 Boom, what's up everyone? Welcome to Simulation. I'm your host Alan Sakyan. We are still in the Boston Metro area I'm really excited to be talking about all things related to probabilities Bayesian reasoning, applying that to things like the justice system, the medical system, engineering and project management and execution We have the absolute privilege of sitting down with Dr. Murray Cantor. Hello. Hello. Hey Alan Thanks for coming on to the show. Thanks. Well, thanks for inviting me to your show. It's a real privilege to talk to you Thank you and thanks to our friend Steve. Wait for introducing us. Yes. Huge shout out. Yes And it's so cool how you've been teaching me in just a short span of like an hour Just why exactly probabilities are so serious in our world and why they matter so much And you did a PhD in math at Berkeley and then you've been working over the last 20 years you did work with IBM the last 10 years you've been doing work with applying probabilities to actual execution flows and project management flows and now most recently the CTO and the co-founder of Aptage Which is actually bringing this forth as an entrepreneurial system for people to actually be able to use and apply and make their workflows Much better. So also author two books along the way There's a lot of amazing work that you've done along the way and I'm excited to unpack that and and and Explain to other people why exactly this is so important. So let's start with your interest in mathematics Because math is kind of like the code of the universe Okay Where does why did you how'd you get picked up with it? Why did you fall in love with it? Well? It's interesting. I mean I was always really Math for me was like my easiest topic way back to junior high or something Okay, so I always just assumed that that's what I would be a mathematician and It It I it's more than that. It's easy is that it really gives you a kind of it's both empowering and And and gives you Insights to what's really going on around you. Yes, and and and in a way that is just so The words I Was I thought I used it empowering but I just love to know how things work I sort of get behind magic tricks I get behind how systems really work together and the language for all of that is mathematics Yeah, right. So gives you the foundation for understanding the universe really really well. Yeah That's why I do it And then I discovered over the years that there's lots of opportunity to apply these things to lots of Really important real-world problems And so what I discovered sort of after I got my PhD is I really didn't want to be an academic mathematician You know So I got in that far just sort of a coasting right and then I realized I didn't That that that was fun to know about but I wasn't really turned on by the purity of mathematics I was really turned on by how this really affected real-world things So I sort of retooled myself as an applied mathematician and I've been just sort of Applying that in a whole bunch of different domains ever since so I have papers and physics journals I have papers in geophysics journals. I have papers in engineering journals now. I'm working on Applying AI methods for sports performance And the fact is just wonderful to be able to play in all these different domains And of course aptage in helping people get better in project management. Yep. Yep So this is so important because you're actually Translating the mathematics into applied again applying the mathematics into real-world scenarios that are In many ways either like accelerate or augment performance. Yes, and they also do things like potentially Eradicate errors that occur and this is these are both crucial Right and so yeah teach us about that. Well, I mean the world is getting real complicated perhaps you've noticed Yeah, and There's a fear among some of us that the we are building systems that are too complicated for us to manage And I think you know and and so if we're going to get our arms around that we have to understand the likelihood of events Of what would happen if we do this What would have happened if we didn't do that? That's what Judea Perron in his book of why talks about intervention and counterfactuals And we live in really wondrous times where we have the math to do this and we have the tools online and various Mathematics and power on the cloud that we actually can apply these mathematics to do things so As I tell my friends, it's a wondrous time to be a mathematician right now We have we have this real strong need to work with these complicated systems Think about how hard it is to make medical decisions like you're friend in it Who has who's trying to make the right decision in a In an intensive care unit for a possible most likely terminal patients How do you make the right decision? You need real frameworks to do this. We have those tools people aren't using them And the same kind of thing with building complex The prod the kind of things we're building are highly integrated and highly complex the teams that build them are highly integrated How do they complex? How do we know how they're going to perform? Well, we don't but we can we can make we can do the we can start looking at the probabilities of how they might perform and what's the most likely outcomes And again, we should be applying these tools because we have this problem and we have the You know need meets opportunity It's so cool that you indicated that as we've evolved we've moved towards complexity And how humans find themselves as stewards of this rock and earth and now we have to figure out Okay, we're building these extremely complex systems And yet humans are we're don't we're not necessarily building the knowledge foundations to to know how to manage these complex systems Well, yes, exactly So it so but we have to of course we have no choice. Yes, right So that's what that's the underlying motivation of all this really I love it. I love it. And then we started hinting towards and you you started hinting towards, okay, there's there's There's aspects to this in in athletics or in judicial or in medical or in project management