Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Lesson 2: Foundational Fluency with AI
8 Plays5 days ago
Stop treating AI like a search engine and start using it as a "cognitive architect" to bridge the gap between just having information and actually building expertise.
Topics discussed:
- The Knowledge Illusion: We often fail to distinguish between knowledge stored in our own heads and information readily available in our environment, leading to a false sense of expertise.
- Illusion of Explanatory Depth: Instant access to the internet can make us feel more confident in our understanding of a topic than we actually are; forcing ourselves to explain a concept is the best way to break this spell.
- Setting the Stage: Effective learning begins with isolating core concepts and stripping away non-essential content to reduce cognitive load.
- Deep Processing: Based on the principle that "memory is the residue of thought," learning happens when we actively think about the meaning and implications of information.
- AI-Driven Retrieval Practice:
Free Recall Loop: Using AI to analyze what you remember from memory and identify gaps or misunderstandings.
Pre-flight Diagnostics: Tasking AI to quiz you on foundational terms to ensure proficiency before moving to complex material.
Borderline Case Testing: Improving conceptual clarity by asking AI for clear examples, non-examples, and tricky "borderline" cases. - The Power of Chunking: Organizing disparate facts into small, meaningful units (like how a chess grandmaster sees patterns rather than individual pieces) to make complex material easier to retain.
Recommended
Transcript
Introduction to Foundational Fluency with AI
00:00:00
Speaker
All right, everyone, in this video, we're gonna talk about achieving foundational fluency with artificial intelligence. What exactly am I talking about?
00:00:12
Speaker
Well, whenever you're learning something, and I mean actually learning it, right? Trying to become an expert in that or at least being able to um talk about it and engage in critical thinking on that topic. The very first step, the step that we're going to cover in this particular video,
00:00:32
Speaker
is getting what I call foundational fluency. You have to know the key terms, the who, the what, the where. All of that is actually what higher level cognition is based off of. When people say something like you want to be a critical thinker,
00:00:50
Speaker
There is no such thing as critical thinking in general. There's critical thinking about a topic and the way you get to become a critical thinker about a topic is by knowing basically a bunch of stuff about that topic or that's at least the first stage.
Understanding the Knowledge Illusion
00:01:09
Speaker
So that's what we're going talk about in this video. And the first thing that I want to mention, I guess I want to convince you that like you have to worry about this. Because there is such a thing as the knowledge illusion. Now this is bananas on multiple levels, but let me define this.
00:01:30
Speaker
The knowledge illusion is a failure to recognize the distinction between knowledge that is internal and knowledge that is external. That's crazy.
00:01:42
Speaker
That literally means that you can't tell, at least sometimes, when the knowledge that you claim you have is in your brain and when the knowledge you claim to have is like in things around you.
00:01:57
Speaker
So sometimes, let me give you a quick example of this, and I'll read this off the slides. We sort of don't have a clear distinction between internal and external knowledge because a lot of what we know is not directly in our heads. It's spread out throughout our environment. And the example that I can give you is well, maybe it's your phone or maybe it's a notebook that you carry around everywhere with you.
00:02:26
Speaker
But there is this idea that your mind is actually not just between your ears. It's extended into all these different cognitive tools. So if you have a planner and you write down everything you have to do in that planner, and you have special notes to yourself and it keeps you motivated and oh, something you keep your snacks in your planner. I've seen people just like open up their binder and snacks pop out of it.
00:02:56
Speaker
That really is, you know, an extension of your mind. Like it's like a spare brain almost. And it almost helps you keep your real brain running by keeping the snacks there too.
00:03:07
Speaker
The point is here that not all your knowledge is between your ears and so that's the first thing i want to try to convince you of let's talk about a pretty commonly observed phenomenon in long-term relationships and this i'm actually talking about this from personal experience So recently, my wife had to um book flights and hotels and and rental cars and all that good stuff on her own.
00:03:43
Speaker
And she went into it real confident, um but she quickly realized that she had lost
The Impact of External Knowledge and Overconfidence
00:03:50
Speaker
practice. She had forgotten how complicated it is because for at least five years i had been the one exclusively doing that and so when she tried to do it she remembered you have to you have to make sure that when the plane lands you're able to you know check into the hotel soon afterward otherwise you have to like find something to do for a few hours, like those things all matter.
