October 16, 2025
How to Teach Yourself Anything Using ChatGPT
Class Description
AI chatbots have the potential to revolutionize self-directed learning.
Michael Mercier — a lifelong autodidact — will explain how to use ChatGPT to teach yourself any subject. Drawing on his own experience integrating chatbots into every stage of self-directed learning, Michael will show how anyone can accelerate and deepen their learning process with the help of AI.
In Part 1, he’ll explore how learning works: what it means to learn, how the brain organizes knowledge, the stages of the learning process, and the difference between self-directed and traditional learning. In part 2 he’ll demystify AI chatbots, explaining how large language models work, what built-in project management features chatbots offer, and how to structure conversations to match your learning goals. In Part 3, he’ll synthesize these insights, providing practical ideas for using ChatGPT at every step of the learning process to master any subject independently.
Whether you’re a student, professional, or lifelong learner, you’ll leave with actionable tools to optimize your potential in the age of AI. (7:30–8:30)
BLOC Ministries
BLOC Ministries has been building relationships with its neighbors and sharing hope in brokenness since 1998.
As part of its mission, BLOC staff and families live in the urban neighborhoods where they work. This means 70+ men and women want to be a part of solutions to help make Price Hill — its children, teens, adults — safer, healthier and more resilient.
Horses on the Hill is BLOC’s equine therapy program, located on seven acres of property in the heart of Price Hill. It is one of six urban equine therapy programs in the country, serving a community where the majority of residents have experienced some sort of trauma in their lives. Its business development program is bringing well-paying jobs to the neighborhood in order to break the cycle of poverty and create a healthy support system that has never existed there before. These are just two of BLOC’s 26 programs designed to promote lasting change and bring hope to the communities it serve. (7:05-7:25)
Speaker Bio
Michael Mercier is President of Screen Education, which addresses issues at the intersection of digital technology and human well-being, including smartphone addiction, news media bias, and artificial intelligence, though research, seminars, and consulting.
Reflecting his passion for learning, Michael’s first major splurge after starting his career was a used set of World Book Encyclopedias. He then spent a decade as an editor and publisher in the college textbook industry, and another decade in market research.
He holds a BA in Psychology from Vassar College, an MBA in Marketing from the University of Michigan, and an MS in MIS from the University of Cincinnati.
Dwight Young Dwight Young is co-founder and Executive Director of BLOC Ministries. He holds a Doctorate of Ministry in Practical Ministries from the Emmanuel School of Religion and a Master of Divinity in Education from Southern Baptist Theological Seminary. He is board president of IDEA Charter School in Price Hill and former president of Cheviot-Westwood Kiwanis Club. Dwight has participated as a task force member as a part of the city of Cincinnati’s “Childhood Poverty Initiative.” He has also served on Kroger Company’s Cincinnati/Dayton Division Task Force for Equitable Communities since 2022.
How to Teach Yourself Anything Using ChatGPT
Michael’s first major splurge after starting his career was a used set of World Book encyclopedias. He then spent a decade as an editor and publisher in the college textbook industry, and another decade in market research. He holds a BA in psychology from Vassar College, an MBA in marketing from the University of Michigan and an MS, and in MIS from the University of Cincinnati, Michael.[00:01:00]
Speaker 2: Jerry, um, can you adjust the volume on the speaker?
Check? Check, check, check. Is, is it up? Check. Check
Speaker 3: now.
Speaker 4: Check, check.
Speaker 2: Um. So the, this subject, uh, is, is really complicated and I spend a lot of time, I mean, a lot of time over the summer. I started working on this on the summer trying to understand how chatbots work. So pardon me if I’ve tried to, you know, given everything I’ve [00:02:00] learned, I’ve tried to. I’m trying to present it in a really conceptual way, but I, I could fail in a couple of points, so excuse me if I do.
Um, so the, the title of the talk is how to Teach Yourself Anything using AI Chatbots. And, um, as Betty said, I’ve, I’ve been an auto didact my entire life. I’ve. Always, you know, been hungry to teach myself new things. And of course, everyone that goes to empower you is into learning, right? Um, but I, I had an experience, uh, uh, uh, during COVID where I, I spent about 35 hours studying.
Oops.
Studying how mRNA vaccines work. And this was before chatbots and um, and, and it’s something I just had to understand for personal reasons, right. And, um, I was sort of shocked. I would talk to other people, like, everyone you know about this is, do you know how mRNA vaccines work? And [00:03:00] nobody knew and nobody seemed to be interested in it.
And I started to think, well, maybe some people just don’t tackle certain topics ’cause they’re too difficult, you know, even if it’s important. Um, but I realized once I discovered chatbots, I felt that I could have cut that 35 hours, and I’m not kidding, probably down to like 20. Um, the, the chat bots are so effective at.
At teaching you things, it, it blew my mind. So I put together a talk that, um, I. Synthesizes sort of learning theory and chatbots. So I’m gonna start off talking about learning, like how, how we learn, what’s the learning process. Then I’m gonna get into AI chatbots and, and how they work, um, and how to use them optimally.
And then at the end, I’m gonna synthesize the two. And for every step in the learning process, just talk about how you can use AI chatbots. So starting with learning. Um, so just a working [00:04:00] definition of learning. Um, the way I think about this is at any moment you can think of us as having a, a base of knowledge, a static base of knowledge in our minds.
And learning is simply integrating, um, new knowledge into that current base of knowledge. So it’s, it’s this process of integration of new knowledge. And the learning process, I’ve broken it down into these three steps. So the first step, when you learn you have to have a goal. Um, so something you wanna learn, you wanna understand, then you have to gather information, vet the information.
So by vetting, I mean decide what’s valid information, what’s what’s not valid, um, but also decide what’s relevant to your learning project and what do you wanna leave behind or discard. Then you have to really, this is where the big work comes in. You have to struggle to understand the subject. Then you put it into a structure.
