foreign to gaining the technology leadership Edge a podcast for Tech Executives we provide strategies and tactics to help Executives succeed and further their career goals with interviews from industry experts leaders and innovators this show will surely get you on the edge of your seat with thought-provoking advice on how to stay ahead of the competition all right well thank you so much for listening to gaining the technology leadership Edge this is our very first episode and the show is going to bring to you the brightest Minds in technology with the idea being that you can learn more about your craft today's guest is Noah Healy he's a game theorist working on better economic systems after training in nuclear engineering he worked for some tech startups studying the mathematics of computation led to patent work on a better Market design welcome to the show thanks for having me here Mike so let's start off right at the top and let's have you explain what a game theorist is so game theory is the mathematics of strategy and it deals with situations where you have multiple agendas um uh that can lead to situations of both conflict and cooperation uh and there's a lot of counter-intuitive things that we've learned over the last 100-ish years uh by studying Game Theory but there's also a lot of optimizations that are available from actually getting your strategic picture sort of written down in in numbers and figures and being able to use some of these tools from operations research and other subjects like that okay so you know in I like to always bring things back to the real world for people I mean we're obviously our audience is technology professionals so they can pretty much handle what you what you give to them but how would you apply this in the real world like what would be a real world use case uh well one of the most prominent use cases uh for game theory is uh an application of something called the stable marriage algorithm uh which is used to match up doctors uh with their residencies in the teaching hospitals so the stable marriage algorithm basically sort of is based on this this problem imagine a village with an equal number of boys and girls each one of which has you know whatever their preferences are uh can you come up with a set of marriages of these of these young people such that all the marriages are stable that is anyone matched with someone who isn't their first choice will not be able to break up the marriages of any of their higher than current Partners choices because in each of those cases the marriage will be to somebody that the other person prefers to them themselves and when first stated it sounds like well that's you know maybe it doesn't always work um it sounds like it would be very difficult it turns out that in fact no matter what the two sides assuming they're you know equal sizes and stuff like that no matter what the two sides preferences are there's actually always at least one stable Arrangement uh and there's actually a trivial algorithm that will give you one of the stable Arrangements uh and the trivial algorithm is you pick a side boys girls usually boys uh and line them up and the guy at the front of the line picks the girl he wants to marry and you just go down the line if any guy picks a girl that's already been picked then the girl gets to pick which one she prefers and then that person goes to the back of the line and he can't pick her again and once the line is finally exhausted you're done you've found a stable Arrangement and that's actually how it works uh the teaching hospitals have the the you know results from the medical schools and they they list out who their favored uh persons are what the criterions are and the residents uh put their you know top few choices in and in the in how we do the match the teaching hospitals serve as the boys in this in this model and they go out and they just pick the residents they want and if two or more residents are picked by the same hospital and the resident gets the one they like better and that's that's how we do it yeah I was looking at um the Wikipedia article on stable marriage algorithm and that was the first thing that I talked about was that the um graduating medical students matching them to their Hospital appointments that's kind of an interesting an interesting concept but does it infer that um that the match will be effective so the match will be effective but it won't necessarily be ideal in fact there was a court case that was brought by residencies because in that particular way of of coming up with a a match uh whichever side goes first whatever side does the choosing gets on average a better outcome than the other side so uh by letting the medical schools go first they would they would sort of get the better part of their picks Over The Residency so residents sued and said hey this this is unfair we should get better things but the courts decided that you know since the hospitals are hospitals and have a lot more on the line and a lot more at risk it was essentially right and fair that they should have this advantage and so the courts held that they should continue to go first um there there are also sort of psychotic cases that that you can imagine um in sort of a simple world where everyone's preferences are identical so every girl's favorite boy is Adam and every girl's second favorite is Bob and every boy's favorite girl is Adele and second favorite is Betty and so on then Adam and Adele and Bob and Betty and Carol and Carl and so on would all just wind up matched up no matter what happens but you could imagine a sort of psychotic situation where the preferences were perfectly inverted so whichever every boy's favorite girl would be the one who likes him the least and second favorite girl would be the one who likes him the second least and so on um under those conditions every single Arrangement would actually be stable um because since everything's perfectly inverted everyone that you like better than whoever you wind up with likes you less than than whoever they wind up with um however whichever side goes first basically gets a heaven hell relationship so if the boys go first every one of them will get their unique first pick and every girl will wind up with the boy that she finds most detestable but if