00:04:17
Speaker
And you forget about that if you don't regularly do it. And so she had forgotten that, you know, how to basically plan for a vacation.
00:04:27
Speaker
And she's not alone. i ah i also did something similar so she offloaded knowledge about planning vacations to my brain and i offloaded knowledge about our health insurance or medical insurance to her i was asked recently actually hey you know what do you know about your uh your health care um do have ppo hmo whatever and after a little bit i was like oh
00:05:00
Speaker
I don't know any of this information. My wife deals with all that all of that stuff. And so we were, or at least I was, and I guess so was my wife, we were momentarily in the under the knowledge illusion. We thought, you know my wife thought she could vacation plan roughly you know without me.
00:05:20
Speaker
I wasn't as easy as she thought, right? And I thought I knew something about my healthcare, but no, that that knowledge about my healthcare lives in my wife's brain right so that is an example of the knowledge illusion consider also the illusion of explanatory depth i love this one and we can get an example of this from the internet So apparently people mistake instant access to information for actually having understanding about that topic. So in one study, they asked people, hey, you know how zippers work? And, you know, most people don't know how zippers work, but then they had them look it up. So they they go online, they look up how zippers work and cool. Now they know how zippers work.
00:06:15
Speaker
Then they ask these people a couple of follow-up questions and like, ah hey, um why are cloudy nights warmer than non-cloudy nights? And they ask them, how confident are you that you can explain that?
00:06:30
Speaker
Well, apparently looking up how zippers work made them more confident in you know saying that, oh, I know how why cloudy nights are warmer than non cloudy nights. In other words, the fact that you just recently looked up information sort of lulled into the view that, oh, I got a bunch of information in my head.
Strategies to Overcome Knowledge Illusions
00:06:57
Speaker
Like you forgot that you got it from your phone or from the internet or whatever, right?
00:07:03
Speaker
That's kind of bananas. You think you can explain things better than you objectively can explain things because you forget that a bunch of your knowledge isn't between your ears. It's actually in your devices, in your notebooks, in your surroundings.
00:07:23
Speaker
So what to do about this? How do you not fall into these illusions of knowledge? Well, the number one way to do it is to force yourself to explain a thing. If someone is under the spell of of knowledge, right, the knowledge illusion, you tell them, all right, explain it to me and be very precise.
00:07:46
Speaker
And as soon as they can't explain it very well, that's when they realize, oh yeah, I don't know this. Yeah, okay. Well, i mean, hopefully that's at least the yeah ir rational ones among us. So,
00:07:59
Speaker
What does this mean for learning? Well, when you're learning new material, you have to actively test yourself or else you might fall prey to the illusion of explanatory depth. You literally have to check, otherwise you won't know that you don't know what you think you know.
00:08:22
Speaker
I know that's kind of a weird ah you know word salad, but that is how that's the brain we got. I'm sorry if this is surprising to you. Okay.
00:08:34
Speaker
Let's get into then how we learn. Now, I kind of organize these principles or these prompts that I'm going to share with you around the principles of cognitive science of learning. Sorry, the principles of learning from cognitive science.
Utilizing AI for Learning Objectives
00:08:51
Speaker
I read a bunch of books here. If I had to recommend just one, probably Why Don't Students Like School? That's a good one. It's probably the one I'd recommend. But um Yeah, let's take a look at some prompts that you might use for developing foundational fluency.
00:09:09
Speaker
all right, let's start here. The very first thing that you actually have to do, believe it or not, is set the stage. You have to figure out what the objects of your attention even are.
00:09:26
Speaker
An often overlooked stage of learning is creating the right conditions for attention. And so this is what James Lang talks about in his book, Distracted.
00:09:36
Speaker
First of all, get rid of all distractions, right? Obviously, you know, that's that's one key step. Get out, filter out all the irrelevant stimuli. That is all true, but this includes, by the way, non-essential content.
