What is AI?

0:00 Today, we ask ChatGPT. We generate images with AI, edit videos with AI, write emails with AI, and even prepare for interviews using AI. But here's the interesting part. Millions of people use AI everyday. Yet, very few actually understand how it works. If you don't understand the technology that's changing every industry, you'll always be limited to using it rather than creating with it. That's exactly why this course exists. Welcome to this complete artificial intelligence course by TelliPat.

0:36 Whether you're a student, a working professional, or someone who's simply curious about AI, this course is designed for you. You don't need to be an AI expert to get started. We'll begin from the very basics. What AI actually is, how it learns, and why everyone is talking about it. Then we'll gradually move towards machine learning, deep learning, neural networks, computer vision, natural language processing, large language models, generative AI, AI agents, and much more. But this won't just be theory.

1:00 We'll work on projects, write code, and understand how AI is applied in real-world products like ChatGPT, recommendation systems, voice assistants, self-driving cars, fraud detection, healthcare, and many more. By the end of this course, you won't just know how to use AI tools. You'll understand what's happening behind the scenes and gain the skills to build AI-powered applications yourself. So grab your laptop, open your code editor, and let's begin your journey into the world of artificial intelligence. Artificial intelligence has revolutionized the way people think, learn, and work in various fields, from finance to healthcare and mobile apps.

1:38 What's more interesting is that AI plays more of a role in our daily lives than we can imagine. From Siri and OK Google to various virtual player games and social media apps, AI is everywhere. It sure is the most happening topic in every business right now. It is the most wanted and exciting career domain right now in the market. Artificial intelligence or AI uses intelligent machines that build in a way that they react like humans. The primary process involved in making these smart machines is to carry out decision-making, which analyzes and uses data available in an enterprise.

2:17 It is similar to the human mind absorbing and synthesizing information and providing the required decision. So how does an AI work? AI systems work by investigating large amounts of data, labeled training data, analyzing the data for correlations and patterns, and using these patterns to make predictions about future states. In this way, a chatbot that is fed examples of text chats can learn to produce lifelike exchanges with people, or an image recognition tool can learn to identify and describe objects in images by reviewing millions of examples.

2:49 AI programming focuses on three cognitive skills: learning, reasoning, and self-correction. Learning processes. This aspect of AI programming focuses on acquiring data and creating rules for how to turn the data into actionable information. Reasoning processes. In this aspect of AI programming, it focuses on choosing the right algorithm to reach a desired outcome. In self-correction processes, AI programming is designed to continually fine-tune algorithms and ensure they provide the most accurate results possible. So there are four types of artificial intelligence: First, we have reactive machines.

3:19 These AI systems have no memory and are task-specific. An example is Deep Blue, the IBM chess program that beat Garry Kasparov in the 1990s. Then we have limited memory. These AI systems have memory so that they can use past experiences to inform future decisions. Some of the decision-making functions in self-driving cars are designed this way. Theory of mind is the third thing. It means that the system would have the social intelligence to understand emotions. And finally, we have self-awareness.

3:56 In this, AI systems have a sense of self, which gives them consciousness. Now let's see what are the applications of AI. One of the best known healthcare technologies is IBM Watson. It understands natural language and can respond to questions asked for it. AI can automate grading, giving educators more time. It can assess students and adapt to their needs, helping them to work at their own pace. AI tutors can provide additional support to students, ensuring they stay on track.

5:32 Banks are successfully employing chatbots to make their customers aware of services and offerings and to handle transactions that don't require human intervention. AI technologies are used in transportation to manage traffic, predict flight delays, and make ocean shipping safer and more and more efficient. Just a quick info guys, Intellipath offers artificial intelligence course training, which will enable you to learn skills like neural networks, focusing on CNNs, RNNs and their application in AI. Natural language processing, computer vision and generative AI.

