I’ve blogged multiple times on AI and ChatGPT, so you may be bored with the topic. Bear with me, because this time I’m going to look at what the real, long-term, threat of the technology could be. It may not be as simply scary as it’s popularly portrayed, but in the long term it could perpetuate and even cap a negative shift that information technology and “automation” started.
Think back to the late 1700s. Many of the really good jobs required considerable strength; blacksmiths, stone masons, and so forth. Others required special skills and dexterity, like cabinetmakers. Then along came machines, and suddenly almost anyone could produce the kinds of goods that specialized, high-strength, high-skilled, workers would have produced, and in much greater quantity. The industrial revolution was populizing in two dimensions. It made almost everyone a potential producer, and it made goods in such quantities that almost anyone could afford them.
Now think forward to the 1960s. Computer systems were still fairly rare, but many large companies had them. There was an elite of “programmers” who were suddenly high-value employees because they could tame these new monsters and bend them to the company will. It didn’t change everyone’s life, but it did create a new kind of elite. If you really, really, understood how to use computers, you were almost literally in a class by yourself. One big insurance company found that about one person in fifty could be taught to program well enough to be commercially useful.
Since that early computer age, we’ve been making computers faster, cheaper, more effective. A personal computer is now within reach for many, but not necessarily within their grasp. If you really know computers well, a PC empowers you to do things that someone without your skill could do only with considerably more time invested, if at all. Jobs that benefit from computer literacy turn out to be the best-paying because of the productivity multiplier computers offer. Computers also started running machines themselves, the first of what we think of as “robotics”. Automation, meaning the conversion of human tasks into tasks a computer-driven machine can do, started to put the kinds of people that the industrial revolution empowered out of work. And that is the risk of AI.
I do not believe that something like ChatGPT could, even in five years, equal a well-qualified expert in a given field. Remember my two friends, one in financial and other in tech media, who said essentially that they believe ChatGPT could do as well as some subordinates they knew? I agree, but not as well as my friends could do. AI is at the low end of a human capability scale in terms of answering questions. One could argue, as my friends seemed to be doing implicitly, that it was “average”. Well, so are most of us, by definition, and what happens if a lot of those “average” people in media or finance were replaced by AI? How do they earn a living?
I grew up in the Pittsburgh area, and when I was back there twenty years ago you could find both the rusting hulks of big manufacturing and mill operations, and also find some of the workers who’d been employed there. They had great jobs for their time, and I talked with some of the sons of those workers who expected to have the same kind of jobs when they grew up. They didn’t. Manufacturing has been automated and off-shored. A lot of the “blue collar” workers who made the US a powerhouse in the space have been automated away too.
Office workers, “white collar” workers, weren’t threatened…until AI came along. If ChatGPT can write a story on tech, will somebody start a media company that relies on that and doesn’t bother employing reporters? If ChatGPT can write basic financial analysis, do all the low-level positions in the financial hubs of the world get replaced by a browser window? Automation didn’t come for the white-collar types, but AI is coming for many of them.
We talk about job categories being lost as though we were talking about parking spaces. If there’s a limited number in one lot, you move on to another. Consider, though, what happens if AI can replace all those “average” people? Say half the total population, maybe more. Where’s the next parking lot that could handle all those looking for space? How does an economy, a politic, handle a situation where such a big chunk of the worker population can be replaced by AI? The industrial revolution was populizing. It made more people participants in the economy. The tech revolution is depopulizing; it’s making more people redundant.
I’m not qualified to talk about the morality of something like this; I’m sure others will give it a shot. I think I am qualified to say that a society cannot survive wholesale economic disenfranchising of a big chunk of its population.
Back when I was a junior programmer, the company that trained and employed me used to give programmers a special “any-time” lunch pass so we never had to wait in line to eat in the cafeteria. I used to see the look in the eye of those in line for the next time as I passed them to eat when I wanted. It wasn’t resentment, it was envy, because at that time most people thought that computers would level the playing field further, create the “second industrial revolution” with the same positives as the first. I wonder what those people would be thinking today, when they realize that the people going to the head of the line were working on a technology that could replace them completely?
A more relevant question comes to mind. Suppose that I’d been working on that super-AI concept? A concept that could have likely displaced many, most, or perhaps even all those people in the line for lunch? It would have spared me any hostile glares, but if we assume that other programmers in other companies were working on the same thing, then it raises a very important question. How would those displaced people get the money to buy insurance, or the products or services of those other companies? Without the products and services, could my insurance company employer have paid for the compute power needed for AI, and have paid me and my programming peers? A consumer isn’t just a body, its wages and spending form the basic element of the economy…except when it doesn’t, in which case it’s not a consumer any longer.
Most of the stories about the impact of ChatGPT and AI are stupid, click-bait stupid. Many who read it will never live to see the kind of impact that the stories predict, and some of those impacts will never happen. The real risk, the risk of creating mass economic redundancy in a world where class differences are already causing political polarization and warfare, is a long way off. But so are potential answers, resolutions. We need to create solutions at least as fast as we’re creating the problems, and I’m not convinced that’s happening.