce affordable automobiles on a scale previously unimaginable. When Ford’s assembly line went into production in 1913, the time it took to produce a Model T was reduced from 13 hours to about 90 minutes. But what matters is not the time it takes to build a car. That’s the speed at which it can be produced. A Model T can roll off the assembly line every three minutes. It’s scale. Ford’s “any color as long as it’s black” did not reflect a need to reduce options or cut costs. Black paint dries faster than any other color, which helped optimize assembly line speed and maximize scale.
Of course, the assembly line was not the only innovation. Spare parts for the Model T were readily available, and the car could be repaired using tools that most people at the time already had. Engines and other critical subassemblies are significantly simplified and more reliable than competitors. The materials have also become better. The Model T used vanadium steel, which was quite unusual for the early 20th century.
But I was careful not to give Ford credit for these innovations. He deserves credit for the biggest picture: economics and scale. Charles Sorenson, one of Ford’s assistant managers, said: “Henry Ford is generally considered the father of mass production. He wasn’t like that. “He was the sponsor.”1 Ford deserves credit for understanding what people really want and coming up with a solution to the problem. He deserves credit for realizing that the problem was cost and scale and could be solved through the assembly line. He deserves credit for putting together the team that did all the engineering on the assembly line and the car itself.
Now it’s time to ask questions. If AI had existed before 1913, when the assembly line was designed (and before 1908, when the Model T was designed), would it have been able to answer Ford’s hypothetical question of what people want? The answer must be “no.” I’m sure Ford’s engineers could have put modern AI to tremendous use in part design, process design, and workflow optimization along the line. Most of the technologies had already been invented and some were well known. “How can we improve carburetor design?” This is a question that AI can easily answer.
But the biggest question isn’t what do people really want? I don’t think AI can look at the American public and say, “People want affordable cars. To do that, we need to build cars at a price and scale that is currently unimaginable.” Language models are built based on all the text they can scrape together, and in many ways their output represents a statistical average. I am confident that a 1900s language model would be able to access a lot of information about horse care, including horse care, disease, diet, and performance. There will be a lot of information about trains and trams, the latter often being powered by horse power. Mostly high-end publications will have information about cars. And I imagine there will be a feeling of “I wish I could afford one” among the emerging middle class (especially if we allow virtual blogs along with virtual AI). But if you were to ask a hypothetical AI what people want for personal transportation, the answer would be horses. Generative AI predicts the most likely response, not the most innovative, visionary, or insightful response. It’s amazing what it can do. But we must also recognize its limitations.
What does innovation mean? This certainly includes combining existing ideas in unexpected ways. This includes reviving good ideas that never made it into the mainstream. But the most important innovations do not follow these patterns or add to them. This involves taking a step back and looking at the problem from a broader perspective. That means looking at transportation and realizing that people need, for want of a better word, affordable cars at scale. Ford might have done that. Steve Jobs did that both when he founded Apple and when he revived it. Generative AI can’t do that, at least not yet.
footnote
- Sorensen, Charles E. & Williamson, Samuel T. (1956). 40 years with Ford. New York: Norton, p. 116.
