A CS Professor on Why Slow Learning Wins in the AI Era | CU Boulder, Tom Yeh

Intro

00:00My name is Tom Yed. I'm a professor of computer science at the University of Colorado in Boulder. I'm also a founder of AI by hand is a global education initiative to make AI inside the black box accessible and approachable. Writing out all the maths with hand in doing so understanding that AI is not a big mystery [music] is something we all can understand.

00:21What is the purpose of learning? Having the answer doesn't mean you know it. People can buy degree by certificate. Do you have ownership of this particular idea core and foundational that doesn't change is evergreen? AI cannot change people but you can change AI.

Why I Teach AI by Hand- The Power of Learning "Slow"

00:48A transformer is to me to process individual words in your sentence. And I can started with say well at the token number four in my sentence and then just then this is a boxes I draw to show that well each token actually has multiple number in this case simple case that says three numbers.

01:05Unfortunately while I was a student I missed deep learning entirely. I was a bit too old. So I studied support vector machines traditional [music] machine learning method. So when then I became a professor all of a sudden all these people are doing deep learning.

01:20I have to learn deep learning all over again. So what I'm doing today with AI by hand is to share my learning journey. How I as a old professor trying to learn deep learning from scratch. All the things I'm sharing has been about to show my own struggle with understanding AI model math and algorithms.

01:41Only way I can get it is I [music] get to draw or write on the paper. This is what oh I actually get it. So when I share my drawing, shared my way to map out the math and a lot of people resonated and people started to comment out, hey, I really like your approach to breaking down by hand.

01:56I started, hey, maybe I just call it AI by hand. There's a reason that people resonate with this. Connect with this thing that I'm doing by hand. Why do we like to calculate this by hand when AI can do this very well? when I was teaching introduction to programming and I have a feedback like I going too fast.

02:21I'm going to s slides and so on and just [music] too fast because I really like to share a lot of my teaching and knowledge with my students. So I decided to teach the entire semester of C++ programming on the blackboard instead of doing live coding.

02:36So I decided to do that. So I have a whole semester of doing writing the I this my notes about I wrote it down on the piece of paper as a semester progressing I see a few benefits. One is that I can only go at the humanly possible speed of my writing.

02:51I cannot go any faster they can write. Second student can only learn at a humanly possible speed. They can only follow how much I write. And number three is that if I get my student to use their hand to copy my notes on a notebook, their hands are not on their keyboard checking their Instagram messages.

03:10So help with focus as well. The by hand is really about a way to connect back to our humans. So using your hand, you can go at a human speed. You commute in a human way. And over time I learned this. I started to see this values of going back to the old school by hand.

03:31What is the purpose of learning? Is it about that physical or digital artifact that prove they have learning or something you feel like you actually internalize you actually own this kind of [music] stuff? Do you have ownership of this particular idea?

03:43Well, AI give me an answer right away. Having the answer doesn't mean you know it people can buy degree by certificate. Whether you or not you own something, you value something is actually proportional to how much [music] time you spend um acquiring that piece of knowledge.

03:58You have to first define what learning actually means to you.

The Foundation That Doesn't Burn

04:08I remember when I was a undergrad, we're learning linear algebra as part of a requirement for getting a CS degree. We have learned linear algebra and we have learned max computation. I have no idea why it is even important. But then it turns out over time computer graphics became really popular because of Jurassic Park.

04:26They put this CGI up the front and people talk about hey everybody need to learn CGI and CGI uses a lot of matching application and then after a few years there was big data movement then it turns out you also need maximum application to do some sort of processing and then move to machine learning again forget about data science we should be machine learning specialist engineer again maximum application and today's AP is AI AI everybody need to be AI native we should raise our kids send into AI [music] M application in few years we all all we talk about is quantum computing is

05:02possible right and guess what m application again so you see this a trend that every time there's something that tool keep changing but it's always something that's core and foundational that doesn't change it's evergreen you could revisit a year from now two years from now it's still relevant people still care a lot about whereas the deepseek there was popular at the time but there's been a while now so deepse not as popular as before there's a new new thing for instance like this cloud super popular but let's see in two two months is it still popular we don't know

05:32but I'm pretty confident transformer topic is still going to be popular couple summers ago I had a opportunity to visit uh South Korea and I got to visit where everybody else will go the what's called Kukong there this palace that's that was just a view of the history very beautiful palace and then what just struck me when I learned about a bit more history the entire thing was burned down in like a 1500s except for the foundation that was made in Saudi rock.

06:03So in 1800s they rebuilt the entire palace based off the same foundation. Like to talk about this story because that reminds me of how this technology has been changing over and over again. But if you had a foundation on the maximum application, you could just apply it to AI.