the girls were to go first the exact opposite would occur where every girl gets her first pick but every boy would prefer any other girl to the one he's actually with uh it reminds me I like I'm a chess player and it kind of reminds me of um white and black in chess you know white always goes first and so white has a teeny tiny like about a half a pawn advantage in every game because they get to decide how the game is opened and so you know it kind of sets the stage for what comes next so that's kind of reminds me of that that's very interesting uh yeah a lot of a lot of games have a a first mover you know Advantage type situation yeah yeah and it and I don't know hadn't ever thought about applying it to a situation like a Matchmaker you know type of thing um that's that's kind of an interesting thing and the theory there is that um because the because of the way the interests are set up that if they make their choice that their chances are that it'll be stable uh well so yeah the idea is and and you know this is where it's a math problem and not the real world uh is that if they're if if you have your preferences and your preferences are stable um if you take your top Choice then if somebody else were to kind of come in and say wait a minute wait a minute what about me well that person's not your top choice because they're not the one that you picked so you're not going to abandon your top choice for a second or third or fourth choice that comes along and says but wait what what about me um and as long as both sides feel that way so if two people are with their third choices there are people that could come in and disrupt that relationship they could come in and say hey I'm your Top Choice get rid of her come along with me but if you're that person's fifth choice and there with their second choice they're not gonna they're not gonna downgrade to come break up your situation so yeah he comes from right that would make a lot of sense so let's talk a little bit about you and your career um it says here that you had training as a in nuclear engineering so so yeah got you into that I I went to college in the 90s uh and didn't really have a particular direction in mind I was in engineering school because I thought I'd have to write less which sadly turned out to be false and I I was sort of following teachers around like I I didn't know what to do and so I was trying out things that were interesting and uh some of the things that were interesting were interesting because the professors were good and so then you know the next year I'd be like well this professor is good and he's teaching that why don't go learn about what that is and see what's going on um so I was taking this this class on uh reactor safety out of you know no particular reason other than I like the professor and uh he introduced this concept uh this physical concept called delayed neutrons and said you know without without this physical thing uh it would have been impossible for human beings to have commercially exploited nuclear power and Engineers particularly larval you know student Engineers react to the word impossible you know rather badly so we all you know threw out some things of like obviously you could do this obviously you could do that well you know you shot us down because we didn't know anything um but I put my head down and tried to design this science fiction nuclear reactor that would actually answer his objections and I came up with this pure you know this this is this is Star Wars technology but like this this gamma ray laser nuclear reactor that would have light speed reaction time for shutdown so that you could actually like have a bomb inside there that you could turn off in time uh and he was like that's that's a pretty wild idea that's the kind of thing that you can write a senior thesis about um why don't I connect you to my friend in the department who who you know runs these things he could be your thesis advisor uh when you know I needed a thesis to graduate and I didn't really have a Direction so I was like sure I'll go talk to him so I go talk to this other guy uh and he's mentoring a PhD candidate who's actually building the world's first computer simulation of this this out of left field nuclear reactor design which if you squint and and you're incredibly generous to me sort of looks like the thing that I came up with in my like flight of fancy um but like but real it actually works it's it's this thing called an energy amplifier um and the idea is that basically there's this reaction you can do that produces an enormous number of neutrons and so you don't have to have a super critical reactor anymore because if you've got a lot of neutrons then the reaction doesn't need to carry out itself you can just use this large fund of neutrons to run through the reactor um so he gave me he gave me an idea taught me a couple of classes uh wound up taking a radiation detection lab from him which was amazing I got to do all kinds of incredible stuff like pull the control rods out of a nuclear reactor and uh do radiation activation chemical detection tests and stuff like that uh and that was it that's so that then he then he gave me a fellowship so I could go to graduate school for a year because they were shutting the department down so nobody really cared and so I got in and took a year of grad school and then needed a job doing you know something that people actually do uh you know I I couldn't I couldn't keep up with my sort of sci-fi mad science college career anymore and uh that's that's when I got into computer tech awesome so you see you worked for some tech startups what what did you do for them uh mostly data analysis uh uh or uh some domain-specific language development uh so I I did some like web log parsing and and managing like the newsletter with like customization for you know we call it the spam Cannon back in those days um and then uh I was working for a company that was doing workflow management for publishing houses uh so I built some dsls to automatically generate some of their sites uh and just just sort of moving around a lot because the companies kept failing because it was 2000 and then 2001 and 2002 and and people with incredibly flawed business models weren't just getting free cash flown in anymore yeah it's