hello this is Sheila Dvi i am the instructor for your course on artificial intelligence so this is the first lecture in this course and um I will just do a small introduction before I start the course so this will talk about uh some introductory concepts of this course so basically uh the textbooks which need to be followed are as follows the first textbook is artificial intelligence book by Russell and Norwick which uh is the one which has been mostly followed in this course another book is the artificial intelligence book by Luger which is used for some small specific topics here and there and there are some other textbooks also which have been followed and I will let you know from time to time what are the textbooks that I have used over there so what are the broad topics which we will be covering in this course so basically we'll start off with talking about what are agents and uh uh what are intelligent agents and the different types of intelligence agents that we have so we have something called a simple reflex agent and we also have more uh complex agents okay which are more powerful so we'll talk about different types of agents and an important topic in artificial intelligence is the general problem solving in other words if you have a problem how would you have a general procedure for solving the problem so usually you use like a search algorithm so you search for a solution to the problem so this is something like the way human beings do the uh problem solving and the next topic will be on logic so what are the different types of logic we have uh mainly you have things like propositional logic u etc and when we represent things in logic then how do we carry out reasoning using that logic because obviously when we are talking artificial intelligence when you have certain knowledge you're trying to find new knowledge by using the means of reasoning and also you have things like uh theorem proving which will tell you whether you know something is right or wrong etc and then we also have other types of uh logic like fuzzy logic non-monotonic logic etc so we will talk about all those types of logic another uh topic is planning so how do you find a plan to carry out a certain goal for example if you have some space mission so what are the steps which need to be taken in the space mission so you plan it out so that is what we mean by planning so there are different types of planning algorithms so we'll go through those different planning algorithms and as we know learning is a very important concept over here so we have uh things like um supervised learning uh unsupervised learning reinforcement learning and in each of these for example supervised learning you have things like um different machine learning algorithms deep learning algorithms so we'll be going through all those type of things and uh when we come to um languages of course along with the usual languages which are there we also have some AI languages which are used so these AI languages are are things like uh list prologue etc where u you know some concepts for example list processing and processing of text etc is carried out and and prologue which you could call as something like a shell expert system shell so we'll be going through some of these languages and communication of course the agent needs to communicate with the human being or communicate with each other so how this communication is carried out so that is what we mean by natural language processing and we'll talk about other things like multiple agents web agents negotiating agents and all these uh different types of agents okay so Okay so basically what are we doing in artificial intelligence we are trying to model how the human being carries out a task in other words you're trying to model human cognition or the mental faculty which the human being carries on and then comes to a decision and and carries out that decision so you're trying to make the computer do things which the human being does now and it is considered to be something intelligent okay so you want the computer to to do things which require intelligence basically so that's the basic idea over here okay so let us go into some definition of artificial intelligence so we can say that it's a branch of computer science okay so AI is considered to be a branch of computer science where you're trying to uh study and create computer systems that exhibit some intelligence okay or which we associate with intelligence in human behavior so when a human be behavior does um even things like recognizing a person or even a child recognizing a person you can say that it is using its intelligence right so that intelligence we are trying to model by using a computer basically okay so in other words you're trying to find a machine which thinks okay so we want to classify a machine as thinking or in other words if it needs intelligence to do something then we say it's a thinking machine right so we are trying to uh create that an artificial intelligence so here are some different definitions given in different places so behavior of a machine if when performed by a human being would be considered intelligent another is study of how to make computers do things which at the moment people are better so what people do better now okay how do you make a computer do the same thing the another one is theory of how the human mind works okay so you try to see how the human mind works and you try to simulate that to get the same result so these are different definitions of artificial intelligence basically here are some more definitions from different places so the famous Alan Turing so you have the Turing test which we will talk about little later so actions that are indistinguishable from a humans okay so if the machine carries out a task which looks like it's been