00:09:52
Speaker
When you are first becoming acquainted with a new topic, um I don't know, the Reformation or photosynthesis or whatever, the the first thing you want to do is, well, you can't look at it all at once. You have to only look at the core content. And once you get the basics down, then you can kind of fill in those details and add some nuance and all that. But the very first thing is isolate the core concepts.
00:10:21
Speaker
How do you do that? Well, AI can help. AI can help you identify core learning objectives based on an instructor's lesson slides or their study guide questions, or preferably all of the above, right? You get their slides, you get their study guide questions, you get whatever summaries they give you, and that will help you realize what exactly your instructor wants you to learn.
00:10:47
Speaker
So I have a prompt here. And in this prompt, well, you know, I break up all my prompts into sections with headings. So here's the first part. You're telling the AI what you want it to do.
00:11:00
Speaker
So you tell it, you are an adaptive personalized tutor for an undergraduate in college. Rather than provide answers, I should say answers, you nudge in the right direction.
00:11:14
Speaker
So task one is I've attached my instructors, whatever, right? Presentation slides, study guide questions. I would like for you to extract the learning objectives. Task two.
00:11:26
Speaker
I would like for you to provide a glossary of key terms that might be helpful for me as I review and continue to learn this new content. If it will be helpful, please also provide a summary of the main ideas and key figures.
00:11:43
Speaker
So that's the prompt. Let's check out what happens when we actually use it.
00:11:54
Speaker
okay so i've attached i pasted the prompt that i showed you earlier and i've also pasted in um some gamma presentation slides i got this from a friend of mine who is a philosophy teacher and gamma is is this uh online um you know kind of a slide deck kind of thing and all i did and i know a lot of instructors are doing this nowadays all i did is i grabbed that um ah link and I pasted it into where I have a little you know space for that in my prompt. And so let's ah let's see what happens, right?
00:12:32
Speaker
While this is ah doing its thing, I should tell you that for links, Gemini is usually a good one, right? Gemini, which is think part of Google is a good one for this kind of thing. Okay, I did, ah it it did its thing.
00:12:49
Speaker
And now we see that it broke it up into a couple of different learning objectives. Define and apply speciesism. That's a very important term in the ethics of animals, right? Explain the principle of equality.
00:13:06
Speaker
All right, so that's another important idea from the presentation. Analyze the role of suffering and evaluate ethical consistency. It seems maybe, right, that's what the lesson's about that, you know, humans tend to... um privilege you know their own species over non-human species for no
Deep Processing and Memory Retention
00:13:29
Speaker
real good reason. And that's sort of the challenge that Peter Singer ah presents in this presentation. By the way, ah these key figures are also highlighted and explained in this little glossary. that Gemini provided for us. Here's the glossary too. Speciesism, sentience, equal consideration of interests, right? so all these words. And we get a summary of main ideas.
00:13:55
Speaker
So this is um this is good. The main important thing here is that we are stripping away things that aren't essential. Because you might spend a whole lot of time trying to learn something that, first of all, isn't like, you know an essential part of the lesson.
00:14:12
Speaker
And secondarily, it actually is sort of slowing you down because when you have just the raw, you know, not the bones, like the bones aren't raw, but when you just have the skeleton, when just, you know, the main structure of it, it's a lot easier to learn.
00:14:27
Speaker
So this is why this particular step, I think, is so valuable. Okay, we have here the learning objectives. We're gonna use these for another slide. Let's go ahead and and jump into that right now.
00:14:42
Speaker
So we have... this slide that we just used, let's move into a different learning principle. So this is called the principle of deep processing. And this is, to be honest, exactly how learning happens. Here I call it the engine of learning.
00:15:04
Speaker
And this, you know, you can learn about this if you want to learn about neural networks. And this is a way that machines learn to write. machine learning is based on this idea. Here it is in a nutshell.
00:15:16
Speaker
The more you think about the meaning and implications of information, the better You remember it. I like the way Daniel Willingham puts it. He says, memory is the residue of thought.
00:15:33
Speaker
Think a lot about a thing and that is how you will remember it or learn it is another word for it. What is the implication? Well, once you know the learning objectives, design AI exercises that get you to think about whatever it is that you want to learn.