and execution of engineering and design tasks within companies so You know So any in your in your career as you as you slowly started getting to abdige Tell us about that kind that kind of 20 year chunk of time when you were like, okay I'm applying this these are the and these are the key areas that I want to see this applied As soon as possible. And then how did you start? Yeah, so my main journey is through project management Okay, and for reasons that are Yeah You know, there's you don't want to get into too much history because then that'll take up the whole time I'm old enough that the stories take forever But uh the cliff notes very cliff notes But I I was working on the graphic I was working in IBM on the risk system 6000 Which uh was ibm's answer to the unix workstations Based on uh reduced instruction set technology And so we were building a uh new hardware architecture a new version of unix aix 3.0 and Uh a whole new graphics. I was on the graphic subsystem part A whole new set of graphics adapters whole new apis We had to get this thing out the door in a certain amount of time Ridiculous, you know, you think about on the face of it. Who would do this, right? And we did it Okay, and it came in a year late, but that's In the whole history of that and um What I realized was that um The Ability of people to manage these kind of we were being asked by our project man by our stakeholders When would we be done with this? Effort I had literally Four graphics adapters three new apis on an unstable operating system And they were asking me when we're gonna be done And the answer I don't know Right So I made up a date and everybody was making updates and they were playing what I would call management chicken Which I was betting this guy was going to turn himself in before I was going to have to Okay, oh, we would have been on time if Right or Because he's late. He gets all the blame and no one notices that I'm late. So but and the day it really hit me I was in my I was in my manager's office And he puts up the Gantt chart In my task and I'm supposed to have I have on some nine-month task And all this building out the new Device driver for this new thing and he says according to this you're 48% done, aren't you? and I said I had a choice and I said I could say no and he did not again until a whole bunch of questions is why not I had no idea So I said yes and everybody else said yes And then The surprise came later Right, so I so I what I became to realize is That I was being asked a ridiculous question that really mattered to the business Okay, that and the sort of adversarial relationship that project managers get into With their management where they the management asked them a ridiculous question and the And what happened with the agile manifesto? I think one way to look at it was that The developer said we're not going to do this anymore You know, it's it's it's ridiculous To do this and some people would take it to extreme saying you're going to get it when you get it You can't run a business that way either This is when the math started kicking and I started and I started realizing that the time to complete a project Like this There's no such thing. Oh, and this is the other thing if I if I gave them a Prediction a prediction of when I was going to be done I was measured of whether I was going to be whether that number turned out to be right or wrong Yeah, right. So I started leading a Actually, one of the things I did in IBM and started leading a interest group in project management And I gave a talk at one of the conference at one meeting where actually Actually Grady Butch who was a great Developer of object methods was sick and so they asked me to come in So I played Grady at the conference and I gave a talk on embracing uncertainty And my main idea was at the time to complete isn't a known quantity It's a probability distribution or what we call random variable And the best thing rather than get into a fight and went and so you don't there's no one number Which is right. There's just a probability of things that might happen And and so the right question is do we understand the shape of this distribution and should we manage it together? That's the management and the thing And the room got real interesting about a third of the room thought I was nuts And a third of the room got sick and a third of the room didn't understand me And a third of the room said, oh my god, thank god someone's been saying this. Yeah, it's about time Okay, which and and so I've been building out those techniques ever since then Which is essentially this lets you and I Management agree that this is the likely set of outcomes This is the the time to complete is a random variable. Let's look at its distribution And oh by the way, let's manage that together because we can affect the odds It's not true gambling. We can change the odds by working together. Yeah, okay, so that and um There's a good example of where you go and to say ask an engineer Manager of engineers. Hey, we have this project that we'd like to get done. How long is it going to take to get done? Well, I have no idea. Well, it's going to take three years. No, it's going to take one year. No, can you get it done in six months? Yeah, yeah, we could potentially get it. Yeah, six months would be the you know worst-case scenario Well, what would be the best-case scenario? Well, well, maybe two months Well, then that already brings it to a more of a period of certainty and then assigning probabilities to It looks like probably four or five months would be a pretty safe bet Well, we actually use this is is is triangular distributions where we Where we look at and what do you think is the most likely thing? And we build a distribution where it's no probability below the lowest No probability above the highest and the highest then we sort of slope up to the max and By the way, this is a technique that's been widely talked about by Douglas Hubbard in his book how to measure anything You know, so the other thing that's interesting is all these ideas are out there It's just a matter of putting them together. Yeah, it