Edification vs. Enhancement—The Non-Transhumanist Vision of AI in Education

An essay on the wise use of technology in education where humans thrive as humans and not as technological freaks

This essay is adapted from the preface to a book I will be publishing later this year on the legacy of Jaime Escalante. The book is titled Defying Low Expectations: What Jaime Escalante Taught Us About Learning.

The story of Escalante’s success is widely remembered, at least in broad strokes: Escalante took a failing math program at East LA’s Garfield High, an inner city school whose students were poor and largely Hispanic, and transformed it into a math powerhouse that at its height in 1987 accounted for more than a quarter of all Hispanic students in the U.S. who passed the AP Calculus exam. Only a handful of high schools in the U.S. had more students pass the AP Calculus exam that year. But Escalante’s academic success at Garfield High didn’t stop with math. Across the school, academics improved at Garfield, and not just a little but dramatically. A rising tide lifts all boats.

Many technological developments have occurred since Escalante’s work at Garfield High in the 1980s. The internet didn’t become a thing until the 1990s. Charter schools also took off at that time, with the internet lowering barriers to entry and acting as an enabling force in their expansion. In the last decade, digital technologies for online education have proliferated. And then, beginning around 2021–22, artificial intelligence (AI), in the form of LLMs (large language models) such as ChatGPT, disrupted all aspects of education.

I want therefore to turn to technological developments that have impacted (or promise to impact) education, and what they mean in connection with Escalante’s legacy, which is a story of one human inspiring other humans to learn and shine. The short of it is that nothing that’s happened since the 1980s or in the last decade in any way invalidates Escalante’s legacy. If anything, subsequent technological developments strengthen his legacy. Let’s turn to a few of these developments, though in no particular order and without any pretense at completeness.

Let’s start with the biggest development—AI. AI is widely viewed as a threat to education, allowing students to cheat in ways unimaginable in the past, helping to write their papers, do their problem sets, and in general bypass the hard work of learning. All such problems for education, however, arise because teachers are not monitoring their students closely and making sure that they can do their academic work standing on their own feet. It’s a false dilemma to think that students will either cheat using AI or must be prevented from using it to learn successfully. The third option is to use AI as a way of honing students’ skills and knowledge, helping them learn more effectively than before.

By analogy, consider computer chess. In 1996, IBM’s Deep Blue defeated Garry Kasparov, who at the time was regarded as the strongest player ever to have played the game. Since then, computer chess has become far stronger than any of the world’s grandmasters. Yet human chess players today are the strongest they’ve ever been because they are able to leverage computer chess in their training. It would be one thing if players in their play constantly asked a chess program what their next move should be in a game. That would be using computer chess as a crutch. But except for machine versus machine chess tournaments, chess tournaments pit humans against humans and prohibit machines from interfering in the game. Consequently, human players can now use chess programs to make themselves better at playing other humans.

There’s no reason the same cannot be done using AI in education, especially with large language models such as ChatGPT, to facilitate the learning of academic subjects by students. What’s needed is to leverage AI in the training of students, and yet also to block AI in educational activities that require students to think on their own feet. This difference needs to be enforced because students tend to take the path of least resistance. Students need to be able to act without the prop of AI and know that they are being watched when they need to act without that prop.

Interestingly, leveraging AI in this way doesn’t require a full-fledged teacher, though it always helps to have one available. Once the AI is set up to deliver instruction through an LMS (learning management system—a technology that did not arise until the late 1990s), it mainly needs a monitor. A monitor needs to know a lot less than a full-fledged teacher. A monitor just needs to confirm that the student isn’t cheating and is answering questions correctly. The teacher, by contrast, needs to know the subject, set the lesson plan, and impart knowledge to students (whether directly or through the LMS). Armed with an answer key, a monitor can give tests to students, collect them in the allotted time, grade them, and submit the grades. The teacher, by contrast, is needed to construct the test and to specify the instructions needed to grade it.