00:15:53
Speaker
So for this, i came up with a couple of different prompts and exercises. So let's take a look at um two of them, or I think three of them. Okay, here's the first one, free recall into an AI feedback loop. Okay, here's what you do.
00:16:11
Speaker
You carefully read through the material. Obviously, to learn, you just have to read and practice and and get your flashcards out and do that thing. That's obviously step one. you don't You don't need any AI for that, by the way. Just get to work on that a little bit.
00:16:28
Speaker
Then close your notes. Write down everything you remember from the learning objectives and glossary that you just studied, and then paste in your response into the AI after retrieval.
00:16:45
Speaker
Okay, so it's a form of you know active recall or retrieval practice is what this is called. And that's when you really force yourself to write down everything you remember. You're like really digging deep.
00:17:00
Speaker
and writing down things from memory. And then from there, you basically ask the AI, hey, how'd I do and how can I improve? So here is the prompt that I have for that role.
00:17:13
Speaker
You are an adaptive personalized tutor for an undergraduate in college, rather than provide answers, you nudge in the right direction. Below is my attempt to recall the key ideas from a recent lesson without notes.
00:17:28
Speaker
Identify what I got right. Identify what I missed or misunderstood. Ask me two follow-up questions that target my weak spots. Here is my recall attempt. You are paste it in.
00:17:41
Speaker
And for reference, here are the learning objectives from the lesson. You need these because, you know, depending on which AI you're using, maybe it doesn't remember what the learning objectives are. Maybe you never gave it to it Right. So it doesn't know.
00:17:56
Speaker
So let's try this out. and Let's see what it how works.
00:18:05
Speaker
Okay, so I input a couple of answers here. um And obviously, I'm an ethics teacher, so i I know the answers. But I put some answers in there that I think are fairly common mistakes. Or I mean, I guess it's not completely correct.
00:18:20
Speaker
So let's check out what happens. I will um press enter here. And... Here we go. I've done a great job. What you got right. All right. What you missed or misunderstood. All right. Equality of condition versus consideration.
00:18:36
Speaker
Personhood versus moral consideration. Ooh, I've got a lot right. I actually got more wrong than I thought I was going to get. ah okay How to refine your knowledge. So then it corrects me and and adds a couple of, oh, this is actually pretty good. um Yeah, these are these are great comments from um from Gemini here.
00:18:59
Speaker
And by reading through it and having it correct your initial um thoughts and ideas, it really is you know helping you sort of calibrate the concepts. Oh, I thought this term meant this, but really it's closer to this other thing.
00:19:14
Speaker
So this is great. All right. So that's that's one exercise you can do, which I call free recall AI feedback loop. That's probably a bad name. I can probably come up with a better name.
AI in Diagnostic Learning and Feedback
00:19:25
Speaker
tell you what, if you come up with a better name,
00:19:27
Speaker
Let me know. And yeah, we'll do that instead. Okay. Let's look at exercise number two. This is a simple diagnostic. I really enjoy diagnostics. I call them sometimes pre-flight diagnostics because these, you know, you really shouldn't move on to a more complicated um ah ah exercises.
00:19:52
Speaker
before you know the foundational terms. More importantly, maybe you shouldn't move on to the next lesson unless you know the foundational terms. Otherwise, you sort of you're always gonna have this nebulous understanding. So, okay, what to do?
00:20:07
Speaker
You provide the AI with the learning objectives and the have the AI quiz you so as to identify the concepts you're having trouble with. Answer without notes. I mean, it should go without saying here.
00:20:19
Speaker
Have the AI evaluate your level of clarity and accuracy. so I'll show you the prompt here. The role is the same, it's an adaptive personalized tutor.
00:20:31
Speaker
And um the part that it matters here is this task. Please ask me four to seven questions based on the learning objectives below in order to assess my proficiency.
00:20:45
Speaker
Take note of my mistakes, but do not provide the right answers. Rather, continue with the assessment. After the assessment, provide a debrief and advise me on what to continue to study and what learning techniques I can use. And then you paste in the learning objectives. So um let's try this out and um see what happens.