6:07 Through this course, you will also master deep learning concepts, AI model building with TensorFlow, Transformers, BERT, GPT, Langchain, and NLP pipelines with advanced techniques like image recognition, chatbot development, and much more. All taught by top 1% experts and IIT faculty. And with this course, we have already helped thousands of professionals make successful career transitions in AI and data science. You can check out their testimonials on our achievers channel. The link is in the description below. And without a doubt, this course can set your career to new heights.

6:43 So visit the course page link given below in the description and take your first step towards your dream career growth in the field of artificial intelligence. Let's start from here. So you will see that, uh, you will see that guys that there are some handwritten digits. So this is a very famous dataset coming from MNIST dataset, right? It's a handwritten digits and generally these are in what? 28 by 28 pixelated form. I'll show you the real data later in the course.

7:18 But although the scale is not very obvious, but if you see this is 20. The last is 28. So basically imagine that every digit falls in like a grid. Okay? Something like this. Since it's a pixelated form, right? It looks like 28 pixels by 28 pixels. Okay. Every number, every digit. So suppose you say three, it looks like that. So this is coming from real data. You will see. This is a very nice image, although it only shows you till.

7:59 Can you guys see that? I know it's a little tiny, but let me zoom in a bit. You will observe. So it's a matrix of, you know, starting from zero. Till 28. If you start from zero to 27, how many, uh, rows guys, come on. This much array, everybody has studied. 28 rows and similarly guys, if you observe 28, although I'm not able to show the whole picture, right? But if you guys focus here, you will realize that in each cell, there are zeros and there are numbers, right?

8:37 And generally, this is zero to 255 digits. Or 255. So if you do the pixelated grid, it will go from zero to 255. As you can see here. So as you go higher and higher, the darker color will come. It can be black and white depends on your setting, right? So generally it's a digit, right? It will be black and white. Right? But there are also shades in black and white, you know, grayish, dark, light, right?

9:10 There are a lot of colors. But if you try to shade it, right? This particular pattern, can you see wherever the numbers, if you try to shade it, right? And where it is, you know, if you try to do it, you will see some edges are coming out. If you are able to pay attention, I know it is little tricky because you see that you will see that there is a pattern, this forming, right? And if you are able to understand the pattern, right?

9:41 It is hinting like a S. Because since I'm trying to do it, but if you go little clearly, right guys? Cool. So this is what this image is telling you. Okay? So now although you can see here wherever it is zero, it is dark black. Uh, right? In this particular pattern, right? So wherever it is zero, right? It is a dark black. So I'll tell you, if I make this image a tiny image because I have this power to do it, you will see that now you will you will thank me.

10:16 You will get goosebumps. Why I'll get the goosebumps guys? Because now you can see there is a pattern, very clear pattern coming out. Are you feeling it? Nobody will tell you this. Let me tell you very calmly here. So now I'm going to add a pen. It seems that I've already taken all the pens in the world. So I need to make my this guy a bigger guy. Hold on. Right? So it seems that I start, you know, now I'm going to become MF Hussain.

10:00 Don't kill me for this. Okay? But you will realize, right? You will realize that obviously the colors are not going to be all same. You know what? What I mean by that? So whenever the digits are like, you know, zero means black. This is very clear. Zero means black. So you don't need to hesitate. So all these zeros are there. I already know that. So I'll do it. Why I'm doing this much of effort to make you feel something intuitively.

10:23 Obviously, I'll code, I'll show in the hands-on, but feeling is very important, guys. Okay? Which people don't do this, which I pity them. The two different colors here. It seems so. I don't want to have different colors. You will start feel missing the pattern. Right? You'll realize that if you're very careful, to do this pattern thing. So you realize that this pattern somehow will turn to what observe this and there will be slight color coming on the border as well.

10:59 Right? Little colors here. So we try to see the pattern. This is what it resembles to, right? Can I, obviously I'm not the champion of it. But what I'm trying to say, it definitely looks like a number S. Are you guys feeling it or not? Come on. Letter S. So if you try to fill it. If I try to fill it, you'll realize that, yeah, you're right. Akash, you'll say that yourself. So see, if I start filling the pattern, since that is looking like an digits, right?

