Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Should you use AI to help you learn?
75 Plays20 minutes ago
In this episode, Kevin discusses the complexities of using AI tools for learning, particularly in the context of front-end development. He emphasizes the importance of critical thinking and careful vetting of information provided by AI, highlighting both the potential benefits and significant pitfalls of over-reliance on these technologies. Kevin shares personal experiences and insights on how to effectively integrate AI tools into the learning process while maintaining a strong foundation in traditional learning methods.
Recommended
Transcript
Introduction to 'General Musings'
00:00:02
Speaker
Hello, my front-end friends, and welcome to my podcast, General Musings. My name is Kevin, and in this podcast, I talk about whatever is front of mind for me in any given week, usually in some way that's related to front-end development.
Is learning from AI tools advisable?
00:00:14
Speaker
And this week, I just had someone the other day DM me asking if I thought it was a good idea to learn from an AI tool. And the the phrasing of their question led to me basically saying no.
00:00:28
Speaker
But I do think there's more nuance to this and I do think it is something that a lot of people are trying to figure out and some people are over relying on them and some people are maybe not using them enough.
Exploring AI as a learning tool
00:00:41
Speaker
I won't say enough, but there're and we'll we'll get to that. But what I want to say first of all is, is it a good idea to learn from an AI tool? I will say no. in a way I think that's the simple answer right there. ah Is it a good idea to learn using an AI tool is a little bit of a different question.
00:01:01
Speaker
because there's a lot of nuance that can come in there. And it's possible that the AI tools can be useful in helping you learn just like they can be useful for writing code if you're careful. And that's the problem if it with them in general. And the code that they can output can
Experiences with AI tools like Claude
00:01:21
Speaker
be okay. And actually, at times, I've i've done a lot of testing with it because it is something that I've wanted to See, I will say certain tools are better than others, Claude, as far as I'm concerned, is heads and tails above every other one as of the time of recording when it comes to writing code. And it still makes mistakes. So you you always need to be careful with it.
00:01:44
Speaker
Some the other tools that I've tried like they often just give you crap. So be really really careful ah with it and that's why I said if you're doing it to use them as a learning tool can be one thing is if it's part of your arsenal of stuff to learn from where maybe you're watching a video of mine or you're reading an article and something happens where you're kind of confused by it and or you don't understand exactly what's happening, or you come across a bit of code that you just can't understand. You have the code in front of you, you're looking at it, you have no idea what it's doing. Copy that, paste it in, and ask the AI to explain it to you. Often, not always, but often they will do a decent job of explaining what the current code is doing.
00:02:29
Speaker
So in that context, I think it can be useful. And especially if it's with ah you, that's part of your learning process. So if you are following a tutorial or a blog post and you're looking for extra clarification because it can help you vet the information that the AI tool is giving you.
00:02:46
Speaker
Because if what it says aligns with what you're already learning and it just clarifies the concept a little bit, that's perfect. And other times it might contradict what was in the lesson that you're going through. And that's when a red flag should go up.
00:03:03
Speaker
And that's why I think if it's part of what you're doing and you're just leveraging it in small ways that you still have, you know, if it's done in a way where you can know if what it's saying is wrong, then it's okay. Because
Challenges of AI confidence and learner impact
00:03:19
Speaker
that's always the biggest issue with all of them now. And I'm going to probably go on a rant about these AI tools in a minute. But the issue is that when they get something wrong, they sound like they're still right.
00:03:30
Speaker
high level of confidence, no matter what, even if it's just spewing nonsense, right? And so that as a someone who's learning a subject,
00:03:42
Speaker
who doesn't know any better that's a really easy way to just get things completely wrong and then misunderstand things completely right because if it tells you something that's just wrong but it says it in a way that sounds very right and it explains itself and it you know it always does that where it it goes in detail to explain why its answer is correct and so it makes it sound like it is and even if it's write 95% of the time. If that 5% of the time is wrong and that might be a high, like I don't know if they're right 95% of the time, but if they are right 95% of the time, that 5% error rate is really high and it's enough to screw you up as you're going through and as you're learning new stuff.