Ben Carson, the renowned pediatric neurosurgeon for many years at Johns Hopkins, describes how his mother got him to read two books a week when he was young (for her motivation to get him reading books, see the Carson interview on this Substack). She herself had only a third grade education, and so was limited in what she could teach him. But she could ensure that her son spent time reading the books and then quiz him on their content, getting him to summarize and answer questions about it. Carson’s mother here acted as a monitor, not as a teacher. Yet she had a profound impact on Carson’s education, and he credits her insistence that he read books as the key to his success in life.

Carson’s mother took an old-school approach to his education. She did not, in the progressive tradition of Thomas Dewey, ask Carson to peer deeply into his heart to determine whether he really wanted to read books, and from there decide whether to become a reader. His mother did not give him that option. Initially, reading was for him a chore. But eventually he came to love it. And it made all the difference for him. The pain was worth the gain. That’s always the way it is with a sound education.

Artificial intelligence is promising to fundamentally transform education, leading to vast increases in efficiency of learning. Consider Alpha School. Founded in 2014 and refashioned under tech entrepreneur Joe Liemandt, Alpha School represents a radical experiment in K–12 education—an AI-driven, mastery-based model that compresses core academics into just two hours a day. Backed by $1 billion of Liemandt’s own funding, Alpha’s TimeBack system uses adaptive AI tutoring, personalized pacing, and focused time management to accelerate learning.

Alpha School students are said to advance an average of 2.6× faster than peers on nationally normed MAP (Measures of Academic Progress) tests and frequently score in the 99th percentile. The remaining hours focus on life skills such as entrepreneurship, leadership, and communication. Teachers, called “guides,” earn salaries starting around $100,000 and function as mentors rather than lecturers. Tuition runs from $40,000 to $75,000 annually—the high end being charged in places like San Francisco, where the cost of living is high.

Alpha is at once a luxury brand and a highly disruptive educational model. It vastly reduces the time spent on learning academic subjects. It completely restructures traditional schooling. And it justifies such disruptive changes by claiming stupendous increases in academic results. Obviously, its cost puts it out of reach of most students and their families. But even though early adopters are paying premium prices, this educational model may in time become affordable and widely available.

Actually, a model similar to Alpha’s that is affordable is already available. It’s not as deluxe as Alpha, focusing instead strictly on academics, and thus, at least for now, lacking Alpha’s life-skills amenities. Victoria Garmy, founder of StudiaNova.org, created this model (see the in-depth interview with her on this Substack). She draws inspiration from her upbringing in a one-room schoolhouse, her background as a mechanical and aeronautical engineer, and her passion for eliminating inefficiency in education. Applying an engineer’s precision to education, she has designed a “microschool” model that maximizes learning intensity while minimizing waste. Her approach reimagines the one-room schoolhouse for the digital age—lean, high-tech, and yet personal—allowing students to achieve mastery in a fraction of the time consumed by conventional schools.

In Garmy’s model, students attend for just three hours a day in concentrated learning sessions. During that time, they work independently at computer stations, moving through carefully sequenced online courses while a teacher-supervisor circulates to answer questions, provides one-on-one help, and monitors progress. Garmy’s school runs like a modern office. Like employees, students begin work by logging into their workstations. They then move through rich, interactive courses complete with videos, eliminating the need for live lectures or Zoom calls.

This “white-collar” classroom ensures that every minute serves a learning goal. The short day is enough, Garmy argues, because it eliminates non-instructional time: no cafeteria meals, no hallway transitions, no lengthy assemblies, and no distractions from unfocused instruction. The result is sustained intellectual engagement—three hours of genuine learning intensity, surpassing the diluted six- to eight-hour grind of traditional schools.

Her business model is as streamlined as her pedagogy. A single storefront equipped with twenty desks and repurposed Linux computers can accommodate forty students—twenty in the morning, twenty in the afternoon. With each paying $3,600 for nine months ($400 per month), annual revenue comes to $144,000. Studia Nova’s tuition cost is a small fraction of Alpha’s, making it widely affordable.