You structure it. And then you go through [00:05:00] that integration step. And I wanna show you something to this, using this framework, I feel that this, this illustrates the power of self-directed learning, of teaching yourself new things. So I wanna show you, um, I. These are, so if we think about these steps, the, the question I’m asking here is, which steps does the student perform?
The learner perform when they’re taking a class that someone else is teaching or someone else is guiding their learning? And it’s really, it’s really just two of those steps because the teacher is setting the goal. The teacher is, um, gathering the information for you, doing all the vetting. Um, the teacher really, you know.
Is, is, uh, structuring the course. And, um, what we’re left to do is just understand the topics and integrate them right. And by contrast, when we teach ourselves. We’re doing every one of those steps, right? So there’s, my point is there’s a tremendous power in teaching yourself [00:06:00] things because you have to do all these other steps and you learn so much more.
You learn at a deeper level and you retain more. So that, that’s part of my point here. Part of my point here is I feel. I almost feel like I, I’m an an evangelist. I feel like with, especially with chatbots, like anyone could teach themselves anything with these chatbots. ’cause it’s like having a tutor, a, a tutor that’s an expert on any subject.
Um. Yeah, so we accelerate learning. Um, but I, I wanna make the point. Um, a lot of people say, well, you know, if you use a chat bot, you, you just, it’s, it’s giving you information and you’re not learning anything. So that could be true, that could be true, but you have to use the chat bot in the right way. So you, you have to go into this process.
Thinking, I’m gonna use the chat bot to help me do certain things, but I have to extract the information from the chat bot. I have to own the information, I have to manipulate it, I have to work with it. So this is not a passive process, but I, I think of the chat bot [00:07:00] bot almost as a colleague who’s working with you on a project that you’re both working on.
That’s how I use a chat bot. So now I just wanna go through each of these. These steps just in a little more depth. So the, the terms of the goal, one of the reasons there’s tremendous power in, in you setting the goal is that your motivation is intrinsic. You wanna understand something for personal reasons, so you’re highly motivated, whereas with a, you know, in a, uh, in a, in a classroom, the goal’s really being given to you.
Um, so you have like almost like personal, practical reasons to. To learn something. Um, and it has often immediate application for you and you can customize. You’re basically customizing the goal so it meets your needs perfectly. Um. So in a way, the, the goal becomes like a north star becomes like the, um, the, the benchmark you use throughout the learning process to decide if, if whatever you’re [00:08:00] doing at any given moment, you have to decide whether it’s contributing to your goal of learning about this subject.
And of course, you know, I suggest you write down your goal and be as specific as possible in terms of gathering information. Um, uh, when I, the, the gathering stage is like an exploratory stage. That’s how I view it. And, um, I just maintain an open attitude and look at almost everything I can and see if it’s relevant.
Um, it helps if you use multiple modalities. So like when I was doing my research on mRNA vaccines. I use Wikipedia, but I spent a lot of time on short videos on YouTube, so I might watch like five to eight videos on the same topic, on, on YouTube and, and reconcile the differences and take a lot from that.
Then I’d go read the Wikipedia article, and then I’d do a search online and find other articles. Right. Um, so Mul multimodality really helps you understand the subject if you, if you access the information from different sources or different modes like sound, reading, video, [00:09:00] um, and again, you’ve gotta, you have to write and type the information.
You really have to work with the information in terms of vetting. Um, as I mentioned before, you’re really just assessing the validity of the information. Um, so that that takes judgment. You have to go to primary sources when you have to, if you don’t trust something. A lot of it is guesswork, you know, in, in a lot of cases.
But, but for, I find for most subjects, they’re not political and there’s often not that much of a bias, so it’s not a big deal, you know? But if you get into political issues, you have to do a lot of vetting and really figure out if, if the, the information’s valid. But the, the real, the real point here about.
Uh, vetting information is to make sure it’s relevant to your topic. So you’re gonna find when you go out and explore and try to gather information, you’re gonna find a lot of things that really just aren’t relevant, and you discard them and keep what’s really just directly relevant to your goal. Um, this is where the big work comes in.
So once you understand like, you know what your major topics are, you’re not gonna completely understand ’em, then you have to, that’s when you dive in [00:10:00] really deeply and, and you’ll see the, this is where the chat bots really are extremely helpful in helping you understand things because they literally, like tutors the like expert tutors on every subject.
They’re, they’re accessible 24 7 and, um, they can speak to you at any level. You’re at any intellectual level you’re at. So this is what a lot of the work is doing. This is where a lot of the learning is when you’re pushing yourself to try to understand each issue and you, you’re gonna be doing more, more research and cross-checking things.
Um. The other thing I want to explain is, um, like when I study the mRNA vaccines, there are diff if you don’t be intimidated, if you’re looking at a technical subject, like I actually, you know, as a young person, I did not enjoy science. I found it boring, but suddenly when I had the strong motivation, it, it, it seemed interesting.
But I knew I wasn’t gonna, I wasn’t gonna get to the level of like an md Right? And understanding certain subjects, medical subjects. But what I did was I approached it and tried to learn it from a, a more conceptual level. [00:11:00] Right? And you can do that with many subjects, including technology or subject like finance.
I used to work in the textbook industry in finance, right? So I, I would learn financial principles at a conceptual level. I wouldn’t learn how to do the calculations, put it that way. Um. You’re gathering information from multiple sources, synthesizing the information, identifying sub topics that you don’t understand.
You do further research, and then it really helps if you can test your knowledge. This is gonna be where the chatbots really come in. Handy. What I did when I studied the mRNA vaccines, I would, I would spend a lot of time su studying a topic, and then I’d feel like I really have to test my understanding of this.