funny you say that I I worked I worked in 1999 I worked for a startup that essentially was the precursor to the iPad um they in they had this really ambitious idea that you know and kids go to a restaurant they need something to play with and so they were going to do these um tablets that they were going to put out and as a as a software developer it was really interesting to me because the the hardware Engineers I'd say to them you know I I can't get it to do this because it doesn't tell me when it's completed this cycle and the next day I'd come in and they'd say we updated the firmware on your device try it now it has a code for you like just cool stuff that they would do but yeah the demise of that company was that you know back in that day uh you know 19 to 21 inch monitor was like a couple thousand dollars and every one of the management team there's 10 of them had two on their desk and the guy that was funding all this came in one day and just said my wallet's closed until you guys start spending the money on the product and not on your you know your office toys you know and and that's you know it was a sad it was a weird it was a strange time period there at the the late 90s early 2000s so that's kind of interesting to hear what do you think is the um like the biggest problem you had to solve during your career so far um that's that's sort of difficult to measure in different ways it's probably uh the the project that I'm actually working on right now which I didn't actually have to solve I I sort of wanted to so um uh around 10 years ago now um I had I had finished with the company I was working with uh I basically completely automated the job they hired me for and this company had a policy that uh uh nobody could do career advancement there so so I quit because there was nothing left um and uh and I had cash and I had been very interested in math for quite some time so I did a little bit of traveling mostly I wanted to take some time to think about uh computational systems and uh I was I got around to being curious about finding consensus under networked communication conditions and I found this game theory approach to that problem and then after talking to a friend about you know sort of the cool properties of this I decided to turn that that approach into a Marketplace and that took me about six months uh it's only a few hundred lines of code but there was a there was a lot of work before I started writing the code to get it all like you know lined out and getting the math working and stuff like that uh and that I've never worked on anything that I had to think about for for more than a week before before that one took a lot of planning and thinking yeah yeah but I mean that's it's worth it in the end though I mean when you I I always love that when I when I used to write code I used to love um hearing someone else especially in another member of my programming team telling me Oh you're never going to get that done that in that amount of time well then it became a challenge you know um but you still have to follow all the steps to make sure you you are dotting all the eyes Crossing all the t's you've thought of everything you've mapped it out because um I I get it like when you're you don't want to get down the road and find you know we use a house analogy and find out that you put a wall in a place that you don't want the wall and now in order to move the wall it's like this giant nightmare um to to move the wall you know you have to go through hire an engineer to figure out how to move the ding wall it's better to have your plan up front think it all the way through and then sit down and write the cut especially if it only 100 lines of code that's not a lot of code um but that's also pretty impressive um so of the startups that you worked for um is there any that really stick out in your mind and why um well uh I guess the the first and and then the the last one before my Hiatus uh are both they're both pretty out there the first one um they they were basically one of the pioneers of social gaming it was Merv Griffin's ex-wife and her sister had hooked up with a couple of guys who had wanted to get into like film and television production but then decided that the internet was a cheaper production platform that they'd be able to get more use out of and so they had these two women who were game show machines and these two sort of ambitious kids with money and they were like what if we put game show type systems online where the sorts of people who like to watch game shows I.E married women in their 30s and 40s could not just watch the contestants but play the games themselves and chat with each other and so on um and so they did that and they were they were getting a a pretty decent number for you know the late 90s of of customer base I think they had uh well the the list was you know 50 to 60 000 or so and was growing fairly rapidly uh but they didn't really have a monetization plan um and it turned out that their their grand plan they they were selling ad space but their grand plan was to come up with these prize thing and and they were gonna have these internet coins that were going to be worth money at other locations and stuff it turned out that going to their their prize fund area was associated with leaving the site and never returning um their customer base loved the games love playing the games hated their monetization system and the other thing is it turned out that while they had a a fund of 50 to 100 versions of each game because there's a lot of content in these in these game styles the players actually really liked chatting with other players on these games so the most people showed up and left without really engaging a handful of people would engage for three to five and then there was another group of people that would play hundreds they would play every single game five or six times because what they what they liked doing was just coming on and chatting with other people uh so they overspent on content by a factor of 10. they overspent on their monetization strategy by a factor of infinity because it was a complete waste of time and money uh and they folded because they'd spend a bunch of money and their plan for where to get more never materialized um in the in sort of the last company I was