WEBVTT 00:01.590 --> 00:02.410 Hello, everyone. 00:03.240 --> 00:11.310 This is a very special session on data science, so let us try to discuss about what is data science. 00:11.880 --> 00:19.590 Let us try to understand what are the major films which are used interchangeably and what is the real 00:19.590 --> 00:20.660 meaning behind them? 00:21.900 --> 00:24.870 So the first thing which we will discuss is. 00:26.580 --> 00:33.570 The difference between artificial intelligence, machine learning, deep learning and neural networks. 00:35.890 --> 00:44.530 So when we talk about artificial intelligence, artificial intelligence is a simulation of human intelligence. 00:45.680 --> 00:54.110 So you remember the movies which we used to watch, the movies where we had several Roubaud which were 00:54.110 --> 01:02.180 trying to replace humans, which were capable of thinking, which were capable of doing different things 01:02.420 --> 01:02.930 like. 01:04.300 --> 01:12.400 Opening a house, like talking to people just like humans, talking on a phone call or doing different 01:12.400 --> 01:20.500 tasks like preparing food or maybe doing our database could do so. 01:20.500 --> 01:30.850 Artificial intelligence focuses on bringing these changes into a small robot or any technology, anything 01:31.000 --> 01:43.180 which is capable of thinking like human or expressing itself in the five senses that human has is called 01:43.180 --> 01:44.650 artificial intelligence. 01:45.660 --> 01:48.340 And they are a little far from it. 01:48.570 --> 01:56.910 We have not reached that stage when a robot can actually replace a human being, but we want to make 01:56.910 --> 02:06.090 a world where a robot can take a lot of walk from human and give a little measure down to us human beings. 02:07.250 --> 02:14.990 So thinking like robots can actually replace humans is very far from right now, and this is a completely 02:14.990 --> 02:16.570 different point of discussion. 02:16.850 --> 02:20.600 So without going further into that, we need to understand. 02:21.550 --> 02:31.450 That artificial intelligence refers to the simulation of human intelligence in machines, and these 02:31.450 --> 02:42.700 machines are programmed to think like humans and mimic the human actions to personify the human behavior. 02:43.300 --> 02:50.960 We want these machines to do walks and different tasks, just like a human being would do. 02:52.060 --> 03:01.480 So that may also be applied to any machine that exhibits the traits associated with the human mind, 03:01.660 --> 03:04.840 such as cloning or Problem-Solving. 03:06.310 --> 03:17.520 So if I have a particular robot or anything which is helping me to maybe cook, then I want that machines 03:17.530 --> 03:23.820 to tank in a way so that it can help me out in creating new recipes. 03:24.810 --> 03:34.830 So I can use it for that or different scientists are doing different searches on medicines, so for 03:34.830 --> 03:44.070 a typical kind of behavior legacy, for a typical kind of disease, a particular chemical works really 03:44.070 --> 03:47.900 fine, or that is one component which works very fine. 03:49.410 --> 03:58.740 So a machine can actually learn what all of the chemicals work in a way to improve from that disease 03:58.740 --> 04:06.570 or prevent from that disease and create different chemical constituents or different components, which 04:06.570 --> 04:10.190 could actually help in curing the disease. 04:10.590 --> 04:18.510 And then it can actually assist a human being by doing all the research work, which a human would have 04:18.510 --> 04:27.120 done behind it and helping a human with those components, which he can basically try and experiment 04:27.270 --> 04:28.530 and test for the. 04:30.160 --> 04:38.080 So here is artificial intelligence is actually reducing the time of the research, which a human being 04:38.080 --> 04:45.910 would be doing and assisting a human being with reducing a lot of time and a lot of effort. 04:46.270 --> 04:48.880 This is what artificial intelligence is focused on. 04:53.610 --> 04:57.880 Now, what is machine learning when we talked about artificial intelligence? 04:57.910 --> 05:01.180 We wanted a machine to mimic the human behavior. 05:02.220 --> 05:10.530 A machine was supposed to think like a human or walk like a human, like it could walk like a human 05:10.530 --> 05:12.480 or maybe could talk like a human. 05:13.050 --> 05:20.310 But to do these things or simply saying, let's say we want a machine to let us, they just simply go 05:20.310 --> 05:21.250 off like a human. 05:22.080 --> 05:27.930 Then in that also we need the machine to learn a lot from different languages. 05:28.560 --> 05:31.920 Let us say we want the machine to learn the language English. 05:32.580 --> 05:39.120 Then it would have to understand the difference between different words and different spellings. 