Imitative AI Will Cause a Permanent Underclass -- Just Not How You Might Think

The Substack growth hackers would be so proud of that headline, God forgive me.

One of the articles of faith — and it is very much faith, given that the current economics of imitative AI, its actual output, or the history of technological change support the notion — is that imitative AI will create a permanent underclass. The drive to become a cyclops that outsources all its work and thinking to imitative AI is a sacrament of this new religion. If it’s adherents do not leverage themselves up into the rarefied air of imitative-AI native, their prophets tell them, then they will be permanently cast into the hell of the underclass, with no hope of ever ascending to the heaven of unlimited funds and power and immunity. And they may be right — but not for the reasons they want you to believe.

Using imitative AI is bad for your brain. This makes intuitive sense. My ability to multiply in my head has declined with the introduction of cheap calculators, much to my chagrin when playing 8 out of 10 Cats Does Countdown. My spelling, never good to begin with, and also an impediment while playing 8 out of 10 Cats Does Countdown, has become even worse as I have relied more and more on spellcheckers. You lose skills when you don’t use them — this is obvious. It is also backed by research.

Teachers — good teachers, at least — have seen from the start that the use of imitative AI is crushing their students’ ability to think and learn. Take the professor who, after suspiciously high grades in the take-home midterm, moved the final to an in-class exam. The final was, except for a handful of people, a disaster. Some didn’t even bother to show. But there are plenty of studies that show the deleterious long-term effects of using imitative AI. A recent study showed that imitative AI improved homework, but tanked in-person exam scores. A Microsoft study showed that adults who used imitative AI became, well, dumber. Another study showed that the use of imitative AI impaired conceptual understanding. I could go on, but the point is pretty clear: using imitative AI reduces your natural ability to think through problems.

So what if the use of imitative AI reduces thinking skills, though? Isn’t that what you just said happened to your ability to do multiplication? Yes and no. Yes, it is true that my multiplication memorization is not what it used to be. I cannot, in my head, do multiplication as fast as I used to. But I still completely grasp the concept of multiplication — what it is actually doing and what it means when I multiply quantities together. And I can still sit down with a pen and paper and work out multiplication problems. I haven’t lost anything but a bit of retrieval time to calculators. Imitative AI steals away your understanding; degrades your ability to reason. It would be as if the use of a calculator permanently altered my ability to work out multiplication problems on paper, or made me incapable of understanding when multiplication was useful. And that is how imitative AI could contribute to a permanent underclass.

The future does not belong to “AI natives.” Imitative AI has too many errors, too many situations where it needs to be overseen by a human being. Unlike a calculator, imitative AI does not always give you the perfectly correct answer. And since AI is just a regurgitation of its training data (yes, yes, I know. An oversimplification. But close enough for the broad discussion), it is not good at working out new ideas or adapting to new situations. But firms are already seeing that “AI natives” are poor thinkers, with limited and non-original ideas. They are turning from those people to people who have been trained in deeper thinking, without the help of imitative AI. And that lack of thinking will not be evenly distributed in the future.

Imitative AI firms are pushing it into education at alarming rates. They tout its benefits and claim that anyone who does not learn to use their tool will be left behind, unfit for any work that would pay a living wage. They have agreements with major universities and politicians willing to insist that education budgets include imitative AI tools so that the kids aren’t disadvantaged. But using imitative AI is what will harm kids. And the truly rich will ensure that the schools they send their children to make certain that their kids are taught the way we all used to be — painstakingly, with an emphasis on figuring things out ourselves. That way, they are prepared for the real world in a way that the kids who use the tools they produce cannot be.

The future does not belong to people who prompt. The future belongs to people who think. If you want a permanent underclass, and many of the people in control of these firms certainly seem to, then you would make sure that education for the masses consisted of lessons in prompting rather than lessons in thinking. Unfortunately, too many of our educational and political leaders are ensnared by the siren song of the prompt.