AI Will Send the Cost of Education to (Practically) Zero

What is the cost of an “adequate” education? The state pays approximately $3,800 per student, while the average district spends over $19,000. Several NH school districts are suing the state, alleging it fails to meet its Constitutional obligation.

But what if the cost of an adequate education is (essentially) zero?


  • AI will displace most classroom instruction.
  • It will take longer than it should only because of resistance to change and inertia.
  • It’s inevitable and unstoppable.
  • You’re probably experiencing cognitive dissonance reading this because the past 100 years have defined “education” as sitting at a desk in a classroom and being guided by a human. It’s hard to dislodge that paradigm but this post will give it a shot.

The Advent of the AI Teacher

NOTE: I’ll try to be careful to distinguish between instructing and teaching. Instructing is the process of delivering raw information. Teaching is the art (yes, it’s mostly art) of turning that information into meaning. Initially, AI will displace instruction. Eventually, I believe it will be superior to humans at both instruction and teaching.

I think that’s where we’re headed. You’ve likely heard the buzz recently about rapid artificial intelligence (AI) advancements. I asked ChatGPT-4 to create a sample homeschool lesson using the following prompt:

You are a 13-year-old girl named Susie. Instead of attending a traditional school, she is homeschooled by an AI teacher augmented by a VR headset. The AI teacher is a virtual avatar of Taylor Swift who is Susie’s favorite artist. Now, please create a one-paragraph summary of a sample lesson and exercise for each of the following six subjects: essay composition, algebra I, evolution, the drafting of the united states constitution, Mandarin Chinese vocabulary, and drama. Each lesson should be about 20 minutes and incorporate visual and audio components on the VR headset.

The following is the verbatim output it generated:

Susie puts on her VR headset and is transported to a virtual writing lab, where she is guided through the process of constructing a five-paragraph essay. Taylor Swift explains the importance of having a clear thesis statement, and Susie practices developing her own using visual prompts projected in the VR environment.

For Algebra I, Susie visits a virtual amusement park to explore real-life applications of linear equations. She calculates ticket prices, ride speeds, and wait times by solving various algebraic equations, with Taylor Swift providing step-by-step guidance and instant feedback.

In the evolution lesson, Susie embarks on a virtual journey through time, witnessing firsthand the process of natural selection and the development of diverse species. Taylor Swift highlights key events in evolutionary history and quizzes Susie on important concepts.

Next, for the lesson on the drafting of the United States Constitution, Susie is virtually transported to Independence Hall in 1787, where she observes the founding fathers debating and drafting the document. Taylor Swift helps Susie analyze the various arguments and understand the significance of the Constitution’s creation.

For the Mandarin Chinese vocabulary lesson, Susie visits a bustling virtual Chinese marketplace, where she learns and practices new vocabulary words by interacting with the environment, listening to native speakers, and engaging in conversation with Taylor Swift.

Finally, in the drama lesson, Susie is immersed in a virtual theater, where she practices her acting skills by performing in various scenes with famous characters from classic plays. Taylor Swift provides feedback on her performance, and Susie learns about the importance of body language, vocal projection, and character development.

It took me less than two minutes to type up that prompt and receive the output. You may think there’s still a big difference between those short descriptions and delivering the lessons themselves. It’s a reasonable assumption, but you’d be wrong. A recursive tool called AutoGPT could iterate through an entire curriculum outline and generate these lessons. Combine this with an AI video generation platform like Synthesia, and you’re done. A 15-year-old could knock this out over a weekend.

“But what about…”

I know this is going to be a lot for people to process and everyone’s initial reaction will be to search for reasons why it will fail or why it needs to be stopped. First, I don’t believe it can be stopped. Second, let me address a number of concerns and objections.

It’s Not Ready Yet

This is true. It’s not. I’m not suggesting the result this weekend would be remotely ready for prime time. But given the pace of advancement, I predict this will be a reality within five years, if not sooner. Here is how much the AI image generator Midjourney has progressed in just one year.

midjourney version comparison

Computers Can’t Replace People

If you think that a computer can never replace a human, consider this. A recent study asked patients to rate their online chat experience. Half were communicating with a doctor, and half with an AI bot. The patients overwhelmingly preferred interacting with AI, rating it more empathetic than human doctors by a factor of 10 to 1.

What About Socializing?

Yes, kids need socializing, but this is a common red herring argument against homeschooling. AI education will be done so efficiently that kids will have MORE time for socializing than they do in our current Prussian prison-styled education system. They will have plenty of time to participate in sports, drama clubs, band practice, etc. Which brings me to the next objection.

What About Working Parents?

The single biggest obstacle to education reform has been the fact that its primary function is daycare, and education is, at best, a distant second. The AI evolution won’t change that (assuming we still have jobs). For that reason, we will still need daycare services that serve as learning pods where kids can experience AI instruction while socializing with others. Our current classroom model, however, will not only be obsolete but counterproductive to the learning experience (as it is now).

The Post-classroom Dividend

The implications are mind-boggling. Educating kids at home at a negligible cost will deliver two major dividends.

Money Savings

NH schools spend more than $3.5B every year on education. That’s about $6,500 per household every year, whether you have kids or not. Imagine if that number shrank to $500. That would still give the government more than a quarter of a billion dollars to subsidize daycare, sports, music, etc. The savings will be enormous.

Time Savings

If you haven’t read Ian Underwood’s book Rethinking Fairness in Education, I highly recommend picking it up. You can read it in less than half an hour, and it is worth every second. In my post How to Fix Public Education in Three Easy Steps, I cover a lot of that territory. In this post, I want to revisit the concept of fairness in education.

What is “fair” for the taxpayer and the student? Certainly not our current system. To believe that our current system is fair, you would have to believe that every student requires exactly 129,600 minutes of classroom instruction to be adequately educated. Of course, that’s ludicrous, yet we all ignore that glaring design flaw in our system and never speak of it.

Nobody would disagree that kids learn at different rates. They even learn at different rates for different subjects. Therefore, keeping a student in a classroom for longer than is necessary for an adequate education is unfair to the taxpayer. And keeping a student in a classroom for less than is necessary for an adequate education is unfair to the student (and society).

education based on time is inherently unfair
Students don’t learn at the same pace. After 12 years, some students will excel and some will be left behind.

AI education fixes this. All students will complete their learning in as little or as much time as they require. Moreover, I predict that the experience will be so superior to our current disastrous methods that the average time required to receive the equivalent of today’s K-12 education will be somewhere around age 15. And I expect that many, if not most, will finish by age 13. The time dividend can be reinvested with more learning, socializing, and even working. The time surplus generated will be unfathomable.


Whether you think this is a good or bad thing doesn’t matter. It’s happening. It can’t be stopped. I’m sure the Education Industrial Complex will try to stop it. You’ll hear all the usual FUD (fear, uncertainty, and doubt). They’ll tell you AI can’t possibly replace humans. The problem is that it’s going to be too blatantly obvious. There will be plenty of parents using it to educate their kids and results and stories will be impossible to ignore.

The other problem is that the technology is too simple. According to a leaked memo, Google is terrified that their “moat” is gone and that AI is heading toward an open-source model, where Big Tech won’t be able to control and profit from it. You can run ChatGPT-4 on a MacBook laptop right now. It’s going to be too easy to develop and use this technology to keep it locked in a cage.

This is going to happen. It’s going to fix our abysmal attempts at public education, finally. The only questions are what do we do with the school buildings, and how do we spend all that money?

Sign Up for Updates

1 thought on “AI Will Send the Cost of Education to (Practically) Zero”

Leave a Comment