00:04:26
Speaker
right so um Yeah, you got to be really careful with that. And if you can use them in contexts where it makes it easy to find conflicting information that it's saying, then you can load up another resource and see that's not, don't ask another AI tool, find a different resource, another tutorial, an MDM post, anything that can tell you what the actual answer is. And it takes a little bit more work, but that's part of learning. Learning should be hard.
00:04:52
Speaker
If you're using the AI tool to try and make learning easier, it's just doing stuff for you. It's it's not making anything easier. so ah There's no shortcuts to learning. It's hard and takes work and you just got to learn stuff and put in the hard work yourself. so I do want to put that out there as well. um I said I was going to go on a little rant. We'll do that now just before I go off topic on the other things because I was going to transition here anyway.
00:05:17
Speaker
of
Critique of rapid AI product integration
00:05:18
Speaker
it's amazing to me how these tools are being pushed as production ready across every company that's out there. And I understand why it's happening. But if this was anything else, there's no way that this would be happening. Their error rates are so high and they're getting better, but they're still terrible. That it's just like, it's this mad rush right now to,
00:05:45
Speaker
integrate these tools into products in ways that people don't want them and to figure out ah part of it is a learning process there but like they're wrong so often and the news it's constantly in the news right they're making huge big glaring errors and like the the latest was the siri one right where it was uh or it was not siri it's the Apple's Apple Intelligence giving a recap of an article and giving like the recaps weren't what the article was saying. And it's just constantly these tools are terrible and they keep getting things wrong. Like a 1% error rate in most things would just not be acceptable. But these things just keep getting pushed and pushed and pushed. And a large part of it is the shareholders that are involved in these companies are pushing on it because I think they're like once we crack this nut,
00:06:37
Speaker
If we can be the first to do it and have it really good, then that is a huge bonus to us. And there's so much money in this space right now with investments going on.
AI's role in startup funding
00:06:46
Speaker
And then I heard that like if you're right now trying to get funding for a startup or something else, you have to basically have some integration of AI if you actually want to get funding. There's probably some people out there that aren't relying on it as much, um but a lot of the funding is coming from some sort of integration. It doesn't have to be the main thing of your company, but it has to be a part of it.
00:07:05
Speaker
And there's just so much craziness going around it. And part of it surprises me there with the the people wanting to invest in these things where AI has to be a part of it, just because like at one point the price of all of this is going to go up. If you're relying on AI tools and not building your own and the funding All the AI tools currently are being subsidized to a ridiculous degree that isn't going to last forever. And then the prices go up. And when the prices go up, if you're a startup that's relying on some one of these AI tools running in the background,
00:07:38
Speaker
and the price goes up tenfold, your product just shut down. Like you you don't have a product anymore. So yeah, I'm really curious ah about all that. But anyway, that's neither here nor there. I wanted to talk more about just the idea of like, they're really It's amazing how bad they are and how much it's being pushed on us, I think. And I think a lot of it is speculation. Like it's just this world of speculation right now on like this will be a good tool eventually and we need to be, we need to integrate it now so people don't start using the other tools that do have it.
00:08:12
Speaker
And then once it's better, that will be great. Sort of how it feels right now. I don't mind selectively using it. Like if you have just your little chat bot on the side and there's other things that I think, I think we're gonna get past chat bots at one point. That that could be a functionality of it. um But I think there's other use cases of them that can be really good. Where we're not even asking them to be smart. We're building in a lot of the steps for it and they can just do some stuff for us. I think it can be more useful use cases.
00:08:40
Speaker
But when you're just trying to like shove these features into something else, I went off track and I forgot exactly what to say. Anyway, yeah, I'm just amazed at how bad these things can be for something that's being shoved down our throats. And the thing there too is most users, I don't think want them.
00:08:57
Speaker
So there's a disconnect there too, which is really interesting.