really is into the systems of complexity to make them more certain That's right. And so each task has a has a Has a has a triangular distribution associated with it Now tasks have dependencies and they go in parallel and all that So computing the joint probability of the time to complete Given all those things. Well, that's hard. There have been tools to do this for a while using Monte Carlo simulation But they're they're hard. They've been hard to use one of things that but this is getting us started But the other thing that that I realized was that the team completion rate or velocity Is also not a number that we can know It is a probability distribution again into random variable And particularly for things where our novel projects where it's not like milling light bulbs where you can get to a steady state You never get to a steady state, right? What does it mean to Have a steady workflow for finding requirements? You know, if I spend if I spend Five guys working at a certain rate to paint a wall I I can predict when the wall get painted. I have five guys working on requirements. How do I know how long that's going to take? Right and adding a sixth person may slow it down as Fred Brooks has pointed out Right, so I can't apply those sorts of mathematics It's applying the wrong math again. It's all about the math But there is a probability distribution that you can learn about rate about rate to completion So now I have to learn that probability distribution and I have to include that with the various sizings And roll that up and now now the math gets complicated because I have to use Bayesian parameter learning techniques and whatever That's what led to aptitude finally. Yeah And so we do all that and the whole trick now is to make that real consumable And what we have found is that teams really adopt this nicely because we're just using Rather than throwing away information that they already have because they know they have known unknowns Let's capture the known unknowns and use that rather than throwing it away Yeah, and you know rather than arguing everybody into a single number. Let's not even bother Let's take the known unknowns capture that as part of what the project notes And all that rolls up and the point is that this is this is this is an example For project management that we're doing But the same kind of reasoning that time that the Likelihood of an intervention working for a drug Or the likelihood that you really are drunk if you test positive on a test All these things are random variables. Yes. All these things are probabilities And as a society we should get to a point. We understand them and use them properly Yes, right. Yes, this where you're just Where you're going there. I really want to get to these examples for everyone Sure those examples are so so important because they're super applicable to our lives and then also I want to make sure as you were talking about things like The triangular distribution and you're talking about how in Something like a rate of a velocity of a team being able to perform and you see that they oh, they hit the four five month mark They're getting faster and faster and more effective at completing tasks and then also the You or you mentioned this to me earlier as well the adversarial component is gone more so because you're right the there's not as much of a Tension going on between Between nodes in a hierarchy in a in a company. Right. Yeah There was also another thing which was the I think this is so crucial when when you when you figure out that you have You know, you're give your engineering team potentially is given a task and the task is to get This done and you initially go for the lowest hanging fruit first That that is going to bite you in the butt later because it looks like you're on on map on on You're going to get there and then at the end you're doing the hardest things and you miss your deadline Because you're doing the hardest things at the end. Yeah, so let's talk about that. So There's Classical project management. There's something called earn value Okay, and the question is how much of the work have you done with the dollars you've spent? Okay, and the idea is is to Claim earn value for work that's completed. Okay, so and uh, there's still This is still very strong and there's various teams that have what would they call the earn value police? Which matter you do which which look at your projects the trouble with that is It you know people People behave by trying to optimize the measures that they're measured by Okay, so that's why I'll tax policy Used to at least encourage home ownership by lowering tax The thing or a few you tell a better example You tell policemen that they're going to be measured by how many tickets they write They're going to write a bunch of tickets. Yeah, whether or not they they should have written a bunch of tickets They're just going to write a bunch of tickets. Yeah, and then if there's lobbyists involved in medical processes And then there's going to be more of those medical processes, but how many of them were actually necessary Goes on and on right. Yeah Uh, there's a whole bunch of stuff about people being rewarded for prescribing drugs. Yes For example, okay, so earn value was so the project manager saying well I'm going to get rewarded by doing the easy task because the length and claim earn value Okay Because there you go. What I noticed was and others that Real progress and project management. I have this novel risky project. There's a lot of uncertainty I shouldn't be measuring progress on whether I've done the easy stuff I should be measuring progress by whether I've reduced the uncertainty Mm-hmm, right because because the way it is the day I started I have a whole bunch of uncertainty If I'm going to ship tomorrow, I better not have much uncertainty Yes And so you're what what you're doing is you're asking people to tackle the hardest things at the start Right as best as they can that way they maximize the certainty Right, they they remove the uncertainty over time rather than just completing tasks. Yes, correct Because that because