05:39.330 --> 05:44.700 It should need to understand what a about a particular word would mean. 05:45.000 --> 05:50.420 And in English language, there are different words which could be used with different meanings. 05:51.510 --> 05:54.530 It could have passed from present form. 05:54.570 --> 06:02.640 So all these things, the machine has to understand that now we humans have learned these languages 06:02.790 --> 06:03.660 over the time. 06:04.850 --> 06:12.980 While the machine would have to understand this in the simplistic manner, in the form of zeros and 06:12.980 --> 06:19.250 ones, because machines cannot really understand what the machines can only understand zeros and ones. 06:20.060 --> 06:28.640 So here the machine would also have to actually be corrupt this entire language and convert it into 06:28.640 --> 06:30.260 a form of zero and one by. 06:30.260 --> 06:31.710 I'm trying to understand that. 06:32.630 --> 06:40.100 So this learning experience and again, it will have to learn this all this information not in 20 years, 06:40.100 --> 06:42.190 but maybe in one hour or two. 06:42.830 --> 06:46.150 This is the kind of training time the machine would be taking. 06:46.880 --> 06:53.900 We want the machine to learn over a smaller period of time and from a very vast space of knowledge. 06:54.350 --> 06:56.510 So what machine learning helps us? 06:56.990 --> 07:05.330 It is an application of artificial intelligence that provides the systems, the ability to automatically 07:05.330 --> 07:10.840 known and improve from experience without being explicitly program. 07:11.030 --> 07:16.670 So I don't want to bring my machine to do a specific task. 07:16.790 --> 07:22.580 So let's say we want to go walk with let us say we want to work with a set of in-car. 07:23.820 --> 07:32.030 So in case of a sense of anger, I don't want to program to walk with, let us say I want I don't want 07:32.040 --> 07:40.740 the program to take a left or right or maybe see a stone and then it should stop or see someone crossing 07:40.740 --> 07:42.350 the road and then it should stop. 07:42.540 --> 07:50.460 I don't want to to each and every line for that one I wanted to learn is I want to give it a few examples. 07:50.880 --> 07:57.570 I will give it a thousand or ten thousand examples where a person would be driving a car. 07:57.570 --> 08:03.480 I'm the machine should learn from it that it should learn that OK, there was a person who was crossing 08:03.480 --> 08:03.960 the road. 08:04.110 --> 08:09.900 This is something which I found from the image at that point of time. 08:10.140 --> 08:14.880 And often this image appeared of a person crossing from the road. 08:15.090 --> 08:20.940 The action which was taken by the guy was to slow down or to put the brakes. 08:22.020 --> 08:29.340 Then when something huge was placed in front of the car, the action which was performed by the driver, 08:29.340 --> 08:32.070 was to slow down again or to. 08:33.000 --> 08:39.290 The thick or divert its direction slightly towards the left or slightly towards the right. 08:40.570 --> 08:46.780 So these are different examples which would be given to the court and it would automatically learn from 08:46.780 --> 08:50.080 this, so we don't want to program anything. 08:50.080 --> 08:51.720 We don't want to program the machine. 08:51.940 --> 08:59.350 We just want to give it a lot of examples and a lot of samples from which the machine can learn different 08:59.350 --> 09:01.180 and find out different patterns. 09:01.420 --> 09:05.860 And from these patterns, it can predict what it should be doing in future. 09:06.730 --> 09:14.050 So machine learning focuses on the development of computer programs that can access the data and use 09:14.050 --> 09:16.000 it to learn for themselves. 09:18.300 --> 09:20.970 The next thing which we have is neural networks. 09:22.310 --> 09:24.920 So we have talk about artificial intelligence. 09:25.160 --> 09:30.050 We have talked about machine known, now machine learning is a. 09:31.240 --> 09:38.350 S field of artificial intelligence machine learning is a part of artificial intelligence when we talk 09:38.350 --> 09:42.470 about a machine to go for a machine to work. 09:42.730 --> 09:50.020 We want to learn different ways of working or we wanted to learn different languages using the process 09:50.020 --> 09:50.950 of machine learning. 09:51.220 --> 09:56.530 So machine learning is a part of the entire process which is undergoing artificial intelligence. 09:56.890 --> 10:03.100 While Neural Network is one of the algorithms which helps implementing machine learning. 10:03.280 --> 10:10.350 Machine learning has a lot and lots of algorithms in place and neural network is just one of them. 10:11.110 --> 10:16.350 And the neural network is based on the how our brain works. 