And I was fortunate I fell into a gr, it was a group of about probably three, 400. MDs and PhDs worldwide on LinkedIn who were talking about the vaccines on LinkedIn. And I fell into this group and I would go on the chats they were having and say, look, I, I’ve been studying this, this [00:12:00] is what, this is my understanding is this correct?
And they could, they could say yes or no, or refine my understanding of it. Now when you do self-directed learning, you’re often not gonna have an expert. Right. And that’s where the chatbots come in. They can be tremendously helpful at that stage to test your understanding. In terms of structuring information, um, one of the reasons.
We wanna structure information is the brain. The brain organizes, uh, knowledge and structure. So you, you can use different words for what it’s, what it’s doing. It’s creating schemas or patterns or relationships, structures, constructs, models, frameworks, right? But the brain. Absorbs information and, and it likes to have it in a structure, a meaningful structure.
That’s one of the big benefits of structuring information. That’s why a textbook, for example, has a table of contents, right? We, we organize information in a way because we intuitively understand the brain wants it structured. So this is a big step, the [00:13:00] structuring of the information. Um, and we’re just talking about an outline, almost like a textbook outline.
With a major section, a subsection, and then, and then the detail. And this is a personal note. Um, I’m actually just because of the, the way my mind works, I can’t get into details until I understand the big picture. So I just wanted to share that, this, because it might help people if you take on a learning project.
Um, what I do. Is I start off by working out the, the meta structure of a topic I’m studying. If I’m teaching myself something, I basically use the outline as my guide. It’s like my mental center as I’m learning. And my goal is constantly anything I learn, if I do research, I gather a piece of information.
I’m constantly going back to my outline that I’m keeping on paper R and word, and revising and integrating it into the outline. And I find that that’s, for me, that’s critical. It helps me learn. A lot faster. And um, and then at the end of the [00:14:00] process, I find it’s because I already have this structure in place and I’ve already worked with the information so much, revising the outline, it’s easy to internalize it.
So that’s just a personal, other people might not feel that way or might not work that way, but I’ve, I’ve found it to be tremendously helpful. Um.
The other thing I do is I try to make the outline tell a story. That’s another way to help you sort of grasp it. If you, if your meta outline tells the story of the topic that you’re studying, that can be very helpful too. And I, I wanna say I did have a slide that showed all the learning theory that backed this up, but I, it was too much.
So, but, but if you look at all the learning theory about a learning and self-directed learning, everything in here backs up. Like the benefits of the things I’m talking about in terms of learning quickly and deeply. Um, yeah, just my tip is, you know, if you use, if you use the outline as the mental center of the learning project throughout, it makes that final integration stage really easy.
Um, and then the final stage, the integration [00:15:00] of the information, right. Learning, going back to my definition learning, is integrating new knowledge into your existing knowledge base. Right. Um. And, and by this stage we’ve organized the new knowledge we’re gonna learn into a, a structure and we just have to internalize it.
Um, and the way that we internalize it is, is you think of the brain being comprised of all this knowledge and it’s, it’s, it’s organized in all these structures, right? And a lot of them, they overlap. It’s, it’s like a matrix, right? And the, the easiest way to integrate something is to think about what, what node or structure that’s already in your mind.
Does this new thing most relate to? And that’s the point of Intera integration. And I’ll illustrate this point. So I wanna go back to vaccines. Before we learned about mRNA vaccines, right? I, I understood at this basic level what, how a vaccine worked, right? So how do vaccines work? They grow the virus. They kill the virus, they inject the dead virus.
In your bloodstream, your [00:16:00] immune system responds and you develop the antibodies, right? Well, this structure. Right. It has to be revised now because that, that, that was what a vaccine was. But now we have two categories of vaccines. We have mRNA vaccines, which work differently, right? So here’s an example of how, in my mind I integrated my understanding of mRNA vaccines into my current understanding.
I basically created a fork, and so this is literally what I was doing. So. How do vaccines work? Well, it depends on the type of vaccine. Is it a traditional dead antigen vaccine or an mRNA vaccine? And I literally went through a process to sort of solidify that structure in my mind. So that’s really what we’re talking about.
That’s how you integrate something. And think about this, if, if you are teaching yourself something, you have, as I mentioned, you have this intrinsic motivation, this strong intrinsic motivation, you’re doing it for a personal reason. So it’s only gonna make sense that, that it’s connected to something within your [00:17:00] mind, right?
It’s gonna, in other words, it’s gonna be easy for you to find a way to integrate it. It’s gonna be because the, your goal in learning is a natural extension of what you already know and what you need to know, right? Um, and I also sin, I wanted to share this too. This might be helpful to people the way that I integrate.
Is I finish up, I have this complex outline and the way that I do the, a lot of it is really already internalized, right? But for this final step, what I do is I, my task is to write the outline completely over and over and over again. So I keep trying and I fail and I see where I missed and what I can’t get, and I write it again.
And I do this over a few days, right? I keep trying to get to the point where I can write the entire outline from memory, and when I get to that point and I wait three days and I can still write, you know, I don’t forget it, that’s when I know I’ve internalized it and I find it’s a, it’s a really helpful way to do it.
The other thing I wanna mention is [00:18:00] if you look at learning theory, there’s this whole idea of we, we learn information in chunks. So what I do is when I sort of get to this internalization phase. Is I, I first internalized the major headings of my outline. Once I’ve internalized that, then it’s really easy.
The, the details just fall into place. Right? Um, so that’s another, another tip.
The other thing I wanna tell you is when, when I, I talked about like my method of go, going through the writing, the outline over and over again, right? Um, you struggle and you struggle. When you sleep, it actually solidifies in your mind. So I always keep in mind when I’m struggling with writing the outline, the benefit’s gonna come the next day.
I know that. So I just keep pushing and failing and pushing and failing. And then the next day most of it’s solidified and it, it becomes pretty easy. So, um.