that's how I know you're going to ship Right and that's not earned value In fact, there are ways you could assign your value to reducing uncertainty and everybody would win, you know, whatever And so one of the other things we did in after just actually show a graph while seeing how the uncertainty is being managed and whether it's going up and down Okay, because what we're trying to get everybody to collaborate on and by the way, it's not just the manager with their management It's also the managers with their staff 360 kind of thing Is that we should all collaborate on removing the uncertainties in our project so that we can make the Goals that the project has now sometimes these goals are time-based and sometimes other things But we should know again if if if I've committed to ship tomorrow I better know for sure that I'm going to be able to ship tomorrow. There should be no residual uncertainty in my project Right Now if i'm four weeks out and there's a whole bunch of uncertainty which can happen from your Scenario of the lowest hanging fruit. It's exactly what happens. They push the uncertainty to the end of the project where they have less time to deal with it So there's an old saying in project management, which is 90 of the project Is spent 90 of the project is spent with 90 of the earned value claimed Yeah Okay, and this is why it was cool about math now. We know why that is Because it didn't work on the uncertainty They they worked on on claiming the they did the early they did exactly what you shouldn't do Which is is is you increase the uncertainty by putting it off to the end of the project Yeah, so not only are you increasing certainty for the teams of project managers and engineers and companies that really want And now aptige is being used by companies and they're actually seeing Great benefits by using the service and i'm i'm totally seeing lots more Companies catching on to this. Oh, we can have more certainty in what we're building We can have more certainty in our timelines more certainty in our teams So then there's also this sort of it's it's important to identify that the probabilities within Within how to actually calculate and synthesize all of the probabilities Because there's so many variables that go into this it becomes difficult to actually Figure out where the where the where the signal ends up being in You know, you started giving these examples of when you're when there's Where people are getting paid paid based on the amount of tickets they give out or based on the amount of pharmaceuticals they give out And so then that's creating these Unjust systems right in medicine and injustice And so this becomes this becomes a problem and so we can potentially mend that problem Yes, yes and There's an important point you make Which is There's two ways to deal with uncertainty And one way is to have don't do interesting things boring No, but so what are the things that happens if you insist this happens though if you insist that I have to make a date I'm not going to take any risks Right and I'm going to move out on all the risky interesting there's A risk reward thing here. Yeah, which do and part of what's going on in our system is the lifespan of Companies is shorter than it used to be Right and so in order because there's more disruption going on so they have to take on novel projects About so it's not necessarily so you don't necessarily want to reduce the initial uncertainty You want to have methods for dealing with it? Yep I've seen company I've seen I've seen Companies who because of this thing I you know Rewarding this is not a false reward thing if you reward project managers for hitting the date every time With the original content at the original budget What you're going to see is more more boring projects less and less risk less less Interesting stuff So they actually move the they actually manage the value out of their development organization And in order to stay relevant you have to take big risks Shot risks innovative risks. Right. So you have to know how to manage those rather than avoid them And this is what I used to call embracing uncertainty Right, let's not let's not be afraid of it. Let's embrace it make it an opportunity But you need the tools to do that and that's where the problem and if you treat uncertainty as the on a risk and on Risk as the uncertainty in meeting a measure that matters to the business Then the way you assume risk is by understanding that uncertainty and working it off Yes, and that's what the tools do and we had this We have this really good example where if you are building up the collective learning of civilization on top of let's say scientific literature and we're building it up and we're building it up and Something that's that was discovered just a couple years ago is Higgs boson And if you're trying to add something like his Higgs boson to our understanding of physics And put that into the knowledge foundation of civilization You have to be extremely certain that that is correct And you have to have a very very low probability of that. We were wrong about that versus with medical literature It seems as though you can you were okay. No problem one out of every 20 times Ah, it's no no problem that that's wrong that butterfly effects and cascades out and Ends up causing a lot of harm because we are letting uncertainties creep into the foundation of what we're trying to build up as a knowledge All right, so the two issues here two two things talking about here one is sort of the standard so A lot of us call ourselves Identify as being basions That's a thing to be What basions believe is Sean Carroll talks really nicely about this in his latest book the big the big picture Which is we believe in things for which we have evidence Yes, okay, and we don't think anything is a hundred percent true The the stronger the evidence the more we believe in it And so the Higgs boson is the example is and I actually look this up So one so they get more and more evidence that the that they had found the Higgs boson They saw a signal then they saw another signal then they saw another signal and