AI and the future of education is part of the timeline of automation. It will be affected by it, and negatively so, as a continual downward trend.
The future of higher education
The value of college and its uncertain future returns
The promise of college is that it is an investment. People expect to attain a higher net wealth by paying for college than if they didn’t. Once that possibility of return disappears, people will stop going. Some people will be caught in the middle of this transition. Where they invested in a white collar education but do not see returns through a job after graduation. Hopefully that’s not my family. But could my little brother. It’s important to think about money and the future.
I wonder if technology-involved parents are already not sending their kids to college because of the possibility of negative return. Maybe just have them run family business. Some families already might do that though who own businesses.
Families should seriously be considering whether it is an investment they should make at this time if they have children about to enter college. This decision should be made according to what you are of course specializing in, as some careers are more likely to be automated sooner than others, however one needs to think in a 8+ year timespan ahead. 4 years of college, 4+ years to recoup those costs / make as much as if you didn’t attend college. By that time, there is a possibility a lot of careers may be automated.
When thinking about automation timelines by types of careers, many people jump to human-oriented jobs like engineering technology or health careers like nursing. Be wary of this when evaluating career pathways. The progress of robotics between 2020 and 2025 has been astounding. Robots in 2033 might be remarkable and have genuine effects on these industries.
Finally, we should consider the fact that entry-level jobs (the ones college leads to) will be the first to be eliminated. Why? Because if you are a company this is your perspective: why invest in a new employee when you could invest in AI and gain a much greater increase in productivity. A new white-collar employee is a $250,000-a-year investment. Why invest in an employee that will only improve your productivity 1 / n and only be able to do basic tasks, when many AI agents are automating those basic tasks, and you can pay 250k for such agents for 200-500 existing employees in your company, resulting in a potential 1.1-10x productivity boost for them? The returns on option #2 are obvious! Invest in the capital that will give you the best returns today. Even if existing employees retire, the enhanced productivity boost given to current employees, as well as greater access to generational knowledge through LLMs, allows companies to continue achieve growth without having to worry about replacements. And everyone is on the hype train. They believe jobs won’t exist by 2075, so developing generational talent is not a concern.
What we’ll see is the available level of jobs and capital move upward. Some people will try to chase it by getting advanced degrees. AI progress will likely outpace them (consider this if you are/going to get an advanced degree!). What we will likely see is entry level job postings, fillings, and expected continue to decrease, just as they are right now, in one industry after another. People will see these trends, stop entering these programs, and colleges will shut them down – just like they did with clerical, drafting, and telegraph operator jobs. I predict we could see major AI-affected degree programs like computer science shut down by 2033. All other major college programs by 2045.
Note: Maybe these programs could evolve / change. Who knows.
How college will change
I hypothesize college will remain the same functionally despite AI advancements. Students will be assigned homework to practice skills, and then demonstrate such skills in exams. Lectures will still be the primary teaching method as it acts as a structured time to learn the material and gives students an opportunity to ask questions/engage. An issue I’m seeing as a student arises when current professors are not holding students accountable for exams which can be easily be cheated on. Colleges must adapt to this in the age of AI.
College professors right now have discretion on whether to issue exams online or in-person. At Purdue, we have online students, so some professors choose to have both the online and in-person students take an online, take-home exam. These exams are trivial to cheat on. Professors know this and typically employ a “trust” in the class, or evaluation based on other components of the course like homework. An issue arises when all of the course is completable by AI. Then the evaluation of the student’s abilities is impossible. Right now this is already happening in some courses due to this policy, and it’s degrading the value of a Purdue degree. Employers should be weary, but it’s not like they’re hiring for entry level positions anyway that much, right 🙂 Purdue is kind of milking their reputation to push these online-only masters programs to make them money, but produce students of questionable value.
An additional improvement I would make is more frequent examinations. Yes, students can have homework and cheat on it, but at least give them more opportunities to be evaluated and get a grade so they’re not stuck with a potentially horrible final grade at the end of the semester just because there were only 2 exams and they didn’t do any of the homework themselves. Exams act as good times for students to reflect where they are at and give them motivation to do better in the class if they are behind.
With less students attending programs, universities may struggle to fund the salaries of existing faculty. How this plays out with external research funding, grad students, etc. requires more research.
The future of lower education
Thinking into the far future, if pursuing higher education is no longer an attractive goal, the value of lower education becomes questionable as well, since the whole purpose of completing lower education for many people is to then attain higher education. This leads to quite the interesting thought experiment, bringing into question what and how we want to teach our kids, and whether it should look anything like how school works today.