10:16.630 --> 10:23.860 So in our brain, again, we have a lot of neurons which are interconnected and each of these neurons 10:24.070 --> 10:33.130 keep sending signals from one to another so that the information is passed and converted and encrypted 10:33.130 --> 10:38.410 internally in our brains, and then we could get a result out of it. 10:38.560 --> 10:40.840 We can make an inference out of it. 10:41.200 --> 10:45.220 So a neural network is a network which is trying to. 10:46.780 --> 10:47.860 Somehow 10:50.110 --> 10:57.430 depict the brain or does show how a brain works and it is a network or circuit of different neurons 10:57.670 --> 11:04.040 or in a modern sense, artificial neurons composed of artificial neurons or nodes. 11:04.300 --> 11:10.720 So these neurons and mode which are connected with each other, they try to pass information from the 11:10.720 --> 11:16.840 input to the output so that it can internally create the network kind of thing. 11:16.840 --> 11:24.490 Or let's us it just try to form a circuit of information which could give us the relevant. 11:25.400 --> 11:29.800 How how does that what is going on behind Windows that later. 11:29.980 --> 11:35.350 But for now, what you need to understand is it is just trying to become a human being, how the human 11:35.350 --> 11:36.100 brain moves. 11:40.170 --> 11:50.520 Next is deep learning, so deep learning is an artificial intelligence function that mimics the walking 11:50.520 --> 11:59.430 of human brain in processing the data for use in detecting objects, recognizing speech, translating 11:59.430 --> 12:01.730 languages and making different decisions. 12:01.980 --> 12:10.110 So deep learning is in hired version of neural network, which would be said neural network is a very 12:10.110 --> 12:13.620 basic symbol of neural network. 12:13.630 --> 12:18.560 And when we have a complex neural network with a lot and a lot of layers. 12:18.960 --> 12:23.350 So if we talk about just this much, this could be said as a neural network. 12:23.580 --> 12:31.050 But when a lot of layers are included in case of the neural network, it becomes a deep network which 12:31.050 --> 12:32.910 is used for deep learning. 12:34.210 --> 12:43.570 So artificial intelligence without human supervision and throwing different data and structures out 12:43.570 --> 12:45.390 of it is called deep learning. 12:45.550 --> 12:50.350 So deep learning is basically a neural network with a lot and a lot of layers. 12:55.100 --> 13:01.550 So now that we have understood about artificial intelligence, machine learning, deep learning, neural 13:01.550 --> 13:04.970 networks, we need to understand what data science actually is. 13:07.880 --> 13:17.290 So data science is the process of extracting knowledge and insight from large volume of data. 13:18.680 --> 13:21.530 It is simply learning from the. 13:23.450 --> 13:29.810 So let's try to understand the simple language, so when we're are trying to study for a particular 13:29.810 --> 13:32.600 exam, what do we want? 13:33.530 --> 13:38.600 We want a lot of material which has a lot of information in that. 13:40.740 --> 13:49.230 I will figure out around two or three books, which will have all the data and all the information related 13:49.230 --> 13:55.230 to the course, which I'm going to have to answer in the exam, which is let see tomorrow. 13:56.800 --> 14:02.410 Now, once I have these books, I will read each and every page from the book. 14:03.600 --> 14:09.960 And when I will read from the book, I will try to find out different pieces of information which are 14:09.960 --> 14:12.310 important, which are the patterns. 14:13.020 --> 14:14.070 So let us see. 14:14.070 --> 14:20.170 I have a science exam, so I will try to pick out some important concepts from it. 14:20.370 --> 14:22.530 I will learn about photosynthesis. 14:22.530 --> 14:25.130 I will learn about how water evaporates. 14:25.150 --> 14:31.320 I will learn about different things which are going on inside this particular lecture or in this particular 14:31.320 --> 14:31.910 chapter. 14:32.160 --> 14:38.730 And then when tomorrow someone would be asking me the question regarding this particular topic, I will 14:38.730 --> 14:40.140 be able to answer that. 14:41.630 --> 14:47.150 So to answer a particular question, I need to have visited that concept. 