A single educator can, according to Garmy, operate such a self-sustaining one-room schoolhouse without grants or bureaucracy, making the model easily replicable nationwide. Garmy finds that total revenue of $144,000 is enough to cover rent, utilities, insurance, and an educator’s salary. The startup cost for laptops and desks is minimal. This is a lean operation. No cafeteria or food costs. No sports. No frills.

Outcomes have been striking: Students quickly reach grade level within a year, even those arriving multiple years behind. After that, the sky’s the limit. For students way behind in their reading level, reading while listening to the text being read combined with a writing exercise in which they give chapter summaries has proven remarkably successful at bringing readers up to speed. Three other features stand out:

1. the model’s adaptability, enabling personalized pacing and mastery learning;

2. its simplicity—Garmy has built a model any committed educator or parent can reproduce, restoring both efficiency and excellence to American K–12 education;

3. its incentivization of good behavior and engaged learning because students themselves much prefer a short three-hour work day versus the typical six to eight hours at a traditional school—they don’t want to jeopardize having to go back to the old model.

The compressed study time at Alpha and at Studia Nova contrasts with the long hours that Harvard economist Roland Fryer, in a recent Wall Street Journal article, found was important for learning in traditional public schools. KIPP (Knowledge Is Power Program), a network of charter public schools, also emphasizes long hours, adding two hours to the typical public-school workday.

Which of these approaches is better for a sound education—short hours or long hours? Long hours have the advantage of allowing more time for students to socialize, play sports, do extracurriculars, etc. Students who attend microschools (or homeschools) may be able to participate in these activities at a local public school while foregoing its academics, but arranging that will require some additional effort by students and parents.

The question now arises, How many of the long hours spent at traditional schools are spent doing intense concentrated mental work that delivers significant learning outcomes? In posing this question, we need to keep in mind that people can sustain intense, deliberate focus for only so many hours per day before cognitive fatigue sets in, leading to sharply diminishing returns in productivity and output. This limit arises from the need for recovery (periods of rest or diversion) to maintain high-level attention and performance, as excessive effort beyond this threshold triggers errors, reduced creativity, and burnout. It seems, then, that Alpha and Studia Nova have found a way to make the learning experience much more efficient than conventional schooling.

Victoria Garmy’s Studia Nova is attempting to spur an educational movement across the U.S. and beyond that builds and empowers microschools at a vast scale. For now, however, her efforts remain fledgling. Yet there is a thriving educational option whose schools are even smaller than microschools. Those are the homeschools. Homeschooling in the U.S., at the time of this writing, educates about 4 million K–12 students, or about 7 percent of the primary and secondary school population. Started in the late 1970s by persons dissatisfied with traditional schooling, the homeschooling movement grew from 10,000–15,000 students in the early 1980s to millions by the 2000s, achieving legalization in all states by 1993.

Homeschooled students often outperform public-school students on standardized tests, excelling in self-directed learning. Homeschooling leverages technology extensively, using LMSs (learning management systems) from their inception in the late 1990s, online curricula like Khan Academy, and AI tools like ChatGPT to personalize learning, enabling efficient, mastery-based education akin to models like Studia Nova’s three-hour daily sessions. On average, homeschoolers spend about 4 hours per day on focused academic work, varying by age and curriculum, with younger students often at the lower end closer to 2 hours and high schoolers closer to 6 hours. Like Alpha and Studia Nova, homeschooling avoids the inefficiencies of traditional schools’ longer hours, fostering high achievement while avoiding, through parental oversight, the pitfalls of AI misuse.

With the wise use to technology, students can learn much more efficiently than past generations of students, greatly accelerating their education. In particular, the wise use of technology avoids so much wasted time and busywork that burdened academic life in the past. Need a journal article? Find it instantly online as a pdf. Compare that to me finishing a book in 1998 and needing to spend half a day visiting a university library that had a journal article I needed, driving there, finding it in the stacks, photocopying it, driving back home, and then transcribing by hand any quotes from it. All this activity was necessary back then for me to finish the book. But it was not edifying—I did not become an intellectually more able scholar because of my library visit.