they started looking What is the joint probability of seeing all these signals? And at what point did they decide they they were ready to declare? And it turned out I don't remember the exact number, but it was something like one in ten billion That they so it's not a hundred percent, but it's it's a really good bet Okay Now in the medical literature as you point out people will publish as there's no If the chance of them being wrong is you know, they pee This significance of point oh five that means that there's only a five percent chance that this happened by chance By chance, yeah But that means that one out of 20 articles just happened by chance And then the medical literature is there's a whole lot of writing about this and then the stuff that Then there's no one ever gets paid to find to actually do the experiment again again And so then we get into a lot of in unreproducible effects and And so it's really hard now the literature is getting better because people are getting smarter about this. Yeah Yeah, and and whatever but it's and shout out to our friends Peter and Lindsay and and uh helen for doing the hoax papers because that's start that's now starting to uh to To to to point out how we can't be letting the the the improper Scientific evidence enter in the foundation that we're trying to build up on yeah, right But the point of that what all this is that there's nothing hundred percent certain So what things so what we need to do is be certain enough to make the right Decisions most of the time Right or to make the right decisions in a way that is beneficial to a system Is economically makes sense Whatever so I One of the things exercises where I'm working with the team where we're looking at sustainability of uh new system New systems and their impact on the sustainability of the oceans Yeah, okay now You know, so they're going to build a fish farm and they're going to do this Okay now we can't be a hundred percent sure this is going to be better than another model But we can build a Bayesian net kind of model to find out What is the what is the likelihood of this improving things? And then we can look at the investment and decide if that's the same kind of Same kind of investment and so their investors in the world who care about this They want to invest in things that both make money and improve the environment Yes, but they need again this kind of Bayesian reasoning For that to for that to work We live in really wonderful times though that we have As I said, there's Judea Pearls book on the book of why but there's a whole lot of mathematics behind that He has a book on causality And then there's fentanyl's book on Bayesian nets and their tools And so we have the methods It's just a matter But these are as we kind of and say these are thinking slow kind of systems But we need to be applying them. Yeah, that's really the point And and the more that I think that we think slow collectively geopolitically as a planet The less oops moments will have and we don't really have room for oops moments anymore We just had max take mark on this show from future life institute And you when you had an oops moment with with, you know, the seat belt or the airbag with the vehicles That's fine. I mean a couple maybe hundreds of thousands of people died before that. Okay That's okay But a nuclear weapon a mutually assured destruction oops moment A malevolent artificial general intelligence oops moment. No, we could wipe out It's an existential crisis. So When when we point out the the sort of Um Oh, yeah, this um this like This the the probabilities that that that you can actually Show as you as you go and and and build them out inside of computer models Is so gorgeous because you're able to tweak the different variables like a A blood alcohol a measure measurement device a breathometer, right? That you can't Rely on it a hundred percent of the time and the same thing's true about something that's trying to find some sort of a medical ailment or something and so because of that you end up getting a certain amount of false positives And then now you can actually model that as a probability you can also add another variable like age like sex Like potentially what type of car they're driving what their socioeconomic status is all different types of religion etc And then that those variables we can actually get closer to certainties in understanding these things And then we need to make better tools that are also less uncertain so the the point here is If you just give person a without any other information And you give person a dw i test And they test positive That doesn't mean they're first of all that doesn't mean they're driving under the influence Because if you drive under the influence You're most like you're very likely to test positive with a test But that doesn't mean if you test positive you are actually are driving under the influence Now those and it turns out the it's remarkable how we would like a five percent false positive Of kind of number which is sort of where these tests tend to be And there's only one chance in a thousand or one chance in a hundred that a person is actually drunk anyhow This would be going down a row of a hundred people and only maybe five of them are actually drunk And you try and see out of the 95 people that aren't drunk how many of them would get falsely flagged Yeah, and it turns out that the percent of people who who are Would test positive, but aren't drunk is remarkably high like 20 percent Into most of these things Okay So now but on the other hand if they're driving a red sports car and they're 18 years old You can sort of show now the probabilities two-thirds higher. Yeah now in in reality now also if they're dry, you know now So If you're only relying on the tests you just stopping people at random and just relying on the tests They actually threw the there was a whole bunch of cases They actually threw out in the state of texas because that's what they were doing Now if a person's driving erratically and you give them the test now that's different But that wasn't what was happening So understanding the