(Steve here: This one became a monster, so I split it up. Part 1 provides context and a possible scenario about how artificial intelligence can play out. Part 2, out tomorrow, looks at questions institutions should ask and what this means for higher education. Part 3, which will come out next week, examines strategies that individuals and institutions can take today, and I share some of my concerns.)
Last week I saw the German opera Die Tote Stadt (The Dead City). Fancy, no?
Opera hasn’t changed much since it was initially created—live singers with live music on a stage in front of a live audience.
In a similar way, a student in ancient Greece, at the world’s first university in Morocco, and at the oldest university in Europe in Bologna, Italy, would recognize how we do classroom instruction today— teachers standing in front of students lecturing.
One of the greatest advances for those who don’t remember anything from their college foreign language classes (ahem) has been opera houses installing subtitle generators. That technology helped keep me more engaged.
Professors have gone from chalkboards to overhead projectors to PowerPoint presentations to help keep students with incredibly shrinking attention spans awake.
For about 400 years, the only way to experience an opera was to go to one. The first way to hear it remotely was in 1881 through a local telephone line that carried music live from the Paris Opera. Recorded music forever changed how and when you could enjoy the art form. The first full opera record was released in 1904—on 40 discs.
It is similar for higher education content and a degree. Yes, you could read the books, like you could read an opera libretto, but it wasn’t the same. You had to go to a college or university to get an education.
In 1892, the University of Chicago was the first traditional institution to allow students to earn college credit remotely through correspondence courses. Now, there are over 40,000 online degrees available, including many from “name brand” schools.
A lot of well-meaning professors, experts, and administrators poo-pooed the idea that you could get a quality education online. The University of California still insists that you can’t do a degree that way because they are the University of California, and they say so. (I can’t decide if this is hubris, circling the wagons to protect the status quo, or hanging on like professors insisting you couldn’t be considered educated if you didn’t study Greek and Latin.)
Research has shown that online courses are producing “similar student learning outcomes” as in-person courses. Other studies have seen mixed results, but the trend is moving toward consistent parity between the two.
As recording and streaming technology have forever shifted how and when people can access opera; the internet forever shifted how much time and energy students and researchers must spend gathering information; online education combined with artificial intelligence (A.I.) will forever impact traditional higher education in ways we cannot fully predict, but we can make some educated guesses.
A Scenario
Meet Professor Akin. He has a 101-second lecture on higher education and artificial intelligence that’s worth a quick watch.
I used ChatGPT, an A.I. program, to write the script. I feed that script into Synthesia.io to create an avatar to present the information.
It’s not perfect, but knowing how quickly technology evolves, what do you think that will look like in 3-5 years? How about 10?
That means that people will be able to enter a script, give emotional direction, and provide charts and images to go with the lecture—and you now have an online class. Artificial intelligence will likely be able to generate many of the images needed based on the script, further reducing human effort. (The header image of the university was generated with Stable Diffusion.)
Traditional institutions will stay away from this as they did (and many still do) with online education. But consider this scenario.
An underdeveloped country needs to improve the education level of its population quickly but doesn’t have the funds to build an extensive, traditional higher education system. (Or, a state or country wants to offer something new.)
They hire a small cadre of grossly underpaid and underemployed adjunct faculty with PhDs from respected institutions from around the globe to develop curricula for a new college. (Some would argue that AI will be able to create or co-create the curriculum at some point, but we will set that aside.)
Let’s assume they develop two majors, Computer Science and Teaching. If you have a strict curriculum with no electives, which would make sense at this stage of development, the content development should cost less than $250,000.
120 credit hours for a bachelor’s degree=40 classes in total. Add an extra 11 classes for the second major. Adjuncts average $3,500 per course in the United States. (51 classes x $3,500=$178,500.) Toss in lower labor costs from other countries, more time for some classes than others, and round up.
The adjuncts write the scripts on a work-for-hire basis (no copyright or royalties), use OER material, and provide them to the country’s higher education department. (Another option would be to grant royalties to the adjuncts on a per-student basis, which would be fairer.)
The country contracts with a firm to put the material into AI-generated professors, using the lessons learned from competency-based education, and creates courses designed specifically for mobile devices. ($2 million?)
A cloud computing company hosts the college and the country uses local government offices as proxy testing centers when students need to take a test to show completion of a course. ($250,000 per year?)
As A.I. presenters and materials improve, the country can upgrade using the scripts they already own. When the subject matter needs to be updated, they can hire new adjuncts on a class-by-class basis.
While there would be some administration to run the whole thing ($250,000-$500,00 per year?), there would be no admissions office, dorms, libraries, stadiums, sports teams, cafeterias, parking lots, deans, department chairs, alumni associations, faculty, or landscaping to pay for.
Annual tuition is $0. Perhaps there will a modest fee for testing.
Let’s say my math is wildly off (it’s just the back of napkin stuff, so yeah), and it would actually cost $10 million to get the whole thing off the ground and at least $1 million a year to run. That’s about the cost of one U.S. Abrams tank; the Bill and Melinda Gates foundation has that kind of cash; it’s a rounding error for the endowments of Harvard, Yale, Stanford, MIT, and others; and the University of Colorado Boulder had a $24 million cost overrun on a football stadium upgrade. The money can be found.
While many would argue this no-frills, Sprit Airlines' approach to college isn’t college at all, it would be a great option if your current option is nothing.
It’s like if you are an opera fan and your choice is a recording or no opera at all; you will pick the recording. And, if you are pressed for time and might not want to go to the opera, you will pick the recording.
Another thing to consider is how other technologies will work with A.I.
Meta has thrown away billions on their version of a metaverse that people don’t like, VR headsets remain clunky, and people thought we would all be wearing something like Google Glasses by now. So, we dismiss these technologies as more Silicone Valley claptrap.
But some versions of virtual and/or augmented reality will combine with AI to allow a more immersive and meaningful learning experience that will rival the mediocre professors that all campuses deal with.
What are the broader implications of A.I. for higher education? What questions should institutions be asking? Stay tuned for Part 2.
Come on now, surely we can find a different revenue source than giving up an Abrams tank.
Very thorough assessment of AI and how it can help higher education. And I think we will get there. It won't be a one size fits all as we all learn in different ways and we all have different expectations from our experience.