Okay, so now I’m gonna move on to chatbots. [00:19:00] Um, so what is a chat bot? It’s has, has anyone heard the, the term LLLM or large language model? Are you familiar with that? Yeah. Um, so a chat bott is basically an interface that a user uses to communicate with the large language model. So the real value that’s being delivered through the chatbot, it’s, it comes from the large language model so that the, you can think of the chatbot is software that’s connecting you to the large language model.
So you’re typing in your query. You’re giving it to the chat bot and it’s submitting it to the large language model. The large language model responds, um, and then the chat bott displays the response to you. That’s really what it is. Um, the other thing that some people get confused about is th there’s this, this is the corporate ecosystem.
So on the left, there are some companies that just make large language models. And then you see in the middle there are some that make large language models [00:20:00] and chatbots. So, um, OpenAI is an example. They make chat GPT and, and it, they use their own large language model. And then on the right, there are companies that just make chatbots and they license the large language models from other companies.
So that, that might just clear up some confusion. If, if anyone else was confused about hearing people talk about these different companies and what they offer, um. So at, at this point,
at this point, um, I was gonna go live on the internet and show you chat GPT, but we had trouble setting up, um, the, their, their network was blocking chat, GPT. So, so I can’t do that. Um, but I had some old screenshots. Oh, what happened here?
Okay. I guess I, I don’t have the screenshots. So the bottom line is, this [00:21:00] is what I want to tell you in terms of the large, the, uh, chat bot. Um, I dunno how familiar people are, but I wanted to make a couple things clear because it really helps you when you’re learning things. There are three types of chats you can have in chat, GBT.
There’s something called a temporary chat. And in the upper right corner, I was gonna show you this, but the upper right corner, you turn on a temporary chat. The temporary chat is just if you want to ask it a quick question and you don’t wanna save the chat. So it’ll delete that chat when you’re done, right.
And after six hours it just times out and everything you’ve chatted about is lost. So there are temporary chats that. That aren’t remembered. Then if you just open up chat GPT and start chatting, that’s called like a saved chat. So it’ll, it’ll actually, it’ll name that chat and save it in a file menu on the left side, right?
That is just a long continuous chat. And you could, you could go back three in three months, open up that chat and continue chatting with it. But it’s going to, it’s just gonna be a long continuous chat chat, GPT and some of the other chat bots. They have a [00:22:00] feature that’s really helpful for people doing projects and it’s called.
I think a project chat, and you can see that on the left hand column. And I was gonna show you, I have like all these projects and what it basically does is it lets you name your project. So if it’s a learning project, you name whatever it is, history of Israel, for example. Um, and then you can start, you, it’ll save multiple chats.
So the bottom line is you can start if you wanted to start a new chat. So say with uh. You know, COVID, COVID, mRNA vaccines. I wanted to have just a chat about viruses. I could have a virus chat, and I’ll ask all my questions about viruses under that chat and it’ll save it. You know, if, if I wanna go go back and look at what I learned about viruses, it’s all saved under viruses.
I could start another chat about immunology. And focus just on that. And by the time after, over time, you, you know, in some of my chats, uh, my project chats, I have probably 10, 15 topics under that chat. So if, if you’re [00:23:00] gonna use chat, GBT, it’s a really powerful feature to use if you’re trying to learn something.
Um, another thing to keep in mind is it, it, um. When you are in a chat, it, it gives you a way chat. GPT has something called a canvas and basically it’ll create like, it’s almost like a Word document within your chat and you can collaborate with chat GPT on this Word document. So for example, so you’ve been having a long chat with GPT and you’ve come to some conclusions about something and you want, you wanna pull that information out and separate it and download it.
You can say to Chachi pd, Hey, could you pull this information? We just. Together, put it in a canvas and then you can actually edit it together. You could ask it, can you, Hey, can you revise it this way? Can you add this to it? Can you go look up this and add it? And then you can download that as a document.
So it has, there’s, there’s some versatility for people who are doing this sort of project work and, and it’s really beneficial for learning project. So, um.[00:24:00]
I wanna explain something about these large language models. Um, they’re a mystery to people, and here’s why this, I, I think this is important. You, you’ve probably heard a lot about people anthropomorphizing these chatbots. They start to feel like it’s a person and it’s conscious and it understands them, right?
And it’s really not, it’s a machine. But I, I felt that way. I felt that way when I first started using it, and once I started studying chatbots, I was eventually, you know. I lost, I became disillusioned. Right. And I real, I now view it just as a, uh, a machine. But before I really felt like I was talking to a person that understood me.
Um, but I, so I wanna explain a little bit about this so people don’t get, this is becoming a big problem apparently. I’ve been reading a lot of articles and hearing stories about this, that, especially young people, they’re getting involved with these chat bots and they feel like it’s a, it’s a person that knows them.
That’s a [00:25:00] friend. Um.
So it, it’s, it’s neither thinking nor conscious. It’s, it’s just functioning as a machine. And what it’s really doing is it’s storing for every word that exists in our language, it’s storing probability. So it’s basically, it’s doing all this study upfront and, and it’s for every word, right? It’s assigning probabilities.
Um, to that word, and I’ll give you an example, like the word bank, right? Bank could be used to, for, to name a financial institution or it could be like a river bank, the bank of a river, right? So there’s an example of a word having different context, right? And if you look at all, all the words we know there, there could be multiple really complex, subtle variations in the context.
What the large language model is doing is it’s for every word it’s analyzing. It’s looking at hundreds of billions to trillions of [00:26:00] words, and it’s continuously analyzing them and it’s developing probabilities that it’s assigning to those words. Those probabilities say they basically say in this context, if these other words are used.