difference between if you test positive and If you're drunk, then you're going to test positive The probability of that is different than if you test positive, then you're drunk And that's true for by the way a lot of medical tests If you test positive for this disease If you have the disease you're very likely to test positive if you if the disease is rare And you test positive. It's not all that likely you have the disease And what the doctor is supposed to say is I'd like to give you a different kind of test Okay, so there's another thought that occurred to me about Why bayesian reasoning is really important The way bayesian things work is you start out with some prior belief And you adjust that belief with the evidence Now it turns out that you for a lot of things Whenever your prior belief was if there's enough evidence, it doesn't matter You get to the right thing and I can show you examples of that If you have if you 100 believe in something No amount of evidence is going to change you Okay, which is why it's so important to leave Area in your ideology open have openness that way when you someone is potentially Adding to your perspective to help you see the world in a more true or way that you're receptive Right, and I think thomas bayes three hundred years ago. We're actually wrote about this This is a very old insight. Yeah. Okay. Now. Let's look at some of the climate change discussion There are people who would know amount of evidence will let them believe They are so convinced 100% convinced That man isn't changing the climate That no amount of evidence will change their minds The thing is so part of what is important for us as a society Is to understand that there's all that to never be 100 sure of anything And always look at the evidence so that you can update your beliefs. Yes. This is so important I love that always be looking at Learning more evidence adding more evidence to the equation to change your beliefs to change your beliefs that it's almost It's a steering process. It's a steering process and if the ones that are willing to To learn from more evidence as they steer they can steer into more truthful Directions, right? Yeah, and and the math will take you there math will take you there So it's so it's sort of roll this up to the beginning thing we talked about why mathematics see so This little theorem about Bayes theorem Which says that That what you do is you start out with some belief and then you update that belief based on evidence And that if you are a hundred percent certain in something no amount of evidence will change that Is actually a way to understand Some of the social behavior we're seeing that's right and some of the conversations Right and why you know, you can you can expand that to gun control and whatever Portion all those things policy. Yeah, it's because we're that where we got trapped in a binary Thought process that I'm a hundred percent in my ideology right and so no evidence can change that That's what the math will tell you That's so cool. So we can actually bring a mathematics lens to Our discourse amongst one another and so minimally what we can do is have a a percentage of Of openness. Yes on ideologies that way where we leave an ability to stay open or like I like to do and I like to Burge people to do is to stay free thinking stay independent And that way when you see people that are just watching fox news or just watching CNN the day of blind spots That you've got to watch both and you have to learn from other people that actually live You know in different parts of the world and get firsthand information yourself I like taking a math perspective on it binary thinking is 100 percent on your ideology But the gray area the nuance is having 50 50 or 60 40 and even 80 10 80 20 90 10 It just needs some amount of willingness to learn from the evidence and see where it takes you Yeah You know, so i'm 90 sure of this, but if you can show me some evidence i'm willing to change Boom hurt him right here. Yeah, that's it. Yeah, so this kind of thinking is just So so in project management, we need to learn from the evidence of how the team is performing in uh medical you know and the uh Medical things we've got to apply bayesian reasoning and to understand the probability of what various interventions would do properly Right and then in looking at things around us as we learn together society We have to learn from the evidence that we see Yeah, that's so to build the most just systems that enable the best flourishing Right and so us to come to correct conclusions And make good make good decisions Murray this is so Fascinating that you're working on the applying the mathematics and applying these probability measurements into maximizing the efficacy of of these different systems justice medical Engineering projects. Yeah. Yeah, I love it. So Murray Tell us about where you see all of this going because it's going to be applied into all of these different industries We're going to basically put on a a math lens as we enter into probabilities Make it more certain. Okay. Well, I don't know but I'll tell you I'll just sort of just Something I'm just seeing now Okay What do I book? Yeah, I have Some belief in this, okay Uh, because I have I try to actually live and not only just talk about being amazing So some evidence I'm seeing. Yeah, okay Judea Pearls book the book of why is actually number one in its category on amazon Okay, I'm having all my friends read it and they all come back to me and say wow Right. It asks really thought-provoking questions. Right. Yeah, okay the uh You see articles more and more there was a recent article in the wall street journal by wadowski burger talking about the same kind of things so people are beginning to So it's beginning to permeate at least some level of the more general population These techniques go back. You know, I don't know when they started, but you know, they really can be traced back with 300 years, but Judea's pearl work on bayesian net And then fentanyl's a geno risk tool and there's the bayesian labs geno risk tools So we have so what we're seeing now is it's it's