11:24 You're not getting it. But if I start filling it up, there will be some corners cutting out, right? Because this is how it is, right? So you realize that, yeah, it is coming out like a pattern, hold on. And guys, since this is not the this is not the full image, by the way. This is not the full image. Because remember, you had that missing data on 19 was here, right? So you need, you will be seeing that there will be some zeros coming and trust me, it will be there.

12:02 So some zeros will be coming kahan, where on this side. So it will be black. That's why I'm making black from my side. Guys, okay? Observe. So I know, it doesn't look like that that oh, this is number five. It looks very ugly. But if I try to give you the character here, does it look like five? See, I'm fitting something in between. Right? And I have the image also, don't worry about it. So I've already populated it today.

12:21 I should have it somewhere in the model. Although it is done differently, guys, why? Because I plotted the black, but you observe they are very close. Are you feeling it, guys? I have swapped the color in the printing, so don't kill me for that. But are you feeling it? Right? Just the complementary. But are you feeling a similar kind of behavior I was trying to draw? Since our image stops at 19, you're not able to visualize it fully.

12:57 Okay. Makes sense. Good. You can say that. Guys, did you understand the the nuance here, right? So how do we, how do we, you know, how do we get it? So the whole idea of, you know, identifying any image, it works on two, three things. Basically, there are two words. Always you have to keep in mind. And that is what edges and patterns. So question is, very simple question. How do we identify any image? That's the question, guys.

13:22 Answer is pretty simple. Just take. Don't think about any neural network. Think about any image you see on the page, right? So for example, if I show you this image of cat or a dog, right? You try to identify what? How do you say that it is a cat or a dog? You look at the eyes, look at the nose, look at the lips, right? If it's a cat, you will also look at the body, overall body, right?

13:56 Understand the design and the pattern, right? So basically, it's all about, how do you identify it? Very simple. Depending on the on the edges. That's the first word. Depending on the edges, okay? Patterns, texture, shape, etcetera. There are a lot of things you actually look at. Images are unique in nature. Trust me on that. Okay? So these are some, these are some very straightforward attributes, guys, right? These are your edges, patterns, texture, color. A lot of things are there.

14:27 But in digit, color doesn't matter. Depends on the images, right? So generally, color has not a big play. Big role here. So if I am writing nine with the black or with the blue, end of the day, it is nine only. It doesn't change. Okay? Colors don't play a big role, the kind of algorithm we are going to do, it's all same, that's right. But edges and pattern makes a lot of, a lot of things. So you might be thinking, what do you mean by edges?

15:10 Right. So if you see, observe this image which I have put, put here, right. Or maybe I'll show you my favorite image, which is sequence of trees, right. If you observe tree, if you observe tree, you will realize that it has a pattern, you like it or not. And it has also edges. So what are the edges? If I have to show you guys. It's very simple, if you want to see the edges, see. You know that in tree, this is something you will always find, doesn't matter how they have written.

15:49 Guys, agree or no? So, edges is like outer side, right? Pattern is like this. So if I show you the border is like edges, so this is an edge. Right. But overall, when I combine things together, it becomes a pattern. Do you see the pattern in tree, guys? So basically, I put it nicely, if you look like this, this, this, this. On opposite of sigma, right. So you know that people will try to write tree, and even kids learn to write tree like this.

16:25 Right. So different people write, write trees differently, number three differently, but they follow a simple pattern. Nobody writes three like this, right. Come on, guys. Agree or not? Similarly, for nine, if I talk about nine. Right. You will realize that nine also has a pattern, see. What is the pattern in nine? Basically, you see a loop, without fail. There is a loop. So it can be like this, it can be like this, what rotation not not a problem.

17:00 But it still will look like this and simple this. Right. This is how a nine looks like. Okay. You do whatever you want to do. So there are patterns, right. If someone says eight, eight is having a pattern. So whenever there are two loops, it's an eight. It can be like this, it can be like this, it can be like this. It can be like this. But there is a pattern. Okay. So what just like our brain tries to process this information of edges, pattern, shapes, sizes, textures, color, right.

