will tell you exactly what move tomake is theo von podcastthe optimal move to make nomatter what your opponent happens to dothis type of thing is not quite aspossible for a much larger game likeCheckers or chess for example wherechess is totally computationallyuntractable for most computers to beable to explore all the possible Statesso we really need our AI to be far moreintelligent about how they go abouttrying to deal with these problems andhow they go about taking thisenvironment that they find themselves inand ultimately searching for one ofthese Solutions so this then was a lookat search in artificial intelligencenext time we'll take a look at knowledgethinking about how it is that our AISare able to know information reasonabout that information and drawconclusions all in our look at Ai andthe principles behind it we'll see younext[Music]time all right welcome back everyone toan introduction to artificialintelligence with python last time wetook to look at search problems inparticular where we have ai agents thatare trying to solve some sort of Problemby taking actions in some sort ofenvironment whether that environment istrying to take actions by playing movesin a game or whether those actions aresomething like trying to figure outwhere to make turns in order to getdriving directions from point A to pointB this time we're going to turn ourattention more generally to just thisidea of knowledge the idea that a lot ofintelligence is based on knowledgeespecially if we think about humanintelligence people know information weknow facts about the world and usingthat information that we know we're ableto draw conclusions reason about theinformation that we know in order tofigure out how to do something or figureout some other piece of information thatwe conclude based on the information wealready have available to us what we'dlike to focus on now is the ability totake this idea of knowledge and beingable to reason based on knowledge andapply those ideas to artificialintelligence in particular we're goingto be building what are known asknowledge-based agents agents that areable to reason and act by representingknowledge internallysomehow inside of our AI they have someunderstanding of what it means to knowsomething and ideally they have somealgorithms or some techniques they canuse based on that knowledge that theyknow in order to figure out the solutionto a problem or figure out someadditional piece of information that canbe helpful in some sense so what do wemean by reasoning based on knowledge tobe able to draw conclusions well let'slook at a simple example drawn from theWorld of Harry Potter uh we take onesentence that we know to be true imagineif it didn't if it didn't rainthan Harry visited Hagrid today so onefact that we might know about the worldand then we take another fact Harryvisited Hagrid or Dumbledore today butnot both so it tells us something aboutthe world that Harry either visitedHagrid but not Dumbledore or Harryvisited Dumbledore but not Hagrid andnow we have a third piece of informationabout the world that Harry visitedDumbledore today so we now have threepieces of information now three factsinside of a knowledge base so to speakinformation that we know and now we ashuman can try and reason about this andfigure out based on this informationwhat additional information can we beginto conclude and well looking at theselast two statements Harry either visitedHagrid or Dumbledore but not both and weknow that Harry visited Dumbledore todaywell then it's pretty reasonable that wecould draw the conclusion that you knowwhat Harry must not have visited Hagridtoday because based on a combination ofthese two statements we can draw thisinference so to speak a conclusion thatHarry did not visit Hagrid today but itturns out we can even do a little bitbetter than that get some moreinformation by taking a look at thisfirst statement and reasoning about thatthis first statement says if it didn'train then Harry visited Hagrid today sowhat does that mean in all cases whereit didn't rain then we know that Harryvisited Hagrid but if we also know nowthat Harry did not visit Hagrid thenthat tells us something about ourinitial premise that we were thinkingabout in particular it tells us that itdid rain today because we can reason ifit didn't rain then Harry would havevisited Hagrid but we know for a factthat Harry did not visit Hagrid today soit's this kind of reasoning this sort oflogical reasoning where we use logicbased on the information that we know inorder to take information and reachconclusions that is going to be thefocus of what we're going to be talkingabout today how can we make ourartificial intelligence logical so thatthey can perform the same kinds ofdeduction the same kinds of reasoningthat we've been doing so far of coursehumans reason about logic generally interms of human language that I just nowwas speaking in English talking inEnglish about these sentences and tryingto reason through how it is that theyrelate to one another we're going toneed to be a little bit more formal whenwe turn our attention to computers andbeing able to encode this notion oflogic and truth and falsehood inside ofa machine so we're going to need tointroduce a few more terms and a fewsymbols that will help us reason throughthis idea of logic inside of anartificial intelligence and we'll beginwith the idea of a sentencenow a sentence in a natural languagelike English is just something that I'msaying like what I'm saying right now inthe context of AI though a sentence isjust an assertion about the world inwhat we're going to call a knowledgerepresentation language some way ofrepresenting knowledge inside of ourcomputers and the way that we're goingto spend most of today reasoning aboutknowledge is through a type of logicknown as propositional logic there are anumber of different types of logic someof which we'll touch on butpropositional logic is based on a logicof propositions or just statements aboutthe world and so we begin inpropositional logic with a notion ofpropositional symbols we will havecertain symbols that are often timesjust letters something like P or Q or Rwhere each of those symbols is going torepresent some fact or sentence aboutthe world so P for example mightrepresent

(Editor in charge: article)