For quite a few years now there has been a current of opinion that schools should teach computer science as part of a general education.
This has now been joined (or perhaps even partially superceded) by the idea of teaching AI.
Does anyone know of such classes actually in use today? And if so, what is the curriculum in terms of topics and objectives?
Widening the question: do you think this is necessary and/or desirable? If so, what do you think the curriculum should look like?
Topics like AI are do not seem like part of a “General” education, more like a specialized group of topics that may be addressed after at least of couple of years of good background in fundamental mathematics and computer science.
I found websites for a couple of recent courses, for example a general class for all backgrounds https://www.futureofai.mit.edu/ and what looks like a more graduate-level class. how2ai-syllabus - Google Docs
In terms of technical details I would agree. But at a social level, one hears a lot of stories about people being led astray by misunderstanding or misusing LLMs.
Shouldn’t we perhaps try to at least get people to understand that these things are machines, not friends, gods or magic answer generators?
Or is that an overoptimistic view of humanity?
If AI goes the way its proponents hope it does, there’s a lot more that we’ll have to change in schools and workforces… it’s like the dawn of industrialization, but for our services economy.
And like DPRK said, being able to make and train AIs requires a background in at least moderate computer and math stuff. The US is so far behind in STEM that I doubt most undergrads would even be at that basic level, much less primary school. Asia, maybe.
If you’re talking about just teaching how to use (prompt) LLMs, eh, it moves so quickly I doubt schools and teachers can keep up. The kids will figure that out on their own much better than the adults can. Critical thinking, on the other hand, and the ability to evaluate factualness… that might very well go the way of cursive.
I expect students will learn about AI and use AI as a tool. They don’t generally need to be taught how it works or how to create AI tools.
Just like some students learn how to code, but all students learn how to use the internet and web searches. AI will be taught in school, just like students used to be taught how to use a slide rule. They will be taught how to use AI.
I’m of the mind, at least as of right now, that griping about AI is not all that much different than our parents griping about computers, then cell phones, then smart phones. Maybe not in it’s current or expected form, but I suspect, over the next generation or so, AI will be a normal part of life. With that in mind, it’s probably a good thing for kids to be learning about it.
Heh, many worse gods and religions have come and gone. It wouldn’t really take much for AI to beat your average cult leader.
People don’t want truth. They want someone else to do their hard thinking for them and provide easy answers. AI can do that better than most.
And, hopefully, how NOT to use it? But the illusion of conversing with an intelligence, if not a consciousness, is already eerily strong, even for a professional software engineer like me.
There are 193 undergraduate programs and 310 master’s programs in AI in the U.S. There are hundreds of high schools in the U.S. that have courses in AI. There are many places in the world that offer such programs and courses. In particular, China has just made AI a compulsory subject for all primary and secondary school students.
Interesting. Where did you get that data… is there a list?
I was more thinking though of the earlier school levels. What, and how, are they teaching there? Sadly of course there is probably a bit of ‘blind leading the blind’ syndrome there, since training of elementary school teachers won’t have covered this.
I put the sentence “How many universities in the U.S. have degrees in AI?” into Google. Try it yourself. See if you get the same answer. I also got a link to Best Colleges for Artificial Intelligence (AI) Majors — Lantern College Counseling .
To parody the little voice in ‘Life of Brian’: I don’t.
I’m happy to have a machine to do the gruntwork of library searches and the heavy lifting of tedious computations. But I’ll do the overall thinking, thank you.
Ah, OK. I wasn’t trying to be contentious, just looking for sources. Thanks.
It should be noted that that link uses the term “AI” in the same way as a data scientist, as an all encompassing term covering everything from K-Means to Multi-Arm Bandits, to computer vision, to language parsing, to customer profiling, etc. Not in the “AI” means an LLM sense. There is little value in learning how to “use AI” at a high level with that definition, so in reality we’re right back to learning advanced math, learning programming, learning analytics, etc.
So, back to the OP, are you asking about teaching people how to prompt LLMs or are you asking about what a data science curriculum would look like?
It’s just like sex education. Kids will figure out how to have sex on their own, but it is really useful to teach them about the dangers and responsibility. Kids will use AI without the schools teaching it, but it is good to warn them about its dangers.
As to how it works, except the most basic of understanding, no.
I’m curious about both. If courses exist, what are they teaching? And what should they be teaching?
On this part…
As a computer programmer, I think some foundational “epistemology of machine learning” kind of introduction would be helpful, if only to better differentiate between the deterministic outputs of traditional computing vs the probabilistic outputs of LLMs. It’s a pretty fundamental change in the way we use computers as a tool, and has not only practical applications (how to prompt, modify temperature, fine tune, minimize hallucinations, etc.) but also philosophical and ethical ones (not just “is it conscious” but more fundamental concerns like how does it acquire new knowledge, how it approaches things it doesn’t know, how its memories and contexts work), etc.
You need quite a lot of math and software to actually be able to produce such a model, but a layman’s overview of the basic flow doesn’t need to be advanced and would still do a lot of good, IMO. Similar to how you don’t really need to be a rocket scientist and able to calculate trajectories just to understand that objects in space orbit each other and cause all sorts of earthly phenomena from tides to seasons.
I don’t think “how to make AI” should be a required part of the curricula. For that matter I didn’t agree with the idea of computer science/programming should be taught to everyone, either. It’s just another trade skill and absolutely isn’t right for everyone. It would be a hugely boring waste of time for many people.
I’d rather we teach more humanities and things like civics and home life and basic finance and scientific literacy, instead, all of which schools aren’t great at.
Having a bunch of unemployed engineers doesn’t really prepare our society for the future, it just causes a race to the bottom in the labor market as technology and capital continue to consolidate. Having good citizens actually able to steer society and government, on the other hand, might better prepare us for steering the ship as needed.
Back in the real world, though, we’d probably get neither, and just end up with the poor librarian teaching “how to AI” in the 20 minutes after “how to Google”. Our educational system is optimized for spitting out elites at the top, not so much bringing up the bottom at the other end.
Absolutely. This is a specialized skill, it’s not anything most people will need in everyday life, just as not everyone can be expected to be an expert car mechanic or plumber. If nothing else, life’s too short: there’s a rule of thumb that you have to practice a craft for at least 10,000 hours before you can consider yourself close to being an expert.
I guess that is historically true. There is a spread of ability which I think is genetic. We have to accept that some people are born rather disadvantaged, and there is not much we can do about that except take care of them. But for people of average intelligence: good education can (or should) make an enormous difference,
Courses exist and seem to often be at the Masters level. I have no clue how good they are today, but when they first started showing up, it seemed a bit thrown together to take advantage of the interest at the time of becoming a data scientist (and the popularity was often salary-based, unfortunately). Sort of like a stretched out bootcamp that covers a wide array of the basics, but without the depth in core areas like math. Most didn’t require Linear Algebra for example.
From a hiring standpoint, I stuck with PhD graduates in Physics and Computer Science mostly. It’s easy enough to teach them (or just let them learn organically) any skills which they hadn’t picked up while getting their degree. Neither Python, nor SQL, for example, are particularly tricky to learn.
That is one of those things that I never grokked in undergraduate days. But 3blue1brown has some excellent classes on YouTube. Once you realize that a matrix is simply an operator on a vector, it all makes sense. But we digress….