17:40 They recognize it. Right. That's why you are able to quickly recognize. Look at that, if I ask you, even somebody has written one like this. But you know that what he has written, guys, come on. You know that it's one, there's nothing else looks like, similar to it, right. Guys, what is this digit? Come on, you tell me. One only, na. Although nobody writes one like this, huh. Very, very few people who do it. Everybody writes one like this.

18:12 Okay. Or this. But yeah, this is, this is very clear. Yeah, why not? People write that. Seven will be like, write one. Two. Okay. So slanting line come in any direction. Now guys, let's talk about the detail idea of, you know, neural network structure, right. So I'll again, because I think I had just shown you guys that structure. So let's talk about little bit more in on this. The same structure I'm going to talk about. Okay.

18:47 So there are, if you observe, there are three layers in terms of difference, right. One is input layer. One is hidden layer, but hidden layer can be a lot. In number, see, one, two, three. And then the finally is you get output layer. Right. So what happens, you're trying to do things, right. The input will go through input layer to the next, then next, and then next, and then finally, it itself. So let's also keep talking and write, so I think I did not explain each, each of these layers, right guys.

19:29 That's correct, na. Before I jump to weights and biases. I do not give you any textbook definition, right. Okay. So what is input layer? Input layer is helps you to input the data features. Input the data features. So in this case, in our case of digit, pixel value or pixel values of an image. Now if you are thinking about pixel values, but before we go there, a lot of people don't understand data feature. What do you mean by data features?

20:04 Are input variables. Input variables, predictors, whatever you name. So if we take our image, how many by how many we have? I'm drawing them, I have it here. Okay. Take this guy. How many by how many we have? Come on, tell me. See, if I take the example of our digit, three. 28 by 28. Take this one, na. So take any of it, right. So take this guy. So there are 28 pixels here, and 28 pixel here.

20:30 Do a quick math, guys. Come on. So 28 into 28. How much it? How much? How much it is? I know it even in sleep, 784. 784 are going to get connected to the first layer. Although don't don't go with this diagram, there are just 10 units. But in reality, there will be 784 guys getting getting flattened and connected in terms of neurons. Okay. So this becomes your number of neurons in the input layer. Number of input or features, basically neurons connecting to a hidden layer.

20:00 I'm talking about which use case, which I'll do it later. That is coming from your MNIST dataset. Okay. So to show you this, you will see this later in your course. Just focus on this. So when I'll do this, I'll show you this that we are going to connect how many input to. So if you look at that, how many layers are getting connected to it? It is next connected to 784 neurons are getting connected to.

20:20 How many are those? 784 connected to how many? 128. Clear? In this particular example, which I'll do it in the hands-on. Okay. This is I have assumed that in this input layer, in my hands-on, in this input layer, there are 28 by 28, so if you do the math, it becomes 784. And this is my first hidden layer. I am saying that I'll keep how many neurons? Come on. I'm saying that I'll keep 128. Okay. Right?

21:00 So it's a matrix. Listen to me, guys. It's a matrix of order what? 128 by 128. Right? So basically, if I flatten this, it's a matrix, right? Of 128 by 128, which basically means Sorry, what I'm saying. 28 by 28, which basically means that there are 28 rows and 28 columns. So what is this array, my dear? This array is what array? It is a 2D array. If it is a 2D array and I want to convert into a 1D array, right?

21:00 How will I do it? What is the concept? Array concept? You know it. You can read about it in case you are not aware of it. And the function is, right? Array flatten. Right? Which is np. flatten, if I recall correctly. So that's what I mean by that. Right? That's how you come like a more like a, you know, one line. Okay. Now the question can be like, why 28, right? Why 128 in the next layer?