sort of coming out of the It's moving down in in sort of the intellectual planes to more and more thinking and because we have The tools to do this We have the Compute power of the web Like aptish could not have existed if it weren't for the cloud It just couldn't have okay The tools that I built it in are python modules and things that Make the programming efficient So we and people can build other tools so the technology is here and people are beginning to think about The next gen of AI being more about Can I read an image better or play chess better or play go better? But can I start making better decisions about interventions? Yeah, right, but that is going to be the next generation of AI We're training it on narrow systems now to also make and now making it more general on better decision making for Human oriented Yes, exactly. Yeah, so that's that I already see that beginning to happen. I love it that so that's that's Who knows where that's going to go, but it's already It's already beginning to happen Because humans are so irrational at times and we we think you think You think yeah, yeah, we we are we we have a lot of emotion. We have a lot of uh sleep deprivation We have a lot of addiction. We have a lot of yeah tendencies that we we don't even Believe that we actually do but admitting that we do is the first step to this Dissolving it as a problem and letting things like a non-biased Artificial intelligence help make these decisions when that's not sleep deprived right so emotional So these systems won't take over our jobs, but they'll really help us do these kind of jobs better and give us a An objective non-biased You you told me the stuff that you believe about These tasks and how we've looked at how these teams have performed or you tell In the other examples, I'm going to I've been looking at it Medical and you've told us this data or this is data that you've measured from population studies Right and now you have to decide whether or not to administer this treatment or this surgical approach That might kill somebody should you do it? Yeah, you know people aren't People are lousy at that Building a system that would know how to at least give you the best answer Would be real helpful. It wouldn't replace the doctor. It would really help the doctor help the doctor But there's because Murray how can we form this basian network in our head calculating dozens of Variables people are real lousy. So there's the classic example of the Monty Hall problem Do you know the Monty Hall problem? I don't think so. Okay, so the Monty Hall problem is is You know based on uh, let's make a deal Right So the deal is It's a standard problem. There are three doors Okay, okay, you know the three doors problem. I think so. Yeah continue. Okay. So behind one door Is a is a car and between the other two Doors is a goat Okay, and the presumption is you'd rather have a door you'd rather have a car than a goat I know some people would rather have a goat than a car, but most people would rather have a car than a goat Okay, and so you open up you choose a door And Monty Hall opens up a different door And shows you behind that door there's a goat And he says do you want to stay with this door or do you want to Or do you want to switch doors? So now you have a 50-50 chance of a car? No, you don't You don't that's that's always the that's the thinking fast That is that's the data counter insist the thinking fast and Paul Erdogan got this wrong, by the way Paul Erdogan got this wrong right when this was first published when this was first published by uh, what's Laurel von Savant or whatever name was She got the answer right and like one of these popular magazines And a lot of people wrote in and said she got it wrong Including Paul Erdogan Okay, so there's two goats in a car you shown one goat and now now so did you switch Do you switch and the answer is your chances are doubled if you switch of getting a car Now almost no one can get this right in their heads Okay, it's um, I think I started getting a little taste a taste of it that that the that that That the door that you had selected was Was a door that Potentially had the car behind it And then you were shown that one of the that one of the doors that didn't have a car Yeah, right and the question is should you switch to the other door? Yeah, correct and You double your chances by switching to the other door and almost no one gets this right intuitively continue now Right, okay, so the answer I have my explanation of why that's true Please you could by the way you could run you could be a frequentist and run a hundred experiments If you discovered that yeah, or you could just think it through Continue, okay, so the way you think it through is the chance of you being right was one and three Yes at the beginning right Okay, so that means there's a two and three chance that it's one of the other two doors Yeah He showed you the so did that change by the him opening a door Well, yeah No, it's still two and three that it was one of the other two doors It was still two or three, but now it's now, you know and he did you a favor by showing which of the other two doors not to open You see Yeah, so the point is you got a new piece of information. He opened a door that didn't have a goat You know more than you did before So you had a one and chance three of being right He opened the door. He still had a one and chance three being right And not only that you actually know where the two and three door is. It's the one he didn't open Oh, so that door has two and three chance now Yeah Why does the door that he didn't open that I didn't select have a two and three chance of being right because it had Because he knew it had a goat That's the point. He didn't open one by random. He opened one. He knew had a goat That's why this works Okay Uh-huh. You see So the whole system. So this is the point that pro makes there's a whole system here Right, the system is you choose something. He chooses a door based on your choice It wasn't an independent choice The people think it's 50 50s because it's an independent he thinks that they think