Or consider a baseball analogy. I used to coach baseball at the middle and high school levels. In a game, nine players from a team are listed in the lineup, and it’s necessary for each player to take a turn at bat before a player can bat again. Understood as a learning opportunity, each at bat is therefore only one ninth of the total at bats, and so a player is only learning one ninth of the time that players on his team are at bat. For learning to improve one’s game, this use to time is therefore highly inefficient. That’s why games are not a great way to hone a player’s batting skills, and separate practices are needed. Unfortunately, much of traditional K–12 education is inefficient in this way, focusing on the activity of one or a small group of students in a classroom while the rest of the class sits idly watching.

Besides improving the effectiveness and efficiency of learning traditional academic subjects, AI-driven technology promises to open entirely new frontiers in K–12 mastery, extending learning well beyond the traditional curriculum and enabling students to acquire high-level, once-rare skills. Consider the following possibilities:

1. Accent and Pronunciation Refinement in Language Learning
   Speech-recognition AI can analyze intonation, rhythm, and articulation down to the phoneme, offering individualized correction and visualized feedback, especially for non-native speakers who want to eliminate an accent. Students can practice repeatedly with voice models until their accent aligns with native fluency—an outcome rarely attempted, much less achieved, in traditional classrooms.

2. Creative Writing with Rhetorical Precision
   AI editors can train students to identify and employ rhetorical devices—analogy, parallelism, irony—and revise for tone, cadence, and argument strength. Instead of vague feedback, students get detailed stylistic analysis mapped to classical rhetorical methods. Classical schools would likely be early adopters.

3. Polyphonic Music Composition and Performance
   AI-guided keyboard and ear-training software can help students not just play Bach’s music but think contrapuntally. Thus they will learn voice-leading, harmony, and music theory not in the abstract but in real time as they compose or perform. Feedback systems can identify harmonic tension, suggest corrections, and let students iterate toward mastery of multi-voice textures.

4. Advanced Sight-Reading and Aural Skills for Musicians
   AI ear trainers can generate exercises dynamically, listening to a student’s singing or playing and giving instant corrective guidance. Learners can progress through increasingly complex rhythms, keys, and harmonic progressions—achieving conservatory-level fluency before college.

5. Scientific Experiment Simulation and Inquiry-Based Discovery
   Students can design and run virtual experiments in physics, chemistry, or biology with AI labs that model real-world variables and data. The system can prompt hypotheses, predict outcomes, and help interpret anomalies—cultivating genuine scientific reasoning. Anyone anywhere with an internet connection and a computer can thus become a budding scientist.

6. Mastery of Mathematical Intuition and Visualization
   AI-driven virtual-reality environments can help students visualize abstract mathematical concepts dynamically, letting students manipulate parameters and see underlying relationships. Over time, learners internalize mathematical structures through direct experiential interaction rather than rote formula use.

7. Fine Motor and Artistic Skill Development via Gesture Feedback
   Vision-based AI can analyze brushstrokes, pen pressure, or sculpting motions, giving immediate feedback on proportion, perspective, and technique. Students can thus master skills like calligraphy, drafting, or figure drawing once possible only under expert tutelage.

8. Debate and Argumentation Coaching
   Sufficiently advanced natural-language processing will be able to listen to and comprehend dialogue, assessing claims, evidence, and logical coherence in real time as students engage in debate. The AI coach can suggest stronger evidence, point out fallacies, and help students strengthen persuasion.

9. Emotional Intelligence and Empathy Training through Simulated Dialogue
   Conversational AI avatars can role-play complex emotional and social scenarios, letting students practice empathy, negotiation, and conflict resolution. Feedback can highlight emotional cues and alternative responses, cultivating social maturity rarely addressed in standard curricula.