21:20 Who is telling you this? These are all questions you should ask when you already know about what is the neural network and all, right? Understand now? My feelings? It's like suddenly why sun is moving around, why we are flattening it, you should ask what is the meaning of flattening. But let me tell the structure. Let me tell the structure first. Because if I don't tell you, you will always be confused, right? Okay. Uh, okay. Input layer, I think I am done.

21:40 One thing you need to understand that once you have the, you know, pixelated view of it, right? You will realize that images has a lot of things. So let's talk about pixels. What do you mean by pixel? Because this is something I'll keep using. What do you mean by pixel? Let me ask some baby question. Now, this is a, this is a question to deviate, what do you mean by pixel? And how is it related to image?

21:40 Related to camera and camera size? So camera to, you know, right? It is resolution of your camera, right? How does it relate to your image size? Let's see how good are you. Everybody will open their computer. You know how to see the resolution of your computer? Because I don't know. What is the resolution of your computer? Yes. So my display is saying my resolution is this. 1920 by 1080. What do you mean by that? So if you do a, if you do a matrix formation, 1920 rows by 1080 columns.

22:00 Okay. So that's how it looks like. So every image of mine will have how many cells, guys? You know that, right? 1920, 1080. So total how many? Look at my image size. 1920 into 1080. If you if you're good with this, you know that it will be what? If I do a simple math, it is how many pixels? 2 megapixel. Yes, come on. Yes or no. You know that. Approximately my dear. Approximately. Why? Because see, it is 10 to the power 6.

22:00 Right? Agree? So don't think your resolution is bigger than 2 megapixel. There is a reason for it. Let's not go there. Okay. So just imagine those who are using. So let's say an iPhone has a 12 megapixel. So 12 means what? 12 into 10 to the power 6. Right? Megapixel means this is only. So you can think like that, you know, more like 4 by 3. So 4 million here, 4000 by 3000, right? Then you will get, approximately I'm saying, making a grid out of it, guys.

25:00 Come on. Which is a lot. How many these are? 4000 by 3000, right? Just imagine that the number of neurons we will see here, we are just looking at 784 in this example. Here you are coming in millions of neurons. So what is a pixel? Pixel is, by the way, what is the full form of pixel? Pixel means picture element. So let me give a pointer here. It means picture element, you know, pix L. So if you bifurcate pixel, pix L.

25:30 Pix means pix, right? Image. So L is what? Your element. And pix is your picture. So it's a picture element, right? Which is the smallest unit. So those who wrote smallest unit there, 100% right. Which is the smallest unit of a digital image or display. Doesn't need to be an image, it can be display also. Okay. Now question comes that how it is related to the image size, right? So just imagine, image size is definitely they are measured in pixels only, right?

26:00 If you observe. So you might have heard about pixels per inches, you know, or you know, when you're doing the facts. So there you see the, right? DPM, or printing. DPM, you know what? DPI, sorry. DPI, guys, you know, in facts or in scanning, dots per inch. Yes. So generally, if you take a 4 by 3 image or 8 by 7, right? Dots per inch. So there are different types of printer. That's how you define the you know, quality of your image there, right?

26:40 So I'll not go there. But you know that in terms of this is about your, you know, printing. But if you do a similar kind of analysis here, it is called PPI. What is PPI? Pixel density or pixel per inch. Right? So if you talk about pixels per inch. So what is the interpretation of it? If you take an image, right? If you take an image, more like a as a unit dimension, if it is an inch of, right?

27:00 So generally an inch is a very small unit. How many pixels you are putting there, right? So higher the, higher the PPI. You tell me. Higher the PPI implies more details and sharpness of the image. Basically, you talk in terms of good image, qualitative image. More details and sharpness, sharpness in the image. Okay, that's what you mean by that. Basically, you say you have a better resolution of the image, right? And if your image is better, generally your image size will be also higher.