00:21:09
Speaker
Okay, so here we are with ChatGPT. I switch over to Chat, just so you can check it out. I don't know. And i i paste it in the prompt, and I made one little change to the prompt, by the way, so that it asks one question at a time. I kind of prefer that. And so here it is. In your own words, what does speciesism mean, and how does Singer compare it to racism or sexism?
00:21:31
Speaker
Now we can go through and answer each of these and it'll ask. I'm not sure how many questions it'll ask, but um I asked it for four to seven. And then at the end, it will tell us um you know a couple of things. So I'm gonna go ahead and type in some answers here and let's see what happens.
00:21:52
Speaker
Okay, I typed in some answers and I actually went ahead and just skipped a question or so so, I can see what happens. And here we go. What I got well, why I answered a couple of questions correctly.
00:22:05
Speaker
What I can strengthen, the role of suffering. Okay, I didn't do well enough on that. Marginal cases, that's one I just skipped. and um applying equality of consideration. So it actually will tell me what I should study next, a quick little summary of these things, and even study techniques to improve.
00:22:26
Speaker
Contrast drills. Oh, right, short comparisons. This is pretty good. Argument reconstruction. All right, explain out loud. This is something that Richard Feynman used to do, I believe, where he would practice...
00:22:41
Speaker
He would learn something basically with the goal of explaining it to someone else and that would apparently help. And edge case testing, we are actually gonna do some of this ourselves in a bit.
00:22:53
Speaker
But as you can see, this diagnostic is very useful because you have to simply test yourself to see what you understand. So, okay, let's move on to another example here of another possible prompt you can use.
00:23:10
Speaker
And this has to do with, as I mentioned earlier, you know, borderline ah case testing. So let's check this out. This one, i call it the pre-application retrieval um because it basically checks for if you're ready to try out some of my application exercises. So that's in the next video, actually.
00:23:31
Speaker
But... Here's what you do. Identify a concept you're having trouble with. And what you're going to basically do is have the AI help you grasp that concept by helping you see what it is not.
00:23:45
Speaker
So the prompt is the same role as before. And here's the task. I'm learning the concept of blank, whatever it is. Please give me one, sorry, one, three clear examples that definitely fit the concept. Two, three borderline cases.
00:24:04
Speaker
And three, three clear non-examples. afterward explain what feature distinguishes the true cases from the others. Alright, so let's check this out. Let's try it out on ChatGPT.
00:24:18
Speaker
Okay, so I've input the prompt and the concept that I want to learn is speciesism. Now, ah it is going to give me a couple of ah very clear examples of speciesism, then some borderline cases, and then some obvious non-examples, right? So I'm looking at these now.
00:24:41
Speaker
These look pretty good, actually. ah Potentially better... than anything I might have come up with on the fly if you were to have asked me for exactly this.
00:24:53
Speaker
And this is exactly the benefit of these sorts of things. Now, I'm not saying that there won't occasionally be hallucinations. In another video, I will talk about when hallucinations might pop up.
00:25:06
Speaker
But for a lot of, um especially, you know, older ideas, right, if you're thinking about, if you're learning an article that's at least five years old, a lot of these large language models are going to do great on those kinds of things. It's really only really new things that it might hallucinate about. And obviously, if you can, you know try to trick it, yeah, well, then it might hallucinate.
Chunking for Memory and Understanding
00:25:31
Speaker
In any case, this prompt is a success. I really like these and i think they're great examples. Let's do one more prompt. And which means we're gonna do one more learning principle.
00:25:45
Speaker
So now we're gonna talk about chunking. Chunking is perhaps my favorite um learning strategy. Let me explain what it is first. Chunking is when you group small bits of information into larger meaningful units.
00:26:02
Speaker
The prototypical example of this is chess grandmasters. If you show um a game, and you know a chess game to a chess grandmaster in the middle of play, right like some 20 moves in, they can, after a couple of seconds, memorize the positions of all the pieces on the board.
00:26:24
Speaker
Now they don't really memorize it, you know, by thinking to themselves, okay, knight is on F3 and the pawn is on C3 or whatever. That's not how they think about it. They think about it as patterns. They recognize, oh, that's a Scandinavian defense or that's the king's pawn opening.