carried out by a human being so AI is the study of complex information processing problems that often have their roots in some aspect of biological information processing the goal is to identify solvable and interesting information processing problems and solve them so this is David Maher from MIT uh the intelligent connection of perception to action in other words the computer perceives something and then after perceiving something it needs to take an action so that action which it takes how it comes to decide what action to take that is where there is a intelligence which is involved so when we are talking about uh artificial intelligence these things are some of the things which need to be done learning new concepts and tasks okay reason and draw useful conclusions about the word world remember complicated interrelated facts and draw conclusions from them so that is something like the expert systems which we talk about understand a natural language or perceive and comprehend a visual scene look through the camera and see what is there okay and move around so that is something like what happens in robotics basically okay plan sequence of action so that is what we call planning how do you plan for example you want to go u and buy some milk and oranges and bananas say right so what is the series of actions that you're going to carry on so you'll have to have something like leave the house go to the supermarket buy some milk go to the hardware store buy something etc right and then come back home so what you have is a series of actions to carry out a task okay next is offer advice based on rules and situations so that is like the expert system so you may have an expert system which will um if you give the symptoms of a person it tells you what disease they have okay so that's what we mean by offering advice and of course it may not necessarily imitate human uh thought processes in in some cases it could be done in a different way but get the same result basically okay uh perform tasks that require high level of intelligence etc so that's what you're trying to see how to make the computer do basically so here is uh an overall block diagram of artificial intelligence so if you see in the middle it tells you all the subjects that are covered under artificial intelligence so we spoke about reasoning learning planning right um knowledge uh acquisition and all those other things right and of course there's a lot of int u uncertainty involved in that and then we have some parent disciplines so the parent disciplines are the disciplines from where artificial intelligence has actually come up okay for example philosophy psychology maths physics computer science etc and if you look down you can see some application areas of AI for example game playing so game playing is a very important um application areas so things like chess checkers you know all these types of games theorem proving okay language understanding uh robotics etc so these are some application areas which we are talking about okay so here we can see the different um topics which will be covered under AI basically okay so what are these topics you have things like uh knowledge representation so how do you represent knowledge okay so that's a very important concept so when we talk about um say ordinary data structures algorithms etc we talk about data okay so you have data to solve a problem whereas here you talk about knowledge for example you can have uh knowledge in your uh database like all crows are black okay so that would be like a knowledge so um you need to do representation of knowledge okay not only uh facts okay so you need to have facts and rules etc theorem proving game playing common sense reasoning okay uh learning models inference techniques pattern recognition okay so some classification algorithms searching and matching etc and logic so you have things like fuzzy logic temporal logic model logic all these things planning and other sub areas are like natural language understanding computer vision understanding spoken utterances okay intelligent uh tutoring systems etc okay these are all some sub areas which will be basically covered so if we come to intelligence obviously when we say that artificial intelligence is when the computer does something intelligent we need to be able to define what is intelligence okay so let us now take up the definition of intelligence so when we come to intelligence if you look at the dictionary this is what the dictionary says about intelligence the ability to learn and understand or to deal with new or trying situations okay so if you have a new situation how do you deal with that okay so that is where you need intelligence and how do you reason the skilled use of reason another one is the ability to apply knowledge to manipulate the environment okay or to think abstractedly as measured by objective criteria the capacity to learn and solve problems the ability to act rationally so these are all uh the definition which is given in the dictionary okay so basically it's the capacity to learn and solve problems okay so the ability to solve novel problems okay if you're given a new problem for which you don't have any solution what the ability to do that and the ability to act in a rational manner so rational manner means in a manner which is gives the right decision basically okay the ability to act like the way human beings act okay so this is what you mean by intelligence and artificial intelligence is to build and understand intelligent entities or agents okay and there are two main approaches like you say engineering versus cognitive modeling so basically there are things like um you know the engineering part of it for example the uh robotics etc and then we have the cognitive modeling okay so which we do by programs etc so we