27:20 Agree or not? Come on. So if I talk about a 12 megapixel camera. I'll not go to our school days when I used to use 1.3 megapixel camera. So 12 megapixel camera, right? If I just say, listen to me, the same thing I'll call it here, guys. It is 4000 pixels here and 3000 pixels here, right? So that's the reason it is called 12 megapixel. You must have been, so don't be now uneducated. Right? So now since you are sitting in my class, you should be, should be aware of these things.

28:00 So 12 megapixel camera releases every time, right? 12 million pixels for each images. Now, if I give you a simple example, that if your images like, you know, a simple image of let's say 10 6 inch wide and let's say it's a portrait one and 10 inch height. Okay. So generally, if you see these kind of standard photo size, right? How it is being, being framed. Understand this. If I go with a standard 300 PPI, which basically means 300 pixels per inch.

28:20 Remember that, guys. Come on. Listen to me and this is very important. Doing chat GPT will not help. Right? For a 12 megapixel, what is the image size in inches it will come? Let's say this is A and this is B. What is the length of A and what is the B? What is the image size, we'll get it. If I give go with this assumption. Let us assume the pixels per inch is 300. If I say 4000 pixels, sorry, 4000 pixels and one inch, you're putting how many?

28:40 300 pixels, right? So basically, it's very easy, it's not a rocket science. So you can do this math, 4000 by 300. How much it is? 13.33. So this is your 13.33 and this will be what? 10, so it's not 14, it's 13.33 and this is going to be 10 inch. Right, so if you, if you get this PPI and divide it, you should get an image and that is an image which will not be, you know, it will not get stretched or anything right, it's a, it's a natural fit for it.

5:32 It has different, different edges, patterns, and it identifies that through series of computation and only fire the ones which are required for it, right? Looking at the input image. Okay. So let me show you this. This is how it looks like. Now, you might be thinking. And if you're observing, guys, right? Whenever you are trying to give, let's say I give a different type of thing. This is my image, right? It looks like a part of four.

6:20 It looks like a part of nine. Can you observe, guys? That's why the model is confused, right? I'm showing the confused behavior. So when you give an image like this, which is having a pattern, something like this, right? So you know that what it will do. It will try to, try to find the. And if you observe there are 784 input. Let me also write. So it is 28 by 28 pixels, which actually becomes 784 pixels in the input, right?

7:14 Focused. Okay. Then it gets into the next one. What is this one? If you count on the finger, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16. I'm not keeping 128. It is 16 neurons. Don't get carried away. That 128 I took it from the hands on. Then another 16 neurons in the next hidden layer. If you see, right? And then finally, how many? 10 neurons, right? We need 10 neurons for output, right?

8:07 0 to 9. 0 to 9 is 10. Are we clear, guys? So if you observe, you are doing a math of how many combinations, 784 into. Both the links. But if you see, what I'm trying to say here from the image by using a filter, this is your filter. Okay? But this is very high level introduction. I don't want to get into the detail yet, but you got it, right? But how do you explain convolution to a kid?

6:01 That's question remains there, right? You can't say that. Oh, you are taking a, you know, filter and all. How do you explain a kid? Okay. How do you explain convolution? Okay. So let me explain. Question is, how to explain convolution to a kid? Let me help you. How to explain convolution to a kid? So let us break this. Suppose, so understand, they have a small image. Right? So let us imagine. Checkerboard you have seen, right? Let us imagine.

7:13 I'll write and talk through. Let us imagine we have a small image. We have a small image. So checkerboard is something if you have not seen it, like you are adult and you have not seen a checkerboard. It looks like this. Right? Now what is about? You will do something about it. But if someone tells you that, dude, your heart has failed. What will you do about it? Nothing, you can do about it. You will just spend your all the savings in those 10 weeks or 20 weeks, but the end result doesn't change.

5:50 90%. But in three years, your five years, if somebody predicts something can happen to your body, looking at your lifestyle, Have proven very effective in image recognition and classification. Image recognition and classification. Okay. So what, what is the implication? When I say this? So, you know, it's a really successful in identifying faces. Sketch drawn, uh, drawing, sketch drawing, putting into a digital mode and identifying the criminal through the database. Guys, If you just scan an image, no fingerprint, how will you identify a criminal?