Skills for effective AI learning
00:08:59
Speaker
But let's circle this back to learning of the nuance of using AI tools to learn. I really do think there, I don't want to tell people not to use it at the same time as much as, i'm I started off with the no, don't do it. And then I went on a rant. So if some people would have stopped listening, that's fine. um if but If that was the answer you you were looking for and that's all you want. If you're going to go on one side or the other,
00:09:24
Speaker
not relying on it, especially for learning is probably a good thing. And just always remember, learning is hard. But there's definitely a skill in using these tools. And I think these tools are here to stay. So learning how to do proper prompting. It's like how we used to learn how to Google things properly. ah Learning how to prompt them and get the right information from things is really important. And some The better the tool, the better it is at needing less specific prompts. But just like if you're asking, I don't use TypeScript, but I was working on something recently that had to have it. So I was like, oh, i'm goingnna the AI is going to help me with this, so I don't just have type errors everywhere. And it would help, but then it would also just randomly remove stuff that I needed in there that was really confusing.
00:10:16
Speaker
um or i'd give it In other things I've done where I've just given it like regular JavaScript, it turned it gives me a TypeScript solution. Or the most frustrating for me is anytime I ask it anything related to code, this is closed specifically, it will create an example using React and Tailwind. And then it's like, so you have to prompt it saying like with whatever stack you're using every single time. It it's it can be a little bit frustrating on that, which is why I don't use them very much.
00:10:49
Speaker
at all, really. The the main thing i'm I'm using AI tools for now is this really quick, like I need 10 titles for a YouTube video that's about this. Can I have some ideas, please? And actually now it's interesting if you ask Claude for thumbnail suggestions, it will actually create SVGs for you. ah They're all terrible and really bad, but it's kind of funny that it it can do that.
00:11:09
Speaker
um But yeah, i'm not even then, I don't think I've had one title that I've actually used that was just like, oh, yeah, that's the title I want. It's either way too sensational. And then I'm like, oh, make it less sensational. And then it's just way too dry. And like just going over and over and over, I'm like, oh okay, I'm just going to make my own titles again. Let's do the sealed way. Again, maybe I need to get better at prompting. I don't know. But I also feel weird every time I use them because I know that they're not great for the environment and everything else and that you were stuck in this weird
00:11:42
Speaker
world now where it's sort of they're here, but I don't really want them to be here. And what is the long-term future of these tools once the funding for them or the subsidization of them, I should say, starts drying up a little bit. Um, I'm curious on that front, but yeah, I think it's in today's day and age, an important skill to learn how to use them and how to vet the information they're giving you. Cause
AI in coding tasks
00:12:07
Speaker
again,
00:12:08
Speaker
they give you really bad stuff sometimes and that's the advantage is like if you know the subject that you're asking it about and it's giving you some code you can know if that code is good or bad like you you do a code review of it because you know it whereas if it's something you're learning you don't know if it's good or bad an example actually of one thing i use it for quite a lot is if i have um like i'll grab copy and paste a whole bunch of like just um color values that out of Figma, there's ways I could do it with a plugin too that I probably should use instead. But if I give it a whole bunch of color values, I can have it turn into custom props. But then I can say like, in if I give it the single color value, it can then expand that out for me where I can say like, every one of these values should have 10 variants that are getting 10% lighter or darker, all based on HSL. And it will do that.
00:12:58
Speaker
You can have it do, you know, that then make utility classes and it can do that. But you have to be really specific with your prompting. And then you can, again, vet the code. It's hard to get that wrong, but it can get that wrong. But yeah, I think learning to use them and what the limitations of them are and the types of things they get wrong and be able to vet the information is the best thing you can be doing.
00:13:21
Speaker
And if,
Conclusion: AI's place in learning
00:13:22
Speaker
again, the way I talked about it a bit earlier about if you are using it as a learning tool, don't learn from the AI, don't believe a word that it says, but definitely using it for clarification on things I think can be ah that That can be valid because you're putting it in the context of a bigger picture where you already know about the thing and it's just like, okay, that makes sense within this context. Then of course, yes, that was good information and now I understand it better or that doesn't make sense based on what I was learning from that lesson. I need to look at this more and figure out what's actually going on. I think that's the best way to use it as a learning tool, at least in my own opinion.
00:14:02
Speaker
um But yeah, I think that's it for this week. I didn't know exactly where this one would go, a little rant there, and then circle back around. So sorry if I repeated myself a little bit. I hope you enjoyed listening anyway, and until next time, and until next time, of course, don't forget to make your corner of the internet just a little bit more awesome.