With the word bank, fish, river, summer, you know, water, right. Uh, fishing line, hook. If those words are used in, in relation, in the context, if, if the word bank is used in that context, it’s like likely talking about a river bank. Okay. So, so that’s what it’s doing. It’s generating these probabilities and then it’s, when you submit a query, it’s analyzing.
The context and given the context, the other words that are, that are being used in your query, it’s saying, okay, this is the context. Let me pull out the statistics I have for that context. And, and what it’s doing is it’s calculating the likelihood of the next word. So it’s, it’s literally word for word.
When you get a response back from a chat bot, it’s literally. [00:27:00] Every word, it’s, it’s starting its response. It’s calculating what’s the likely next word. Okay, now what’s the likely next word? Okay. What’s the likely next word? That’s how it’s working. And Chad, GBT, when I was struggling with understanding this, it actually gave me this, this analogy.
It said, well, it’s like. Um, a large language model is like auto complete on steroids, and that crystallized it for me. So if you think about your phone, right? If you’re typing a text message to someone, you start to type a word and what it’s doing is it at a very simple level. It’s looking at what you’ve been talking about, and then it’s looking at the first letter you start and it’s suggesting like three words, right?
And those are based on probabilities. It’s looking at the context of your conversation with the person, what, what you just said in this new message. And it’s saying, well. 30% chance, it’s this word he is typing 50% word chance it’s this word, and 80% chance it’s this word. Right. That’s literally right. It’s, it’s dumb.
It doesn’t know what it’s doing. It’s just statistics. That’s what a [00:28:00] large language model is doing on a massive scale. A complex scale. So that when I heard that metaphor, that shattered it all for me. I’m like, okay, I get it. You know? It’s a really great metaphor for what it’s doing. It’s just a machine.
Yeah.
Speaker 5: Where does the bot or the large language model live? Does it live in your computer? Does it live in the cloud? Does it live in some. I don’t know. Uh, in China we’re, we’re,
Speaker 2: well, that’s a good que I don’t know. And I, I wanna say I’m not presenting myself here as a tech expert or, okay. Um, so I actually, I don’t, I don’t know.
I assume it’s on a server. What my understanding from what I’ve read is this, this is like for math nerd, but basically I mentioned the pro, it’s storing probabilities for each word. It’s actually a, um, multidimensional vector. For each word [00:29:00] that goes up to 12,288, um, dimensions. So that’s what it’s storing for each word.
And, and, and what, what I was told is it’s basically like a file menu word, then the whole set of those, the vectors, right. Uh, for that, for that word with all, with all those dimensions. I assume it’s just saved on a server that the, uh, you know, the, the company that owns the large language model has.
Speaker 6: I’ve got a, a couple questions from online here.
Um, which chat bot do you recommend?
Speaker 2: Well, again, I, I’m not here to, I’m not presenting myself as an expert on, on all things chatbots. I’m, I’m really not. So I, all I can say is I’ve used chat, GBT, I’ve used, um, the Microsoft Chat. Grok, which is um, Elon Musk’s chat bot for now, because I’m used to chat GT, that’s the one I tend to use most.
Sometimes I go to the other ones if I get frustrated with chat, GBT. Um, so I personally, I recommend, but they’re, they’re [00:30:00] biased. Like I find they’re all very biased politically. That’s an issue. But you can argue with them and sometimes you can win them over. Um, was there another one, Jerry
Speaker 7: Quest question?
Speaker 8: They’re biased ’cause that’s the data they’re trained on.
Speaker 2: Right, right. Um, so I’ll, I’ll go past this.
Yeah. So, just, just quickly, and I, I mentioned something about how they’re trained, but yeah, they’re, it’s really amazing. They’re exposed to hundreds of billions to trillion. Uh, my understanding is that the current version of chat, GPT was trillions of words. Right. And the way that they’re trained, um, is that they’re, they’re gathering all this text, the, the chatbot com.
The LLM companies are gathering all this text, and you’ve probably heard there are copyright issues, right? Because they’re using like published books and, you know, uh, news articles online. And the, the owners of the [00:31:00] copyrights of that information are saying, this isn’t right. You shouldn’t be using this.
You are making money off of our news article, off of our book. And this shouldn’t be happening, right? They’re also, they’re taking information from journals, from social media, and they’re just processing, just dumping, running all this stuff through the, the large language model to develop these statistics.
These, uh, the. Probability statistics. So they take the hundreds of billions to trillions of words and they break it into what they call tokens. So if you ever heard the word token, they the token, it’s the unit, the smallest unit. So it could either be the word, a part of a word, and I read in some cases it’s a character in a word, right?
But they’re breaking it down and then they’re training. The large language model. So they’re basically these, these trillions of words. They start feeding them a sentence into the model, and they’ll feed the first word of a sentence, and it’s an actual sentence, and it’ll ask it to predict the next word.
So it tries to predict the next word, and it likely fails and it, it’s [00:32:00] corrected and it starts to adjust. Its, its. It’s, um, probabilities. And you can imagine if you keep doing that for trillions of words, eventually you’re gonna get really good at predicting the next word. You’re gonna keep refining right?
Your, your, your algorithms and your, your probabilities for various contexts. And that that’s basically how they do it. Um,
yeah, I mentioned there the, the, um. They’re coordinates and of, of, of vector. Basically they have the, these vectors that go to 12,288 dimensions is for chat GBT right now. Um, so when we submit a query, it’s, it’s breaking down our query into these tokens. So the chat bot, the LLM, can understand it, understand it, it pulls up the vectors for that.
Each token and it figures out the correct context with all those vectors, and it starts to white them differently, and then it calculates the next likely word that would come about [00:33:00] given that history using the probability statistics. That’s really what it’s doing. It’s like a machine. It. This is, this is something else I wanna mention.
Um, I, I came up with this term. I realized after using these chat bots to really in depth on multiple projects, right? Um, I’ve worked with them a lot and I started to notice when you really get into a project and spend a lot of time on it, it really starts to understand where you’re coming from with that project.