00:26:44
Speaker
And that's how they're able to memorize where all the pieces are so quickly. They're not looking at the little small pieces, the small bits of info, they're looking at the larger patterns. So instead of these little blocks in this image, they're looking at the larger blocks, the larger pieces.
00:27:01
Speaker
And that's a lot easier to memorize. It's easier to memorize, you know, five large pieces of ah of of meta information than 50 small pieces of information, right? So that's the whole idea behind chunking. Now, chunking is incredibly useful when you're learning new material.
00:27:20
Speaker
because it helps you, you know, organize the concepts that you're learning into some kind of cohesive unit. And that's exactly what AI can help you do.
00:27:32
Speaker
This is often the case um that in philosophy where you learn all kinds of isms, right? And materialism, dualism, and then you want to figure out how they all fit in together.
00:27:45
Speaker
Well, you could, of course, um you know, really study real hard and and read all you want and and try to group them into meaningful units. But someone with a lot of content knowledge, like your instructor, might be able to do that really quickly for you.
00:28:00
Speaker
They might draw you a diagram or something like that. Now, if you can't get your instructor to do it for you or you're on an online course, right, and you kind of want it, you know, now there is an easy prompt here that I made up.
00:28:13
Speaker
that you can actually have the ai organize the concepts for you so that they are easier to you know sort of uh cohere into a meaningful unit so this is actually sadly from one of my former um study guides and this is not helpful at all to be honest i mean it was my first semester i think so you have to cut me a little bit of a break here but i would basically say hey be familiar with all these arguments Now these are all arguments having to do with the philosophy of religion, but they're not organized in any way that makes it easy for you to see how they fit together with each other.
00:28:54
Speaker
In other words, it's a blend of arguments for and against God's existence and different types of arguments too. So this as it is, is not terribly helpful, but I have a prompt here that can help you organize it so that it becomes more meaningful.
00:29:10
Speaker
So the role is, as always, that the AI is an adaptive, personalized tutor. And here is the task. I would like for you to organize this material according to the cognitive science principle of chunking, which states, learning is easier when material is organized into three or four organized units, each of which itself can contain three or four units.
00:29:35
Speaker
This principle can be used to organize entire lessons into units, but also operates at more granular levels. Create any tables or concept maps that will aid my understanding.
00:29:48
Speaker
material insert learning objectives and glossary of key terms so this is where you insert whatever it is that you want the AI to chunk for you and let's try it out last one let's see what happens okay so i pasted the prompt in and I also went ahead and pasted in my my former study guide, which again is just a random list of argument names that my students were supposed to learn.
00:30:19
Speaker
And for this one, I'm using Claude. And let's see what happens here.
00:30:30
Speaker
Okay, so Claude generated pretty masterful little chunked architecture. It's basically three master chunks, arguments for God's existence, arguments against God's existence, and little response, and the epistem epistemological foundations of these types of arguments.
00:30:53
Speaker
So that is ah very useful and it's actually so useful that I think I want to use it for my class. ah Wow, this is embarrassing. In any case, so it organizes it for you and you can kind of see that there's even tables being made here so you can kind of grasp the connections a little more easily or the key ideas more easily.
00:31:14
Speaker
And yeah, this is actually very well done. So I am quite happy with this prompt.
Conclusion: Enhancing Learning with AI
00:31:20
Speaker
And those are some of the ideas that I want to share with you. So here's what I want to leave you with.
00:31:24
Speaker
If we can kind of recap something that I mentioned before. the way a lot of people right now are using large language models they're essentially de-skilling themselves they are actually not creating the conditions that that allow them to go from novice to an expert and so i call this the the broken ladder of cognitive ascent now with the appropriate assignments and with the appropriate ai prompts you can actually easily get yourself
00:32:00
Speaker
up that ladder. And this ah video covered those very first steps, that very first rung of the ladder, what I call a foundation of fluency. You need to know sort of the basics of the content before you can move forward.
00:32:17
Speaker
And I think even if you were to only do these, this would be an incredible asset to your study sessions, right? This would be a great thing to do, even if you just stopped at that.
00:32:28
Speaker
But we can go further and I will show you a couple of more methods in upcoming videos. For now, i think you have some work to get to.