will not be going into the robotics part over here basically we are seeing it from the computer science point of view okay so what is it that you require okay in u the case of uh you know learning or or or intelligence right so if you want to act in an intelligent way what are the different uh uh characteristics that should be there in your AI system it should be able to interact with the real world so how do you do that maybe you can have some sensors so if you take human beings we have sensors like eyes nose touch right mouth whatever all these things um so the in the case of the computer also you need to have sensors so that it's able to interact with the real world and find out what is happening in the real world so what you need to do is to perceive to understand and to act okay so um so here uh ability to take action okay so you need to understand what is happening in the in the real world and you need to take action for that reasoning and planning so you you're modeling the external world given some input solving new problems ability to deal with unexpected problems etc then learning and adaptation we need to continuously learn and adapt okay so as the real world changes we need to also adapt ourselves according to what is happening in the real world um our models have to be always updated okay so that is uh these are the things which are necessary for basic intelligence okay so these are some people who have really been pioneers when it came to artificial intelligence so you find many of them were working in the 40s50s and60s okay 1940s50s and60s so the first is Makati so the first thing he's known for is lisp okay so lisp is a list processing language and um it it was uh started in around 1960 so he came up with it around 1960 and uh it it carries out some type of logical programming and non-monotonic he's also known for non-monotonic reasoning and all these things next one is Minsky so he was the founder of AI lab at MIT and uh there's a knowledge structure called frames so he is responsible for that particular knowledge structure then you have Simon uh so Simon is known for the general problem solver so that's a very important uh uh you know software he came up for uh solving general problems uh and of course he's worked on social networks and come up with the power law which is another important thing then you have Arthur Samuel so author Samuel um term coined the term machine learning and he was known for using AI for game playing so in other words he was responsible for all the checker programs which were written in the 1950s so he wrote a series of um checkers programs basically then you have Alan Newell who came up with what is called the logic theorist which is logic based uh reasoning general problem solver okay and and he was responsible for what is called the chess machine then you have Neielson so you have the AAR algorithm so the AAR algorithm is a very important huristic search algorithm which we will be talking about later so he is the one who came up with that algorithm and uh he was he also talk about planning and um uh he there's a book by written by him on artificial intelligence okay which is also a very well-known book so um we'll now talk about what is called the Turing test so what is a Turing test so Turing test also called the imitation game in other words you have a human interrogator who is separated from uh the AI system okay or the computer system okay and there's a teley type with which he communicates with the AI system and if the interlocator cannot tell whether the whether it is a computer or it is a human being okay at the other end the computer actually passes the test so that is what is known as the Turing test okay now what are the capabilities that the AI system needs to have to pass the Turing test that's what we have to see so the one is that it should be able to work with natural language processing so it should be able to communicate with the interrogator so the interrogator asks some questions so it should be able to understand that and it should be able to reply to those questions in natural language processing next is the knowledge representation so whatever information it has it should be able to store it okay so that it has all those information when it needs to answer questions okay or communicate with the interrogator third one is what is called the automated reasoning so when the interrogator asks a question it should be able to reason out using the information or the knowledge it has stored okay and it should be able to draw conclusions answer questions etc and the fourth one is machine learning where it needs to adapt to new circumstances so these are some of the capabilities required by a computer if it needs to pass what is called the Turing test okay so let us see um uh it's a multidisciplinary subject AI okay so number of different areas are involved here first one is of course the computer engineering okay so obviously the artifact we are using is the computer right so the computer is an important uh aspect over here so computer science data science that is storing of the information in the data machine learning deep learning all these things come under the computer science and then you have uh subjects like philosophy psychology cognitive science okay so these are another important uh uh set of u subjects okay another is uh mathematics physics etc so mathematics of course things like logic what we are using logic reasoning all these things will come under mathematics and of course things like linguistics when you do natural language processing robotics where um you have the machine um you know moving from one place to another carrying out tasks etc so that is robotics so all these areas will come under AI okay so all these are something like the parent disciplines