06:20It doesn't really matter. We build your skill based on your solid foundation. [music] And so that's why I'm focusing on foundation because I believe there something you can rebuild. Doesn't really matter whether the new tools is obsolete. If you keep focus on the surface features, [music] the tools then forget about foundation.

06:38You just have to keep rebuilding your houses. You still never have your foundation up or can build a pump. So how can it apply to your own situation? Think about the way you grew up. Maybe your parents sent you to a soccer game or maybe the piano and you have something some skill you have become good at.

06:57Is it piano? Is it chess? And think about the way you acquire the skill and that is actually something that doesn't change. So as AI tool come every day the fact that you could acquire very difficult skill that is something that part of your identity that doesn't change.

07:14If you continue to focus [music] on that and you have ability to realize, hey, I could acquire this, I can learn this skill, I can become really good at it. That's when you could continue to apply that skill to a new AI tool. So, I'm a good example.

07:28I was falling behind on deep learning for quite a while. But I have learned skills really trying to break down difficult [music] topics by patiently writing everything down on paper. So that skill I able to eventually catch up. I caught up on deep learning by from a position that's way behind from people who actually been working on deep learning for a long time.

07:50So for you your piano skill, your soccer skill is not useless. That will be a skill to help you eventually once we figure out all this like crazy stuff and this one tool. You just skip this tool. I think it's absolutely fine. Who knows what are you going to next?

08:05But if you skip your next piano practice, you skip your next soccer practice [music] and you give up on that, that's not fine because that is going to be eventually a long run define who you are. But not this tool, not just one tool. At this moment in my career as a educator, I started to care more about not that they learned this math.

08:24A lot of times when I teach, oh, you showed up, you listen to me, you're trying to go through it, but I bet maybe a year from now, you don't remember anything. But what you can remember is more about you are able to understand this at the moment.

08:38You are willing to come here to understand foundation. You are willing to open up the black box. That willingness is what set you apart from others. [music] Others would never try, never attempted, never took out a challenge. That's why differentiate.

08:51It's not really about how much you remember the equation about [music] transformer about attention mechanism. It's really about there was once upon a time I try hard to memorize this. I stay in the library for hours of study. I was successful.

09:05So next time when there's a learning challenge I can learn this. So that is more important about what differentiate that people who know foundation they implies the process time the person invested in learning it. So that is what I value versus person who never really learn foundation implies the lack of effort the lack of willingness to invest in time and effort to take on a challenging learning task.

The 'AI-Native' Trap - AI can't change people, but people can change AI

09:32When I was teaching the intro to programming, you would really spend a lot of effort making new assignments a whole every semester because of check places that people share solutions and we have all these technical solutions. We're trying to check the IP addresses see whether people are accessing this and we even put some kind of a trap on the site.

09:53if somebody access that. We know they access this but at the time I was like okay well I hope Czech could get out of business maybe shut down by the government that they will help us educators and my dream came true because AI become a new treating tool and the check out the business [music] and then when I realized hey check was all the business my dream came true but problem is still there what's going on so it reminds me again we should keep thinking going back to the source there was the reason why people have to cheat at the first place.

10:27That is the main cause. It's not and check and AI just a symptoms. [music] So check is gone, people still cheat. I bet when AI is gone, people can still find way to cheat. Is this AI cheating that distract us from the bigger fundamental profit of the society's incentive system?

10:45Why [music] are student compelled to cheat? Why is it that this the system doesn't encourage real learning that has actually have spent time So when you hire somebody, what do I care about? Is does student have a good work ethics? What I care about?

11:01Is a student a good problem solver? Is this the person a team player that can actually communicate willing to work with others? [music] So when you hire somebody, we go back to those basic. That's what you actually care about. You want to keep those people as employees.

11:14And this AI thing is just going to be a byproduct of this. Think about it. When we hire someone because there's a problem solver in [music] order to solve problem that person is automatically just going to learn AI you don't have to tell them the reason why you had to force your AI native you had to go back maybe you didn't hire the right person you forgot to emphasize on person being a problem solver again similarly if you are hiring this person because this person team player because [music] the person is team player the person will learn how to use AI to facilitate team collaboration

11:49so you don't even have to tell the the person will automatically do that as well. Trust your instinct. [music] Continue to hire people like that because those people would automatically adopt AI. You're not a team player. AI not going to make you a team [music] player.

12:01You always look up your own interest. You do not respect others. AI not going to fix that. How can AI fix that? AI cannot change people only you. You [music] but you can change AI.