6:40 How will you find the social security number? Have they seen these movies? Don't think it is all fake. It's all real. So they have a database of all the top notch criminal, right? In CID, they are using. Even in India, it's not like they have to do it outside India, they have this mechanism. So identifying faces can be a real use case in these kind of world, right? Like somebody has a, you know, red corner notice, right?

7:40 So let's say I'm the Vijay Mallya, or likes of them, right? And I am done some fraud, and I can't travel outside. Right? You have some restriction. So one way is that to stop it right, or find that person is that, you know, capturing images from traffic signals, right? Because they will be traveling and you know, process that in the database. Makes sense? So objects, traffic signs, right? You know, apart from powering vision in robots and self-driving cars.

9:00 So just understand the length of the use case. The output feature map will be smaller. Sometimes my English. Output feature map is smaller than the. It is 0. 0009%. But not much, right? Guys, nothing is happening, right? Guys, come on. Agree. So do you see any losses happening? Nothing, yeah. But understand this. For rest net architecture, is the latest one in the market. Right. For rest net architecture and that's how I'm talking about 128. For rest net architecture which is 128, when I'll show you this architecture just for the sake of knowledge.

6:33 You know, it has how many convolutional layer? You know that? 50 convolutional layer. Can you believe it? And each layer of 3x3 size filter. Can you believe it? 50. Now take the same image. Right? Here you think it's not making a lot of sense. Right? No, no. Not it's not about the image. You are missing the points and re. In real world, you'll not have only one layer. You will have multiple layers of convolutional. And each having a filter size of 3x3.

7:20 You understand what I mean? Where I'm going? So filter size is 3x3. Now for every 3x3, you are going to miss on what? Right? 2, 2, 2, 2. Agree? Then I was talking about the pooling thing. So that also something I covered already. All right. So what I'm trying to say, only thinking from, like a big company, and trying to justify this. But think about normal users like us, right? You think your machine is smart enough?

6:00 Your machine is smart enough to do things like, you know, running a big object detection project. No, I don't think it's reality. So it's not that right. So that's why when we talk about these concepts, we are talking about a general machine configuration which the mass uses, right? Not the other way around. But you're right, means you can have some machines. Who are which are powerful. But not trust me, I use a lot of the biggest machine.

6:20 So currently, the machine I'm using in my project is having, you know, how much RAM. 664 GB RAM. I know how many workers are there, like CPUs, threads, 1000, 24 CPU threads. And still sometime it doesn't work when you're giving me. 664 GB RAM, I'm using on a server. So, and I'm the only user. It's not like it is spread around multiple things. So, yeah. Means I know what do you mean. Guys, this is just to show you the recapping the padding once more.

7:20 In a different way. So you can see right, this is your original image. Of size one, one, right? So if you see this here, pixel utilization for convolution of size one by one, two by two, and three by three respectively. So if you guys observe It will confuse you to be very honest. So there are two types of filter. I would always say that this filter. This filter is associated to vertical edge. The other filter is associated with the horizontal.

5:49 Okay. So Sobel wise, horizontal. I'm just also trying to say. I'm just also just confirming. Guys, give me a second. The kernel is vertical is negative on the top. I'm just just trying to recall that. But looking at the image what do you think? It is like a horizontal or vertical? I think image itself says here. So I don't worry about anything else. Just look at that. This is giving you horizontal edges it is picking. Because I can clearly see that these are your edges on the vertical side and it is picking it.

6:34 So this is definitely vertical edges, no? I'll explain everything here. That's why I'm not going into too much of mathematic because trust me, mathematics is super heavy here. But what I'm trying to say, and you have to pay attention. To understand. So I just write one pointer. RNNs allow previous outputs to be used as inputs while having hidden states. Now this line has a lot of depth. How we will see. I'll try to show it's not easy to show within the box.