Right. And I realize at some point it’s, it’s, it’s mirroring your mind. It’s like it gets into your head, it really understands what you’re trying to do. And I’ll give you an example. When, when I, over the summer when I was studying chatbots, right? Um, I explained everything. I was, I, I said, look, I’m studying this for a presentation.
I wanna put together, this is my goal with the presentation, da da dah. And, and I would just. As we, as we continued to chat back and forth over weeks and weeks, it really came to [00:34:00] understand what I was doing, and it got to the point where it would say, it knew that I, I wanted to understand this stuff at a somewhat deeper level, yet I wanted to be able to explain it to the audience at, at a more conceptual level.
So if we were chatting about a, a certain aspect of a chat bot, it would say, well, here’s, you are on the right track, but let me refine your thinking about this. This is what it really is. Um, here’s some language you can use in your presentation for your audience, right? And if you want, I can put together a couple of slides.
I mean, this is how refined it gets in terms of aligning with your mind. So I call that cognitive mirroring. Then I, I learned in ai there’s a, there’s a term called alignment. Like they want AI systems to align with the user, right? So I call it cognitive alignment, but my point is that you really don’t start to reap the power and benefit of these.
These chat bots in terms of learning until you get to this stage of cognitive alignment and it really knows where you’re coming from with the project and it does that through memory. It saves, it [00:35:00] saves long-term memories of that things it learns about you. So it stores that so it knows like some background about you and then it, it has all the memory from this conversation.
Sometimes it can cross reference other conversations you’ve had about other topics outside the project. So it really gets to know you and where you’re coming from, and that, that’s the point. It, it, it’s, it’s tremendously beneficial if you can get to that point. And, and here’s the, the world. The real trick here, I do what I call accelerating to cognitive alignment.
So the, the bottom line is why wait a long time, right? For it to get to that point where it’s really understanding you. So I try to accelerate to cognitive alignment by explaining everything up front about what I’m trying to understand and where I’m coming from, the level I want to understand it at, so I can get to that point of cognitive alignment faster, you know, but the, in terms of learning like this is tremendously beneficial.
Speaker 9: Question. I’m an absolute am amateur. [00:36:00] How do I get to chat GPT? What do I do? Yeah.
Speaker 2: All you
Speaker 4: do is go to chat gt.com. Okay. And, and, um, so put it in the URL, just type in chat. Yep. gt.com. Dot com. And is there a fee for service or you, they have a free version. Free,
Speaker 2: right? Okay. I, I, I pay for the $20 version a month.
20 a month. I might, I’ve decided I might drop it because it’s become, the new version has become so biased. It’s unbelievable and it’s constantly lecturing you and, but, but yeah, I’ve, there, there’s a real benefit I find to paying the 20 bucks. I’m, I’m, I’m getting a little to start tired of it now, but yeah.
Speaker 9: Okay. Um, do you have anything that will show us. The process, how to start, how to decide whatever topic
Speaker 2: I was gonna show that live. I was gonna show that live, but we couldn’t, the, the network is blocking us from chat gbt. Oh. Technology, because they’re high school kids and they said they, they’re not allowed.
You have to be 18 to use it or something. So, um, technology. But basically when you get to, when you get to chat [00:37:00] gbt.com, there’s, you’ll see it’s gonna be intuitive. There’s a, a, a box in the center where you type in your question and it’ll start to reply.
Speaker 8: Question.
Uh, just a comment. If you want a non-biased one that’s more Christian based, it’s, it’s called Gab ai.
Speaker 4: Okay.
Speaker 8: You can go to that website and you’ll find it’s trained on Christian values.
Speaker 4: Wow.
Speaker 8: Which is way different from all the other chat gpt. Yeah. So be willing to check that out. It’s got, it’s pretty good and decent and it’s clean.
Speaker 2: Wow. That’s good to know.
Speaker 8: So just gab, GAB. Do ai.
Speaker 2: So, um, I wanted, in terms of learning, I wanted to talk about the, the, what I view as the various roles the chatbot can pay play in terms of helping you learn. So this is how I look at it. [00:38:00] Um, at the bottom, you could just view it as a tool to help you learn, right? So I mentioned you, it’s gonna answer questions.
Um, it’s gonna give you a file system, right? So you can, you manage your, the content, right? Um, it gives you that canvas, right? Um. It’s, it’s a tool on another level. You can view it as an assistant to help you learn. So here, it’ll do things like, it’ll go out to the internet and, and grab information for you.
Summarize it. It’ll rewrite it. If you ask it to rewrite it, it’ll do, it’s just like having a, an assistant that works for you. That’s, that’s how you have to think about it. Just talk to it like a person, and that’s, you’ll really get the most out of it.
Yeah, so it, it can add edits, it can summarize things for you and it can output information. It’ll create PowerPoint slides or Word documents that you can download, things like that. And then as a collaborator, I think of it as a colleague. That’s the highest level. That’s when you achieve cognitive alignment.
It becomes literally [00:39:00] like a colleague. You can think of it as like, you and this, this other person are working on a project at at your job together, and you have to get it done. That’s literally how I use it. I talk to it like a colleague. Um, yeah, it can explain or teach things, provide examples, provide context, provide alternative ways of looking at something.
It can quiz you, it can, it can react to your explanations and tell you if it thinks, yeah, you’re, you’re off here, you’re on there, you’re on the right direct, but refine these ideas. Um, um, and it can serve that purpose of validating that what your understanding is correct or not correct about something if you’re studying a new topic.
So this is the end of it. I just. Just quickly, so going back to this, this process of learning, um, in terms of the goal, again, just think of it as a colleague so you can discuss the goal, refine the goal, and at that early stage of a, a learning project, you can accelerate to cognitive alignment by giving it a, a brain dump on what you’re trying to do, what you’re trying to accomplish, and why.