that he was working independently But he's not He made a choice based on What you chose Okay, okay. Okay. Yeah, we don't how do we know that he Okay, because that's the rules of the game. That's the rules game. Okay, okay Now you could go into well. Is he going to follow the rule? Sure. Sure. Yeah But the point is that that even that amount of And by the way that solved he's amazing reasoning gets you that real fast. I can build you amazing at the shows you do this That's cool. Okay. That's cool. Yeah. Yeah See this one I'm talking about. Yeah, because it makes gives us better decisions It helps us with slowing down to think right Right. So you and that's exactly it. We have these complicated systems that behave like the monty hall problem Yeah And we have to understand the probabilities of making decisions based on that Yeah, so this is sort of like a toy problem, but it's not really Because it shows several things one is that almost no one gets fast thinking system works I've had real arguments with philosophers. I had to take them through this very carefully Yeah, yeah, yeah, okay I have some questions. Yeah, okay, so let's do that. Okay. Okay. Some questions on the way out um First question is what would you say is a central guiding principle of your life? um As I said, I am I tell people I'm a devout Bayesian Devout Bayesian I learn from evidence. I believe in things for which I have evidence Yeah, right. That's a that's very important. Yeah, no essentially that's the guiding principle. That's beautiful. Okay How about there are others but that's of course, okay um How about How about Do you believe that we're in a simulation? Oh All that stuff What do you think? Oh, you think this is a base reality. Yeah There's there's wheels. Yeah. Yeah. Yeah, even though we're making simulations right now. Yeah And I'll tell you why you know, I actually have thought a little about this Okay, because the who have this simulation is so so for example You lose your keys five days later. You find your keys and then you remember that's where you left it The simulations are real good Right, it's remarkably consistent, right The world we live in It's so consistent. I don't believe it could be a simulation anymore Really? No, it's it's it's so you don't need it either But it's also I'm also sort of amazed at how There's no that things are always entirely consistent. There's no Glitches in the simulation Ever well, sometimes it feels like deja vu is a glitch. Yeah, well, but that's something else we can talk to We can talk about. Um, how about What is something that you believe to be true that almost no one else believes to be true? Oh, I don't know if there's anything like that um Now I don't think there is anything like that. How about What is the deepest emotion you've ever felt? Oh god Oh, well, I don't know. I mean, uh I even hope I can talk about that I mean, I get What's interesting? I think a better way to response this. I'm very rational about stuff But I do respond very strongly emotionally to things. Yeah, you know what? Um, and I'm actually Believe myself be very empathetic. Yeah So like watching some of the things that are going on with like at the border. I get really angry For example, I don't know. That's the deepest emotion I've ever felt. Yeah, but um Why don't you want it? Why couldn't you talk about it? Like why did it come up that you didn't well, I believe I mean There's not a deepest emotional. There's just not a deep one of your deepest emotions. But yeah, right well, uh They're all very standard. I mean I'm deeply in love with my wife of just These and the children but right now I'm just so in love with my grandchildren. Yeah Yeah, that that's very deep. It is. Yeah, I mean and Love them to death And of course because they're still at that age. Yeah, yeah Right, but I but really all of them. Yeah And uh Deep love for family. Yeah, it's like that Murray. How about Last question. Okay What do you think is the most beautiful thing in the world? you know This is gonna be you can't not be syrupy. So, you know or soupy about this but the fact the Thing is, you know, I really believe that we're a small speck in the big universe and the universe the galaxy song and all that right, eric, I don't know all that stuff but You know the fact that We attack them that we really develop emotional attachments for each other that It's it's it's that we have empathy and we can really Emotionally connect I think is remarkable It's really that Yeah, you know, yeah It is this this sort of You know, we evolved I understand how that evolved and all that but love and and and and of Emotional connection is a remarkable thing And it's it's still a great mystery. We'll learn more about it as we learn more about consciousness and stuff We're learning more about it now. It doesn't make it less beautiful Yeah, this uh As you say that it reminds me of when you look at someone's eyes How crazy is it that you can aim to try and understand the last 20 50 70 years of their life. Yeah, that led them to the point where they're at, right? Yeah, yeah You know and the con and you know, you can without with or without a religious thing Our consciousness doesn't it's things that blend Right, and that's what that's a wonderful thing Murray this has been so enlightening. Well, thank you. I've learned so much And I hope we all have taken away the thank you so much. Thank you so much the the opportunity the importance of probabilities The importance of applying mathematics into our lives across all different domains justice medicine projects Mm-hmm This is so important to do and artificial intelligence is helping us with that The links in the bio for aptige go check it out aptige.com go and check them out Also, give us your thoughts in the comments below. We'd love to hear from you Go ahead and see what we can do with math and probabilities and applying that into our world We'd love to see you create and you build go and manifest your destiny into the world everyone much love And we will see you soon. Thank you so much out. Thanks, Murray. Bye. Bye. That's all she wrote. Okay. What fun You