6:39 Here, understand here. So here XT represents sequence of inputs. Sequence of inputs, okay? Now, we will also break it down that what we mean by this. Sequence of input, okay? But what is, what do you mean by sequence of input? It is contextual, contextual. Depends on the context, right? Green boxes at the bottom, denote the input sequence. This is how it's working. At least in Bangalore, this is how it's working. So Rapido is like a betting system, right?

5:40 So you have to bidding system, sorry, wrong word. Bidding system. So if you are bidding more, they'll give you, they'll the you will get a car faster than anybody else. So suppose, a normal route cost you from origin to destination 400 rupees. And you say, I will give 500. Dude, complete it. You know? So if you are busy, you are, you don't care for 100 extra bucks, you're not worried about it, pay and go. I've done it multiple times.

6:00 What to do? Right? Obviously, not something I would like to do, but what to do? So guys, just understand that how they are tapping the customer sentiment. So those who want to make a career in this field, they can study sales and marketing. And there's a very nice book on the marketing research. Right? I'm forgetting the author's name, but in case you need, you know, and you're interested in that, please. This is like a very fulfilling career.

7:00 But join a very good company because sales and marketing can be very toxic job to be in. Trust me on that. Especially the sales, not the marketing. Marketing is like okay. You can do any ad, you can do anything, but targets and all you have to meet, so it's very tricky. Competitive. Okay. But I by by that, everything is competitive in this world. So that is something and watch out for you. And then you can use a lot of.

7:40 So Salesforce is one of the, one of the, one of the biggest tool in this world, right? So I feel, if you're aware or not. Salesforce is one of the biggest platform for this. And I feel that this is the way you can, you can learn these techniques. So, coming back to our topic of interest, which is your sentiment analysis. I'll talk about in this detail. What is sentiment analysis, but I've given one line up. Then you make a box there and saying that if you find this similar, right?

5:40 You do some action on top of it, right? Agreed, correct, no? So if you think OK is there, you have to click OK, right? Now, you know what? I don't know. I want to show you a demo. So shall I do it? Can I take five minutes? Because I know almost we are done, but I'll just take one five minutes. Are you guys ready for it? Can you see chat GPT? So I'm using a paid version, my dear.

6:20 Okay. So this is something I try to use it the most of the time. Can you see chat GPT? Okay. See. I'm just sharing a screenshot. Where is that screenshot? So stop what analysis. Let's say step five. Obviously, this is handwriting. Right? Suppose I just copy paste it. Hey, come on. I paid for it, right? What's wrong? This is not done. Why the attachment is not coming? My paid account is gone or something. Seems so. Coming from a project on Kaggle.

5:31 Okay. Description, I-level description, guys. Passengers tweet. Tweets about airline performance, having sentiments about their journey. See, I could do this on Collab, but the problem is that Collab is really slow and time is less. You have seen, right? Priyanka. So it takes time to react to what I'm writing. That's why I'm not much happy with that. That's all point. Anyway, you're not going to code with me. Just have to see on the screen. So let us do that.

5:36 Additionally, it's a great way to get hands-on experience with Python and machine learning frameworks, both of which are valuable in the tech industry. Moving on to the fourth project idea. Job seekers today face a major challenge: getting their resumes noted by recruiters. Many companies use an applicant tracking system or ATS to filter resumes before they even reach a human recruiter. These systems scan resumes for specific keywords, job titles, skills, and qualifications. If a resume doesn't match these criteria, it may never get seen by a hiring manager.

6:26 Now, the goal of this project is to build a system that can automatically pass and extract important information from resumes. This makes it easier for ATS to process resumes and for recruiters to find the most qualified candidates. This resume parser should be able to extract details such as the candidate's name, contact information, skills, education, work experience, and certifications from various resume formats like PDF, DocX, and plain text. It will also analyze and categorize this information, helping job seekers optimize their resumes for ATS.

7:08 In order to get started with this project, we'll use natural language processing or NLP to extract and understand the content of resumes. For