Robert Farley shared this funny story One city that outsourced school bus routes to a company said, „Artificial intelligence is used to generate routes in order to reduce the number of routes.'' Last year, JCPS had 730 routes, which has been reduced to 600 starting this year. . ” The result was reported as a “traffic accident.”
I'm not sure you can blame the AI here. . . Reducing the number of routes by more than 15% would be a big deal! Initial estimates suggest that the route will be over 15% longer, but this is just an average and some routes are sure to be much worse. No wonder bus drivers hate it.
Farley said: “In theory, developing bus route algorithms is entirely possible with AI… optimizing the incredibly difficult problem of getting thousands of children to more than 150 schools in a limited time frame. It’s for the sake of it.” However,
1. Effective problem solving in the real world requires feedback, but it is not clear that this system included feedback. A company may have signed a contract, executed a program, and sent the output to the school district without even reviewing it. Not to mention getting feedback from bus drivers and school administrators, the results made sense. How many people take the bus to work every day?
2. It appears that the goal was to reduce the number of routes, not to create functional routes. Optimizing element A can come at a significant cost to element B. Again, this is the reason to get feedback and solve problems iteratively.
3. Farley describes AI solutions as “high modernist thinking.” That's an interesting and insightful way to say it! We do not know what kind of „artificial intelligence“ was used in this bus operation program. This is an optimization problem, and the most important thing is usually deciding what to optimize, not how to optimize.
In that sense, I think the biggest problem with „AI“ here is not that it has led to bad solutions, but that algorithms that are not backed by feedback will fail when trying to optimize the wrong thing. But rather, it had a magical vibe that made people accept the results without question. “AI”, like “Bayesian”, could serve as a slogan to get people to stop being skeptical. They might have been better off saying they were using things like quantum computing and indoor conductor superconductors.
Perhaps the connection with „high modernist thinking“ is that (a) we can and should replace the old with the new, clearing out „slums“ and building clean, shiny new buildings, etc. The idea is that (b). Looking only at the surface is similar to how Theranos fooled many people by building fake machines that looked like clean Apple-branded devices. In this case, there is no reason to think that the bus routing program is a scam. Sounds like an optimization program plus some good marketing, but it's just the „AI“ providing a plausible cover story, and a poorly planned corporate/government contract he's one. was.