On "The Reverse Centaur's Guide to Life after AI" by Cory Doctorow ****

This book was lent to me by a coworker because he thought it might be useful for a class we're teaching together. Indeed, AI has changed a lot about our world and about teaching as well. I'm not sure exactly how this book specifically--or passages from it--would be useful in the class, but I'll be curious to hear my coworker's thoughts.

That said, this was a quick and breezy read that was very informative, and it being Doctorow, also provocative. Doctorow takes the stand that the AI buzz is primarily a bubble. I'd agree. I see a lot of parallels to what happened in the 1990s with the dot.com boom. Everything is being transformed by AI the way that everything was being transformed by the Internet. The only problem was that there was so much hype that even things that were not being transformed worked off this idea of the Internet transforming it. A few years after the bust, folks laughed at things like lobsters.com: your one-stop shop for lobsters delivered straight to your home. But even as those ideas gave way to reality, some of the more audacious ideas just needed a bit more time to come to fruition. At this point, you probably could get lobsters delivered to your home, though it would likely come through a marketplace like Amazon rather than a site devoted solely to lobsters. In this same way, we're looking for transformation by AI at things Doctorow says AI cannot do. I'm not sure he's right about that, though he's surely right about the speed with which some folks expect said transformation. Case in point would be his examples of architecture transformation. A group of architects laid off their draftsmen because AI can render pictures in a few minutes what it would take a draftsman days to do. Of course, if that's the case, as Doctorow points out, it won't be too long before people realize they don't need architects either, since people could go straight to the program to design a building. But what the AI cannot do is render blueprints based on those renderings, because it's based on algorithms regarding what pixel goes next to which; that the architect says that's just a step away is false, Doctorow notes. That may be true, at present, but I don't see why such an algorithmic program couldn't eventually be written to map blueprints to renderings, even if blueprints are much more complex and have little one-to-one correlation with renderings. They could be made to have such correlations. On the other hand, it isn't going to happen tomorrow. That's probably a decade or two away, as were many of the transformations made possible by the internet.

So in Doctorow's view, AI is not ultimately coming for most of our jobs. Companies who fire employees based on this will be woefully regretful, because the AI simply can't do what companies claim it can. On some level, this is true. But on another level, I'd say we will still need fewer people in some jobs. I think of the advent of the printing press. That put a lot of scribes out of work--but not all. Where, say, 167 scribes might have been needed to render 500 copies of a 100-page pamphlet in 30 days; 50 people could do that work in 3 days. But it would not have been cost or time effective to use a printer for just one or two copies of a 4-page document; scribes still had a function. Meanwhile, other jobs would have come into existence, like those who create the block type. And also more text would have been made available and prices would have dropped. Books would no longer be so scarce, and so the volume of text would make up to some extent for the jobs that would have otherwise been displaced. Further efficiencies in printing as centuries progressed would reduce work hours further, but each time with offsets in terms of other jobs and more volume.

Doctorow hates chatbots for customer service. Indeed, I'd agree that a company that gets rid of all customer service for AI is stupid if it wants a good reputation. But for basic queries, chatbots may actually have value, allowing a company to queue only truly unique queries to a human. (Of course, the whole point, in many cases, is crappy customer service, in order to avoid returns, which chatbots are perfect for.)

What's scary about what Doctorow writes is the bubble itself. One-third of the stock market's value is wrapped up in seven companies, all on the forefront of AI and the irrational exuberance around it. When the exuberance fades, the market will crash and so will a number of jobs and fortunes with it. It'll be like the dot.com bust or the housing bust all over again, but worse. Doctorow explains well how it is such companies come to have so much value (irrationally, I would say). It all has to do with P&E, profit and earnings. If you've got a taxi cab company that makes $200k profit on a million dollars of earnings per year, you're doing well, but if you do that every year, you don't really have much to offer to a stockholder beyond steady profit. If, by contrast, you're new tech, and you're promising taxis without drivers and ultimately a lot bigger profits, your loss this year on $200k of earnings is not enough to scare me away as a stockholder. It's the potential that's important. Next year, you make $400k in earnings and still show a loss, but those earnings keep going up each year. At some point, you might show a profit. In fact, you might someday show a huge profit. It's that potential that makes me want to invest. If eight years from now, you're making $5 million in earnings and $1 million in profits, that's a much better investment than the stable taxi company, especially if I got in early. Of course, if you're still taking a loss, if your earnings are still going up, maybe it'll still work out. Eventually, though, the tech company either goes under as investors grow tired of losses or it becomes stable, which means it's making a steady profit on a steady set of earnings, and there's no more grand potential. This explains how Ford's stock is cheap while Tesla's is through the roof, even though Ford makes many more vehicles and even makes electric vehicles. Ford is seen as stable rather than as having huge potential. This seems somewhat silly to me, as an investor, though, if one part of Ford's business does the same as Tesla's whole business.

I didn't care much for Doctorow's title (it seems too obscure), but the idea itself is worthy. Doctorow notes that most AI proponents have the usefulness of the technology backwards (though it helps them promote it, which is the whole idea). AI, they say, will do the work, and humans will just check it. This enables tons of many fewer workers. Doctorow notes how AI can help someone do a faster and better job, but the idea that humans become just the quality checkers is a bad one, like putting the brain in the AI (the horse head) and having the humans do the grunt work rather than the other way around. It's arguably less efficient, and it leads to workers/checkers not really knowing how to do the job and thus how to do the actual checking. If I never write code anymore, how will I know where the AI messed up its code? Writing code, at least the nonrepetitious part of it, is more interesting than reading it. Why give the good job to the AI?

But AI, as Doctorow notes, often doesn't even do much of what it's cracked up to do. He writes of the Turk, the medieval machine that could play chess but that turned out to be an illusion: a little human actually guided the machine's moves. Many times, that's what is the case with supposed AI. Some demonstrations of things like self-driving cars have actually had human engineers controlling the car remotely. Again, this is all part of the hoopla to drive stock prices.

This brings us to the role of "creatives." Doctorow notes that AI-created stuff carries no property rights. Some companies want this changed, but creatives should stand pat. AI is a great assistant. But the idea that AI can create instead of the person would render the person pointless. That's what big companies want. The creative jobs, which don't pay well anyway, go away, which renders questionable how much money AI actually saves. However, if there's no copyright for AI-rendered stuff, that makes it public domain, which means there's no profit to be made in, say, books created by AI. Anyone can take that book and reprint it as their own.