School should be where humans get together and learn from personalized AI tutors in a controlled environment. They should discuss things, maybe do short writings, etc. about current work, politics, and history. Below are some quick thoughts about what I think we should teach:
- Learning to speak, read, write, and type will be considered the core and maybe only skills you need in a post-work world.
- I’m a big advocate of history so learning about that should be part of the curriculum too. I think adults today should have mandatory yearly history lessons, just to keep them refreshed about all the things they shouldn’t be doing in their careers / politics that past humans have made mistakes on (ahem ahem).
- I guess prompt engineering is a skill, basic math maybe? There’s lots of other things you can work on.
Figuring out where to place all this condensed learning within the timeline of human development is a task that needs to be done. Society will likely decide on less school per day and more play. But at least some form of school, and have it continuing through adulthood (history).
Last challenge is how to ensure learning because they aren’t applying those skills, they’re not learning. I guess quizzes and exams can be sprinkled in as we have now.
Some other considerations are how we want a child’s day to look like, such as, do we want school to be forced social time? Should it be without electronic devices?
If school finishes early for people and there’s no work, what do we do? We won’t know what to do with dumb kids who probably know how to talk and read, but do nothing but watch videos all day. Forced healthy activity maybe like solving puzzles, playing sports, who knows could be an option I guess.
More thought needed.
Current automation opportunities
- Robot instructors. LLMs have the ability to create presentations now, generate scripts for those, and do realtime text to speech. They have vision capabilities to detect raised hands and can stop in the middle of lectures to answer questions. All it takes is a camera pointed at the whole class and a speaker. Gosh, you could even setup a humanoid for funzies. Unis would like this because professors could dedicate more of their time to research. The cost per student would also come down. This is an example of a system that’s already possible. It just takes someone to build it. Good project.
- We could make a Piazza bot. Automatically responds to questions based on course materials which are taken from as a source, and it’s own knowledge. It may struggle at answering course-specific questions like what a professor knows, but it will at least be a good tutor. Of course, once people see bots can answer these questions, they will likely use bots on their own use instead of using Piazza. This is similar to NotebookLM.
- A Brightspace bot would be amazing too! I’m sure LMSs already are thinking/built integrations like this.
How I feel about AI and education
Unmotivated. Pointless. Sad. has ai ruined the purpose of life?
At college I just watch people in the library learning how to solve problems…
…that LLMs already know how to solve.
It’s like watching someone someone do long division on a large numbers over and over and over again, when a calculator is sitting right next to them.
Its… triggering… bothersome…
These problems they’re learning about, will not be solved by them in their jobs. They will be using LLMs. When you think about this, it’s almost as if they should be learning prompt engineering, or at least the higher concepts instead. Yes, we don’t teach people knowledge just so that they can use it in their job, we teach to make them smarter. But when LLM’s already provide an even greater level of intelligence (at least on a context-window basis as of now lol) then what’s the point?
I’m a big advocate of self-motivated, exploratory, as-needed learning. Put people in jobs immediately after high school. If they need to learn something to complete a new task, let them choose how to learn it, whether from an LLM, online resource, or coworker. If we do this, I believe as a society we will advance much quicker. Higher education has become what appears to be wasted time. If it’s not making our intelligence competitive or specialized, then we should spend our time efficiently learning how to use AI as a tool instead, and specialize learning how to use it in our career/area of specialization.
Place people jobs or encourage startups after high school. No more college. The LLM is your 21st century white collar worker and your new hire is the manager of your company’s newest division of LLM employees. Realize and exploit this.
Now this might be radical, and in the current iteration of this page it might sound like BS (so don’t think I’ll think this way forever). I realize that AI still struggles in applying its knowledge in real world situations means that we still as humans need to understand these things, many of which we learn about in college, however, I think there is a case for truly cutting down on a large portion of modern higher education’s curriculum, and instead focusing on the topics that matter. And the time to do this is now, while we all still have the possibility of attaining a job.
Altogether
With all things AI, thinking we need to consider how things will be implemented at a global scale too. AI is affecting the education of people who aren’t even in school, such as those who can’t afford higher education. They are teaching themselves skills and using AI as a tool to complete useful work. How AI will impact education, lifestyles, and the future of the economy in middle to low income countries is something to consider and something this post could expand upon.
Finally, a post-work economy all hinges on us being able to create safe, controllable AI, so none of this matters unless we can overcome that obstacle (which will be due likely sooner or close to a fully post-work economy).