Full transcript of “🚨Learn AI From Basics | AI Agents | LLMS | Workflows | @tech.boy.deepak”. Click a timestamp to open the video at that moment. Need another video? Use the YouTube transcript generator.
0:00 Hi guys welcome to the A Learning [Music] Community Copilot Studio will allow you to create agents. The department of AI AGT in the future. Worklose diff. 2500 plus students have filled out the form and many people are very excited about what the AI Learning Community is going to do. Many people have left many comments. So yes, as I said, while shooting this video yesterday, I suddenly noticed that many students want to learn AI. So we
0:46 thought, what if we started an AI learning community and we were like a guide for them, helping them learn the basics of AI little by little and improve their AI skills? So that's why I'm starting the AI learning community . So I still don't have an answer to what happens in this AI learning community, what we actually learn and what we do. So what exactly do I need to learn in the next few days, a structured roadmap ? Where to start ? And what tasks should be given to these community members? How to track all of them? We're going to design a complete system, and in this video, as you can see in the thumbnail, let's discuss AI LLMs, AI agents, and AI workflows. So this is the first video of our AI Learning Community and in this video I am going to talk about AI LLM, AI workflows, and AI agents. So I
1:36 will explain it very simply using real-time examples and clearly in a way that all of you can understand. And as I said before, those who want to skip the video halfway should not watch it because I do not want to give you wrong guidance or misguidance . If you don't get a clear idea by watching the video until the end, you 'll be confused again. And what else do I need to say? After watching the video to the end, and during this viewing process, you should maintain short notes. In the short notes, you should keep taking notes on what I am actually saying, what their definitions are, and how far they go. And after the entire video is finished, I need some support from all of you . Like a picture of you watching this video and a picture of your short notes and if you want to write a post of your own, tell me how we can drive our AI learning community, how we can take it. I am expecting a LinkedIn post from all of you.
2:21 This LinkedIn post is very important and among those who have posted LinkedIn posts, I will select the best people and send them rewards on behalf of Tech Boy Deepak and let's dive into the video. The video has started. So, as I'm learning, I'm going to teach you in this video. So, let's take a look at the first topic and what we're going to cover today, LLMs and Second AI Workflows . Third, let's look at what AI agents actually are . So before talking about LLMs and LLMs, like we hear a lot of terminology about AI, like RAG, GPT, LLM, AI Agent, etc. Many students think that all these words are very hard, but all these are very simple and the
3:07 definitions related to them. Whatever terminology I am using, the definitions related to them are given here clearly, below in a very simple and easy to understand way. Read the definition and even if you don't remember, take a screenshot. So okay, let's look at LLMs first. And as I said, I'll learn and teach you, so okay, first of all, I'm going to GPT. Do n't forget the reason for the notes, keep them. Write notes, short notes, but very important explanations in simple terms with examples. Okay, so let me break this down and explain it clearly. Let me explain it to you. So first, what is an LLM? What is
3:54 LLM? Many people don't even know what LLM is in full form. LLM stands for Large Language Model. It is a type of artificial intelligence that can understand and generate human-like language like how I'm talking to you right now. So how does GPT talk to us? When we type, GPT talks to us. It explains it to us. So LLM is what it is, it is an artificial intelligence where it can understand human text, which means if we give it any message, it will understand it and speak to us like a human. So Chart GPT Gemini Cloud Graph all these are built on top of LLM. So how does it work? Imagine you are reading millions of books, websites, articles and so on. You will get super good at guessing what word comes next. That's what
4:39 LA does. This is very important. This is the basic fundamental of an LM. So what does it mean? We read a lot of books, we look at websites, we look at articles, we look at messages. So let me give you an example. For example, let's say we're watching a movie and we can guess a little bit about what's going to happen in the next scene . If there wasn't such and such a villain, we'd guess that this is going to happen, that is going to happen. So LLM also predicts its next word . Now here it says imagine, right? After imagine, this text predicts everything . So that's what an LLM does, this is very important guys, remember this, you are you will get super good at guessing what word comes next in a sentence. It reads
5:26 tons of text, that is, data from the internet and learns patterns in language. It understands the pattern in that language. So when you ask a question, when we ask a question, it can reply like a human put based on what it has learned. This LLM responds only on what it has learned. So, Chargé Pétit Google, GeminiSoft, Copa Pilot, Notion, Instagram DS Auto Replies. All these are LLMs and I will tell you clearly about LLMs and SoLLA. So let's see in a visualization. So, before watching the video that I made, don't forget to subscribe. It is very important to subscribe. So, we have understood the full form of LLM, right? Large Language Model. For example, if there is a human, he
6:11 gives input to a human. What does LLM mean? Chargept, Claude, Gemini. So these are LLMs. We are giving an input to this LLM. So this LLM generates the output for us . So the point here is that LLM cannot be generated on unknown data . For example, an LLM is only awarded based on the data on which it was trained. Events Happening in Hyderabad in 2024 So I am giving you that, so what tech events happened in Hyderabad in 2024, so if you look at the list, it will give you everything, major tech conferences, AI Summit, Future Technology Show, and their locations and venues . So how does it do it? It
6:59 connects to all the news articles and websites, reads from them, and gives us updates . Now let's ask another question, what are the events that happened in 2020? So when asked what happened in 2050, he said yes of now 2050 is in the future so no revealed events happen at all so he didn't say that either. It is trained on data related to what happened in the past and gives us updates related to it. So it is only related to what we have trained on. That is what we call an LLM. It is as simple as that. So I thought I should mention another important point about this LLM. The first
7:46 one comes and whether it is a chargept, a Gemini or a Claude, it gives us answers only on the data it has learned. So the second behavior comes and we only give a response when we ask a question. When we ask, when we put, it gives us an output, but it doesn't automatically respond to it and say, "I'll tell you myself ." Remember these two behaviors. They will definitely help you later. Just make a note of these two behaviors. And the next topic is AI workflows . [Music] So I'll explain AI workflows very clearly . Last time we looked at LLM, what does LLM mean? If we give an input, it gives an output. So let's see what AI workflows are. I will explain SOAI workflows with a very simple example. It will be very clear to you.