14:48.320 --> 14:56.990 So this is one data sciences I want to extract and I want to know one from a piece of information, 14:57.620 --> 15:06.290 and I want to learn different patterns from the information so that if any question is asking me around 15:06.290 --> 15:09.990 this particular topic, I will be able to answer that. 15:11.220 --> 15:19.370 Remember, I am not being given the exact questions which will present in the book if I I'm able to 15:19.940 --> 15:22.670 know from the exact questions and. 15:24.130 --> 15:31.330 Give answers to the only those exact questions, then that is what deboning if I was able to do that, 15:31.330 --> 15:37.960 if I try to do that, then I can simply do that by simply switching the pieces of information in a database, 15:37.960 --> 15:40.680 maybe, and then giving answers from that. 15:40.870 --> 15:42.410 But that's not what I want to do. 15:42.430 --> 15:49.720 I don't want to go to a question paper or a list of question papers or a bunch of question papers and 15:50.140 --> 15:55.990 20 questions and being able to answer only the questions of the come out of those 20 questions. 15:56.000 --> 15:57.250 No, I don't want to do that. 15:58.310 --> 16:05.540 Here, what I want to do that I want to do is I want to learn the concept and this is what data science 16:05.540 --> 16:14.540 is learning and extracting the knowledge and patterns and the insights from these structured and unstructured 16:14.540 --> 16:14.810 data. 16:19.600 --> 16:29.050 So data science is an interdisciplinary field that uses scientific methods, processes, algorithms 16:29.200 --> 16:37.380 and systems to extract knowledge and insight from many structural and unstructured data. 16:38.820 --> 16:47.360 So let us see, I have a lot of e-mails and I want to find out if a particular e-mail is a spam or not, 16:47.640 --> 16:56.610 so I will try to find out some specific patterns which I find in spam emails such as a lottery or bank 16:56.610 --> 17:00.210 account or free or Huddy. 17:00.540 --> 17:07.800 These kind of words will mean then someone is trying to hurry up and try to grab this opportunity or 17:07.980 --> 17:10.170 get something free goodies from someone. 17:10.350 --> 17:11.670 So I don't want to get those. 17:11.850 --> 17:16.590 These are completely spam, so I don't want to get into this family, so. 17:18.010 --> 17:25.390 I'm finding out these pythons died when it was another meal comes out saying you will get these things 17:25.390 --> 17:31.660 free or you have won something, you won the lottery, then I can completely ignore these and put them 17:31.660 --> 17:32.850 in my spam folder. 17:34.150 --> 17:41.530 So this is the pattern which I have extracted from the emails which I have caught, so data science 17:41.680 --> 17:45.300 is related to data mining, machine learning and big data. 17:45.310 --> 17:50.800 So I can have any amount of detail and I can use that to work with these. 17:51.250 --> 17:52.930 So this is one big differences. 17:56.630 --> 17:59.340 Now, why do we really need the signs? 18:00.320 --> 18:03.830 We need data science for making decisions. 18:05.270 --> 18:12.110 In an organization, there are a lot of organizations which are hiring different data scientists and 18:12.350 --> 18:17.490 there are a lot of knowledge on a lot of information which is present. 18:17.510 --> 18:24.350 Nowadays, we have so much of information in different files, in different databases, but no one really 18:24.350 --> 18:26.350 knew how to use them. 18:27.230 --> 18:33.650 But now we have data science and machine learning, which will help to find out different patterns from 18:33.650 --> 18:34.270 this data. 18:36.310 --> 18:42.970 And if I want to make a particular decision, then I can go back to this data, I can learn from this 18:42.970 --> 18:51.100 entire data and then make a particular decision, because nowadays the decisions are not made based 18:51.100 --> 18:52.050 on indications. 18:53.120 --> 19:02.440 Someone cannot really see that, OK, if we want to increase the scenes, then let us add a bungalow 19:02.480 --> 19:03.040 packaging. 19:04.260 --> 19:10.830 We need to have a lot of data to prove that actually having a pink packaging will improve the sales. 19:12.270 --> 19:19.290 So the decision making has to be fact based and it has to be a realtime decision, we need to make real 19:19.290 --> 19:20.250 time decisions. 19:20.250 --> 19:20.860 Legacy. 19:21.120 --> 19:26.790 Is a flight which is about to take off in five or two minutes. 19:26.790 --> 19:31.630 And I want to find out if the machinery is working fine, if the plane is flying. 19:31.860 --> 19:41.430 So for that, I need to analyze all the mechanics of the plane so that things can not be done in one 19:41.430 --> 19:43.810 or two or three hours time span. 19:43.980 --> 19:51.270 I need to have a quick decision making here, so I need to have that machinery in place, the system 19:51.270 --> 19:56.970 in place, which can check a plane in real time and find out, OK, this is good to go. 19:56.970 --> 19:58.470 We can go for it, people. 20:00.180 --> 20:07.230 And also, there is a lot of intense competition which makes having these real time decisions and precise 20:07.230 --> 20:08.580 decisions very. 20:10.570 --> 20:14.540 So this is life data science is very, very, very important.