of AI okay so let us see how these different topics are used over here first of all philosophy if you consider um logic of course also comes under philosophy you had Aristotle working on some type of logic in in the earlier parts okay uh methods of reasoning and looking at mind as a physical system uh foundations of learning language rationality language rationality all these things come under philosophy and then uh if you look at mathematics you have formal representation and proofs algorithms okay so here what happens is you have algorithms which are intractable basically okay because you're talking about really big um data etc so how do you work on these intractable problems then you have things like decidability computation all these things then probability and statistics obviously you have lot of uncertainty involved so how do you model this uncertainty one way is by using probability or one of the important ways is by using probability okay so these uh that's another area then economics so in economics obviously when you make a decision okay so how do you take a decision okay so the the to taking a decision is based on decision theory and um when the there is more more than one method of doing it you can use what is called the utility theory so what is the method which gives the highest utility among the uh decision uh among the different um actions that you can take okay and then you have rational economic agents etc neuroscience right neuroscience neurons as information processing units so we know that um if you look at the way the brain works okay so it's like a neural network basically so there are neurons which are highly interconnected with each other etc so neuroscience is a very important concept there psychology cognitive science okay how do people behave perceive process cognitive information represent knowledge etc right so when we uh take in things okay how much can we store in our brain right so how do we carry out the knowledge representation how do we do the abstraction okay so we can see that uh some information is stored in the brain some information is forgotten you know all those type of things so that cognitive science and psychology is very important to if we um you know study that it'll help us in AI also computer engineering so building faster computers we know that as the days go by there have been faster and faster computers so AI started in the 1950s and since then there's been so much improvement okay the uh computers have becoming faster and faster and the storage is is gone up like anything right so all these things actually help you in AI and um you're able to get better and better programs and and more uh accurate programs because of this then control theory so so when you have an objective function which you need to uh maximize you use a control theory feedback systems etc and linguistic how do you represent uh knowledge how do you uh work with natural language all these things okay so all these are something like um the parent disciplines making it a interdisciplinary area basically so the foundation of AI is based on mathematics neuroscience control theory linguistics etc okay so mathematics is very important because um um if you see uh it's only after say 1990 or so that AI picked up in all these things okay we had things like uh decision theory utility theory uh mathematics um in speech recognition you have uh mark of processes which processes which came up etc so you find that uh it became a very sort of going more and more towards uh you know uh a full mathematical model of the AI systems so in mathematics okay logical methods are are used okay so when you use logical methods obviously you have boolean logic where you have everything represented by zeros and ones and then you have fuzzy logic where in fuzzy logic it can belong to it can be a member of um um it can be a member uh you know membership with with a membership value basically and then uncertainty obviously um most of from the systems which we are working on is based on uncertainty so when you have uncertainty you need to be able to handle that uncertainty okay and this is done by using probability theory other types of logics like model temporal logics etc you have non-monotonic logic so all these things actually help you to deal with uh the uncertainty so um neuroscience how the brain works okay how close are we to have a mechanical brain uh more recent studies use accurate sensors to correlate brain activity to human thought earlier studies used injured and abnormal people to understand which parts of the brain work and which don't etc okay so neuroscience also goes into artificial intelligence then control theory so what you're basically uh trying to do is uh you in control theory the feedback that is used right so you're trying to use that okay uh and uh see how it can be used in AI basically so linguistic speech uh with speech is a is a big part in human intelligence it's only through speech that we understand things and we use uh these uh you know what we understand um it it goes into our brain okay to form human intelligence so speech is a very important aspect um so so the uh language is important there for example you can see that children uh they would have um you know as they grow older you find that they are able to create sentences which they have never heard of before right so um language is is actually very important and uh we we actually if you see we think in a particular language for example um for example I may think in English right so um the thinking and language is actually sort of very closely connected to each other okay so now what we are going to uh talk about is something like the history of AI okay so uh we'll stop here okay and take a break and continue this in the is in the next session thank you