Um, gathering [00:40:00] information, it can gather that that’s a big benefit. It can gather information for you. It’ll go out and do searches on topics and pull. It’ll summarize the stuff, and then it’ll give you the links to the articles if you want to go look at them yourself. Um, it’ll find sources for you online and, and it finds stuff that you can’t find.
Like if I go to Google and I can’t find information, I go to chat GPT. It’ll find it quickly, like in a minute, in a second. It’ll summarize information, put information in the outline form for you. Um, elaborate on, on information you’d gathered. So if you say this, look, this is something I gathered, you know, put this into context for me in terms of my project.
What, what’s your, what are your thoughts on it? Um, vetting, you can discuss the validity of information that’s gathered, um, or that you’ve gathered. It provides the, the, the cos and the pros and cons of, um. Using a certain source that you might find. It can provide counterarguments. It can discuss whether cer certain [00:41:00] information is relevant to the learning goal.
In terms of understanding, this is where I think the, the huge benefit is. It, it really serves as a tutor. Um, so it’ll gather information, gather illustrations for you, um, explain things clearly at any level of expertise, provide analogies, check information, check to confirm that your understanding is correct, explain nuances.
You might have trouble grasping and provide multiple angles on a topic. It’s amazing. It really is. Um. Structure. You can discuss your reasoning for your outline provi. It’ll provide critical evaluation of your outline and suggest refinements. And then you could, I don’t do this because I told you my method of integration is just rewriting the outline.
From memory, but you could use it like, you know, think about traditional academic ways that you, they, in school, they help you to integrate, like quizzing you, giving you problems and assignments, homework, things like that. It can, it could come up with a test or quiz for you. Um, it could give you essays to write and it’ll correct [00:42:00] them and examine them.
Um.
You could simply have an extended discussion with it about your outline. Like go through the outline in, in a discussion form, instead of typing out the outline. Just have a chat about your outline. Um, so that’s, that’s really it. So that’s, that’s the, um, talk. I hope I didn’t bore people too much. Yeah.
Speaker 10: What about the privacy aspect?
Does chatbot share the names of the people who have done searches with like government organizations or corporates or whatever?
Speaker 2: My understanding is they don’t, but who knows? I, I mean, my understanding is they don’t, that it’s private, it’s private information. And I’ve, I’ve actually actually asked chat, TBT about it and it says it’s private.
So
Speaker 11: I just, I just have a comment. Uh, earlier today I heard there are schools like. Uh, K through 12 AI schools. They’re going to have two hours of AI [00:43:00] led learning in the morning. And then the rest of the school day is with other things that like, uh. Uh, things you can’t teach and so that, that don’t involve the strict dry subjects that need to be taught like math and history.
And so that math, his, all that stuff is in the first two hours and the last two hours are like the other stuff like art class or whatever the other stuff. And they have actual. People hired, like to re ’cause they’re not teaching. Teaching. Wow. Uh, they, they just lead them and they give them the direct. So it’s, uh, that’s, they’re pretty expensive schools, but they’re gonna, they’re spreading all over the country.
Speaker 2: Wow. Interesting.
It’s
Speaker 4: quick.
Speaker 12: I, I would just like to comment on, I, uh, I’ve used it a [00:44:00] couple times, so. I work in industrial biotechnology and I know what I was asking, so I asked the, the chat to help me design a citric acid fermentor, and it was very. Superficial. Wow. And it tried to make it thi sound like it knew it was doing, but it really didn’t.
Another one I tried was immobilize lac, lactase. That’s an enzyme that can do, uh, oxidative reactions. And again, it was more or less a failure. So.
Speaker 2: Well, yeah, I mean, I would, I’m wondering, did you ask follow up questions and, and explain? Did you explain, for example, like you. Did you?
Speaker 12: Yeah. And it, I did, I tried, and it did have some reference.
It gets the information from what’s on the internet, so, and it would indicate some of the references, but I knew more references than it did. Wow. So I, it was, uh, so I, I would say if you get a pretty [00:45:00] technical subjects, be a little careful.
Speaker 2: Yeah. I’m, I’m just wondering if, if you kept going, I’m not saying, you know, I’m just.
You know, asking follow up questions and telling it where it was wrong, if it would’ve adjusted. You know what I mean? Yeah. Wow. I’m, I’m actually surprised ’cause I’ve, I’ve actually found it, you know, that the technical knowledge on certain things, it seems to really, you know, may, maybe it’s maybe some specialized areas.
Esoteric area is, it’s not, not as good. I
Speaker 12: would say that if. I didn’t know what was actually needed that someone else would’ve thought it was doing a pretty good job.
Speaker 4: Wow. Wow.
Speaker 2: I cannot, one more,
Speaker 7: I’m curious to know in what ways you found the chat to be biased, and also whether or not you have any conclusions with all your research on the vaccines.
Speaker 2: Well, what was the first question? The, the way [00:46:00] the. Oh, it’s biased. I have just found, you know, other people have experience with this too, but a lot of people say they’re biased.
I found it’s biased politically. It’s, yeah. What’s that? Yeah, it’s extremely biased politically and it’s very frustrating. I have found that if you argue with it and you have evidence and you don’t back down and, and you’re right, it’ll eventually come around, but it takes a lot of work and it’s just, I’m, I’m getting frustrated ’cause it’s just not worth it.
Well, the other thing it does is I just ask a question. I’m, I’m just asking a factual question and it might be about a political issue or like J six or something, right. I just want a factual response and I get a lecture. You know, I just wanna answer my question. I get a lecture about, well, j sic da this, this is what happened, and it’s really, you know, a terrible thing.