10. Advanced Memory and Visualization Techniques
    AI tutors can train students to organize and recall large bodies of knowledge by turning abstract information into vivid, memorable stories or spatial patterns. Using adaptive prompts and personalized review schedules, learners can develop exceptional recall and deep understanding across subjects—from history dates to scientific classifications.

This last skill, if achievable through AI and becoming widespread among students, would be of both practical and historical significance. Before books and literacy became commonplace, cultures were largely oral, and a strong memory was a virtue. One concern about Gutenberg’s press is that human memory would weaken on account of it. That concern is, given the internet, greatly amplified, with so much of human knowledge now being instantly accessible, requiring very little of our capacity for recall. Use it or lose it—and people in developed nations have lost a lot of their memory capacity. AI can reverse this trend, empowering the memory of student learners, and thereby helping them to be better learners generally. Indeed, who wouldn’t find a powerful memory useful in their studies?

One last technology I want to consider takes its inspiration from the Oura Ring, a “smart” ring placed on the index, middle, or ring finger of a person’s non-dominant hand. The Oura Ring tracks sleep, activity, heart rate, temperature, etc. through advanced biosensors, offering detailed insights into recovery and overall well-being. It has since its launch in 2015 gained widespread adoption worldwide, embraced by athletes, biohackers, and wellness enthusiasts as a discreet, science-based tool for continuous health monitoring.

What if something as unobtrusive as an Oura Ring could track student learning? Right now, brain imaging through functional near-infrared spectroscopy (fNIRS) or electroencephalography (EEG) could be used to track whether learners are bored, engaged, or cognitively overloaded. But these measurement devices are obtrusive, impractical, and for now unaffordable for widespread use in a classroom. In any case, students would not want to wear them.

A suitable low-cost unobtrusive device, however, could monitor brain states favorable and unfavorable to learning. Such a device could enable real-time adaptation of instruction, such as adjusting pacing or content to improve focus and retention. Such a device should not be used for surveillance to enforce conformity. Rather, it should be used by teachers and students to monitor when attention is drifting and then to course-correct.

The goal of using technology in education—as sketched here and consistent with Escalante’s legacy—is humanistic, not transhumanistic. The goal is edification, not enhancement. It is to grow by learning things that make us fuller, more vibrant versions of ourselves. It is not to correct defects in ourselves and ultimately replace ourselves through technologies aimed at dissolving our humanity.

My prediction is that the humanistic vision will prevail, leading to far more actual human flourishing than the transhumanist vision, which is all about using technology to fix and upgrade humanity. The humanistic vision is natural, like promoting health through good diet, exercise, and proper rest. The other is artificial, like relying on pharmaceuticals to achieve wellness.

I close with some advice to parents: Edify your kids, don’t enhance them. We are organic beings, not gadgets to be improved with newer and better modules (like the newest computer chips). We are alive and need to strive for becoming fully alive. So far, we haven’t tapped into the full potential of our humanity. So why look to transhumanism, whose aim is to so transform humanity as to render it unrecognizable? Transhumanism tempts us with FOLO (fear of losing out): enhance your children or else they will get left behind, unable to attend Ivy League schools, second-class citizens in the wider culture. Yet edifying your children is not only proven to make them thrive but also known to do no harm. Neither can be claimed for transhumanism.

Postscript: There is much talk these days among journalists, talking heads, and podcasters about artificial intelligence (AI) turning into artificial general intelligence (AGI), where AGI so puts our cognitive faculties to shame that it will take over the world, supplanting us and ultimately ridding the world of humanity. This is a pipedream. Certainly model collapse of LLMs (see here and here) suggests that these systems face inherent limitations that will ever keep them from reaching AGI. But there is a deeper theoretical reason for this in the idea of conservation of information, which demonstrates mathematically that computational systems can never output the novel information needed to attain AGI. For the demonstration, see my recent monograph on this topic (much of which is accessible to the lay reader).