8:32 Compare it with the previous example. So how can there be a human being, we give an input to LLM, what is LLM, and again, if the charge is Claude or Gemini, that LLM is not the output, what happens in the EAI workflow, that is, we integrate it. For example, if a charge asks me what events I attended in 2024, will it tell me? Let's test it once, what are the events that I attended in 2024? If I ask what events I attended, then it will analyze my LinkedIn and tell me. It doesn't know the story itself. So
9:18 what's happening here is that this charge goes into my LinkedIn and reads everything I do, whatever events I attend, and tells us. This is where workflow comes into play . So here I have a workflow where I show the calendar as a workflow for example, which tells us what LinkedIn is called LM, so this LLM connects to LinkedIn and tells us what events we have attended . For example, suppose I don't have LinkedIn. This chart doesn't say what GPT answered. It doesn't know what events I attended . So, if we make a note in our calendar of what events we attended, and then connect that calendar to this LLM, it will tell us what events we attended . So look here, I have set up a
10:05 calendar as a workflow. So if you look here, if you connect that calendar as a workflow to that LLM, that chargeptr cloud, that LLM, will read the data in that calendar and give it to us as output. So this process of connecting this calendar is called control logic. Remember this. So what is control logic? In this workflow, whatever tools we integrate with and train data from, that process, that logic is called control logic. So this is the simple example for AI workflows. So what is the difference between an LLM and an AI workflow? An LLM tells us our past data, not our future data, but
10:50 what we did in our past data, if we train it . So what is EAI workflow? When we integrate this LinkedIn or Calendar with it, the workflow in that sense is trained with the data in that calendar and our output is what events we attended in 2024 . This is what a simple AI workflow is all about, and remember that logic is called control logic. And before moving on to the next topic, guys, don't forget to subscribe. We already have 13.5k 0.5k subscribers. And your subscription is very important to take this e- learning community further . So if you're already subscribed, it's very important that you like this video and comment on how you feel about it now that you've finished watching it . So the YouTube algorithm works in such a way that if there are
11:35 more positive comments and more people like it, it will reach more people. So like and comment for that. Let's definitely reach a lot of people. Let's help more people learn AI. So the next topic is AI agents. So what are these AI agents? In this article, we will discuss AI workflows. There are many different websites like Make.com or Rapier. In the last video, I clearly showed how to build an AI agent. If you missed that, watch it too. So what is an EAI agent? How do we link a calendar? How do we link LinkedIn? What control logic do we think of in an AI workflow? All that is thought of by the EAI agent. So EAI agent for example as a human we give a prompt
12:20 to an LLM the same charge Claude or Gemini so what does EAI agent do is that depending on the response we give it where it should go that agent will do according to the prompt we give like for example like last time AI agent did what we said i.e. go to and search for tech boy Deepak and give me the output like that. So, we went to Google and searched for "Who is Tech Boy Deep?" and got the output . So this is all done by an AI agent . In an EAI agent, we just give a prompt saying, "This should be done, this should be done," and that prompt can also be re-issued by an EAI agent. For example, suppose we replace the human here and put an AI agent here, duplicate it, give the AI agent a prompt, and then the AI agent thinks again and
13:06 gives the same output. If the output is not correct, the AI agent will tell the agent again . So here, what AI replacing AI tells an AI agent is that this output is not correct, give it like this, give it like this, it tells an AI agent, thinking like a human . The AI agent then generates output as a response . So this AI agent generates an output without human intervention, without our intervention involvement, and next in the future, artificial general intelligence (AIGI) is going to generate an AGI, right? So what does AGI mean? It's like completely human, it's like you're talking to your friend. So in this video, we discussed three topics: first, the upcoming LLM, second, the upcoming AI
13:53 workflows, third, the upcoming CAI agents, and every time we mention that terminology, we clearly put the definitions there. Look at them and understand them clearly. Don't miss out on the short notes that you're told to mention. And I'm asking a question, so which of these three is going to replace us, LLMs, AI workflows and AI agents ? How is LL? AI workflows, AI agents? What will replace us in this mood ? Leave a comment explaining why it's being replaced. I'll see. In the next videos, we'll learn more basic things related to AI, gradually upskilling ourselves, and move to the advanced level . So here you see, it says Build Your Own AI Agent, right? This video is about building your own AI agent. Watch this video and
14:40 you will completely understand how to build an AI agent. And if you haven't subscribed, please subscribe. Comments are very, very important. If you want your friends to be part of this AI learning community, tell them to also share these videos with the AI learning community . These 2000 members were the first AI learning community I planned for 1000 members, but I received 2200 responses. So I think I'll give access to all of them. All the best and congratulations to all of you for being a part of the AI learning community. So, in this auspicious moment, put up a nice LinkedIn post and say, "We are the part of AI learning community and we are learning together with Tech Boy Deepak." I am very happy that you are also upskilling and we are also reaching many people . So, in this video, I think I have explained it very simply and clearly. In the next videos, I will come to you with a new topic and
15:29 talk about AI again. Let's learn the basic things and move forward by learning and upskilling. And signing of Tech Boy Deepak Thank you.