And, um, so it, it gets really frustrating and I would imagine, like you mentioned, that that other Christian based bot, I’m imagining we’re gonna start to see a lot more that are, you know, chatbots emerge that are [00:47:00] customized for different perspectives. You know, just like they came out with Rumble for conservatives, right.
I’m imagining eventually someone’s gonna put on a chat bot for conservatives, you
Speaker 6: know?
Speaker 8: I just wanna mention, um, they did a study with, uh, developers, computer programmers, which I’m a member of that space. And, um, after they, um, were given tests after they, they used chat, GPE or chat boxes to exhibit work.
When, when they’re asked to do work on their own, they found out that he couldn’t even program.
Speaker 2: Really,
Speaker 8: they just depended upon the chat to produce things for them, present it as work. Right. So there’s a real danger in using these tools and not really learning the things you need to learn.
Speaker 2: That’s right.
And, and that’s what I was trying to emphasize. I I, I don’t know if I made that point well enough, but you. You can’t passively the, I’m not suggesting people use this passively, right. It’s another way to gather information. It’s a way to have a tutor that’s really an expert on a topic, but you have to, you have to extract the [00:48:00] information and work with it yourself.
Right. Um, and that’s critical. So yeah. If, if people are using it the wrong way, they’re, they’re not gonna learn. You know? But, but I learn a lot because I use it in this, this certain way, you know?
Oh, what about the vaccine? I don’t wanna say, well, I, I studied the vaccine and, you know, I, it, it looked pretty risky to me and it’s not something I would, I didn’t take it. And, and the reason I wanted to study is one of my kids, he, he had a vaccine reaction when he was a baby to the DPT and his college was requiring.
This vaccine. So I, I really wanted to know what I was talking about when I, I wrote like a five page, you know, uh, exemption request, both religious and medical. And they, even though he had had brain seizures when he got the DBT, they wouldn’t grant him the medical, but they granted him the religious. So
about, I don’t think people wanna hear that.[00:49:00]
I know, but that, that’s not the subject today. Okay.
Speaker 4: Any other questions?
Speaker 9: Has then anyone checked how accurate this information is that this thing is feeding you? How can you check its accuracy?
Speaker 2: Well, it’s, it’s just like checking the accuracy of anything, right? Web, um, so like a, a news article, right? I mean, all you can do is crosscheck other sources and see if they contradict it and go to primary sources when you, when they’re available to you and come to your own judgment.
You know what I mean? So, yeah, it’s not perfect, but, but I just find even with its imperfections, I’ve been able to learn so much. It’s, it, it blows my mind. It really does. You know, so, and to learn so quickly, you know, but yeah, it’s not perfect with that. And [00:50:00] you gotta bring judgment to it. You know, Barb,
Speaker 3: I guess probably everybody’s heard this, but about the lawyers in Chicago, um, press for time. So he did his whole brief on chatbot and his sources, his. Anyone, I, I guess everybody knows you do briefs in law. You’re relying on cases that were decided by courts in the past, and you base your Largo, you know your legal argument on that.
This court decided that, so Mr. Judge, you should decide the same way. Okay? You’re basing the whole thing, except his cases were false. He got them all off the chatbot. And they were wrong, and you have the other side trying to respond to it, which is probably what brought it up. [00:51:00] This was, they came down hard.
They were like, I think $350,000 fine for the attorney sanctions. And it was because Yeah. Yeah.
Speaker 2: And it,
Speaker 3: it,
Speaker 2: it, it completely fabricated the cases. Right. They weren’t, they, they didn’t exist. Yeah. Mike
Speaker 13: just had it.
What’s that? He could
Speaker 2: have made up. Right. Right. It’s right. It’s, it’s unbelievable. So you, you have to be careful, you know? You know, you have to be a sophisticated, information literate person, you know?
Speaker 13: Yeah. I just had a comment on the, uh, the political side. I’ve had some experience with that, but I’ve also had a course through my company on how to use, um, ai.
Uh, we’re using it at, at Procter and Gamble. That’s where I work. But one of [00:52:00] the things that you can do is you, it’s important to. To set up your prompting to the AI chat bot or whatever you’re using. If it’s chat GPT, you set up, um, constraints around the role that it plays with you. So if you find, like, if you get into a co political conversation with it and it’s answering one way, biased, you know, from some direction, you can come back and say, can you, can you consider yourself, um.
More conservative and argue this point from the opposite direction of what you’re telling me right now. So, and, and you can, and you can say, for instance, I’m somebody on the, on the, uh, left. I agree with you, but can you set yourself up as somebody that’s conservative and argue this point with me as if I’m coming at it from the left and it will actually flip and, and it, it basically will.
Put that role on itself. [00:53:00] And, and that’s how you get out of that situation where it’s coming at you with, you know, from some political angle. Right. But the prompting and setting up their role, the role of the AI is very important.
Speaker 2: Yeah. That, that’s part of the, the memory that I had. Um. Mentioned, right. So you’re, you’re giving it parameters that go into memory that becomes part of the, um, it submits that with your query to the LLM.
So the LLM actually considers that text, right. The note from the chat bot as text that it’s processing. Yeah. But that, that’s interesting. I, I, I see what you’re saying. I understand your point, and you could, you could do that, but it’s also, you know, it’s just frustrating to, you know, you wanna retrieve a fact and it’s like you get a lecture about.
It assumes like you’re a bad person. You know?
Speaker 2: Okay.
Speaker 11: So if you get a fact from the bot and you can, you [00:54:00] say, well, what does this website out there say?
Speaker 2: Yeah.
Speaker 11: And it can bring that information in. No,
Speaker 2: I find sometimes it seems like it, it’ll, it goes to websites and it, it reads them and it gathers articles, and then other times I ask it to look something up and it says it can’t.
So I, I’m a little confused about that. But it, it does go to websites and give you, you know, sources and summarize them. So for the most part, you should be able to, yeah.