It's Never Too Late to Learn to Code: Insights from Mehran Sahami

Hi, I'm Maron Sahami, a professor here at Stanford University. And today we're going to talk about some controversial takes on AI and other topics in computer science that we've seen on social media. I keep hearing people talk about AI. Do they mean chat GPT? Way too afraid to ask that at this point. So, what I would say is that chat GPT is a form of AI called the large language model. And it's probably one of the forms of AI that people are most uh used to cuz it's one of the things that's most available to consumers. But in fact, a large language model is just one form of AI. When you send your email, for example, there's other forms of AI that are checking email that gets sent to you for spam. So there's actually lots of AI out there, and chat GPD is just one form of it. Anthropic just said that 90% of code is going to be written by AI. What's the point of even having CS classes anymore? Well, there's lots of reasons to have CS classes. One of the big reasons to have CS classes is to understand the notion of algorithmic thinking. How to take a big problem and break it up into smaller pieces. How to then solve those smaller pieces in a way that you're come together to solve the big problem overall. It's also a particular mindset that when you think algorithmically, you think step by step. You think systematically. So, the reason why we teach people to code isn't just because we're trying to train an army of programmers. That's like saying we teach people about physics because we need everyone to study black holes. Really, we get people to study physics because they need to understand how the physical world works. And we teach coding so that people understand something about abstraction and problem solving, whether or not they're actually doing programming as their job or not. Programming languages change all the time. Why should I learn Python? Well, let me tell you why you should learn Python. Python's a great language for all kinds of applications. It's easy to learn. It's fast to use. And it's used for all kinds of things like data science and artificial intelligence because there's a large body of existing code in Python already that lets you build things very quickly. So part of the reason for learning Python is it's actually a great language to start with, but over time it'll also be a great language to build bigger and bigger things with. If I have access to an AI, why do I need to learn how to code? Well, part of the reason why it's important for you to learn how to code is even if an AI is producing the code for you, you still need to be able to check if it's correct. You need to be able to verify it. And in some cases, the AI may not actually be able to produce very good code for you. One of my colleagues in our computer science department, Dan Benet, and some of his students actually did a study a few years back where they gave a group of people access to an AI that would help them to code and another group didn't have access. And what they asked these students to do was basically to write some code that involved computer security. And then after they wrote the code, they checked how long did it take them to write the code, how confident were they about the code, and how secure was the code. And here's the interesting thing. The group that actually worked with the AI, they were faster. They were able to solve more problems. And they were more confident about their code. But the interesting thing is the code they wrote was actually less secure than the people who just wrote the code from scratch. which goes to show you that you need to know something about coding in order to really be able to verify the code that comes out of an AI. And that's why it's important for everyone to learn a little bit of code. I wish I could understand this stuff, but it's too late. I'm 60 years old and barely remember division. Well, to be honest, I'm almost 60 years old and I barely remember division as well, but I still program. And part of the reason why I do is because it's never too late to learn this stuff. Learning the basics of programming is actually pretty easy, pretty straightforward, and doesn't require that you remember a whole bunch of math from the past. It's actually kind of like learning a new skill, like if you were to pick up a new instrument or learn how to play a new sport. And part of the idea there is to understand the basics to actually do some pretty fun and cool things with it. You don't need to do a lot, and it's never too late to learn. So, here's a question lots of colleagues ask me. Why does code work one day and another day it breaks? Well, part of the reason why code breaks is because sometimes things that that code relies on gets updated. So there might be libraries or other bits of code that a particular program is actually using and those things change and suddenly the code you had before breaks or sometimes that code relies on things like an internet connection or a particular server being up somewhere and if the server goes down or the internet stops working then the code breaks. And sometimes it just turns out if the code's really old, some of the things that it actually uses in there, some of the functions that it calls may actually be what's called deprecated, which means they might not be supported anymore. And guess what happens? The code breaks. So there's lots of reasons why code can work one day and break another because it's part of a more complex system. And understanding that complexity is just part of understanding programming. How do you balance the technical and ethical dimensions of teaching computer science especially at the intersection with AI tools? Understanding the ethics of applying AI in different contexts is extremely important because we need to understand how these AI tools which are very powerful affect people's lives. They do things by affecting for example things like employment, what jobs might actually go away or how certain jobs are going to change. They change the sorts of opportunities that are available for people by understanding something about technology and AI. It suddenly opens up more possibilities for the kinds of problems you can tackle or the kinds of jobs you might go after. And it's important to ultimately understand when we engage in this endeavor of computer science, when we use AI tools, what are we trying to achieve? We're trying to achieve a better world. We're trying to solve problems. We're trying to create a more just society. And so when you think about the ethical implications of using these tools, that has to go handinhand with understanding the power of the technology because that's the only way that we make this technology work for everyone. You know, some people I'm sorry once I stop laughing, I will I will answer this question. All right. So, some people wonder, should I be nice to Chat GPT just in case AI rebels against humans? I guess it couldn't hurt, but I'm not really worried about that case. I mean, let's be real, AI is not going to rebel against humans anytime soon. And if that's what you're worried about, good reason to learn how to code, cuz then you can build your own AI to keep you safe if that happens. But seriously, AI isn't going to rebel against humans. But be nice to chat GPT. Not because it's an AI, but because it's important to just be nice in general.