working for the one where you know I sort of finished it was actually a company that made and leased slot machines to Indian casinos and they had just gotten a truly massive order and they were on their way to effectively dominating the Oklahoma Indian casino Market uh I was somewhere between the 70th and 90th higher at the company and I can't be more specific than that because that they probably hired 10 people the day I was hired sure when I left five-ish years later uh they were a little over 500 people at that point they they grew like a weed um but pretty shortly after uh about a year after I was there their primary sort of development plan actually got fractured and over the course of the following four years they effectively were struggling with the fact that they had two separate incompatible code bases that they needed to reunify and they never did and they in fact never have um so the the last game they developed was was actually developed a little bit before I left uh at which point things had sort of fractured so much they couldn't move forward anymore uh and a lot of that was down to the growth rate where they'd gotten so big so fast that they were just throwing money at things and creating problems that the management team wasn't aware of and so kind of the the reason these kind of stick out for me in particular is kind of the same reason in both cases the first one was a failure and it it collapsed and eventually got sort of sold off for scrap the second one could easily be called a success the the guy eventually sold that company for in excess of a billion dollars um but there was a great deal of potential that was essentially left on on the table and in both cases management was making decisions which the computers that they owned had information that contradicted the sense of those decisions um and so one of my sort of motivations was that in organizations particularly once you've sort of gotten out of all living in the same room and writing the same code on the same computer uh most of the information is outside of the executive's head and that leads to Executives making incredibly bad decisions because they just don't know what's going on in their own company and so if if an approach to improve information consensus and awareness uh exists I think that's a very interesting thing to have access to we I always teach my clients that as Tech leaders sometimes they need to be the one that puts the brakes on on things uh it sounds like the first business had a plan up to a point in it in that part of the plan worked really well the game part and I think they underestimated what their um their target audience was going to be really interested in because as you said that they weren't interested in the monetization strategy at all and I mean it reminds me of in the in the late 90s there was a game called um Hollywood stock market and basically you'd buy and sell movie actors and actresses and it's kind of like I should say it's like fantasy sports but for actors and actresses and they did a similar thing launched the game get people interested in the game and then they started with their monetization strategy and the same kind of concept nobody liked it and the game just died because of it because they're not you know when you're forced into that pathway and you don't like it you don't do it then your second example it just sounds like perhaps they scaled a bit faster a lot faster than they expected and like in a situation like that what I would advise people is you know take a take a breather you know take a step back look at look at what your plans were and see how this rapid growth can be harnessed so that you can still control where you're headed um but I mean if they sold it for over a billion dollars though and they yeah it's a pretty dang good success it was yeah it was a a very interesting conflation of circumstances um about I think a month and a half before I was hired uh at which point I think the company was less than half the size that it was when when I was hired um the the chief operations officer and the CEO uh had gone into a meeting with I think the second biggest tribal group in Oklahoma that had decided that they didn't like IGT which is the number one slot machine manufacturer on Earth uh those guys invented video poker basically um and so they go to the meeting and igt's machines are piled up in the parking lot so yeah it's you know you're a car salesman and somebody walks on the lot like it's it's it's on so they went in and they got contracts for 40 of the space um and and they had one game um they'd been sort of poking along just kind of barely eeking something out and then they took that contract and they went to the next people and they were like well the number two people that's this is what their floor is going to look like now if you want to compete your floor needs to look like this too and they just sold they sold 35 to 40 percent of Oklahoma uh before they'd actually gotten around to sort of having uh an actual production chain in place and and it it they got they got as far as they could get and then they couldn't they couldn't get any more they couldn't get outside that bubble um so you know it's it's an enormous thing but on the other hand IGT that invented video poker um that was that was sort of a single good idea and they went from nothing to 80 of an industry that has revenues in the hundred-ish billion dollars a year which so so you know these guys got you know these guys got less than one tenth of one percent of market share uh while they you know they became very big fish in a very small pond effectively that's really interesting though because your description of them being piled up that's just funny um so as we're getting kind of short on time I wanted to ask your opinion you know right now in today's world um artificial intelligence is really like at the front you know people talk about it a lot what what do you think about artificial intelligence like some of these softwares that write for you Etc what are your thoughts on that uh I think AI is is reached a it's it's sort of most exciting Peak uh and I think as a tool um it has a lot more advantages than than we sort of traditionally used it in the technical space I think it's disadvantages are still pretty poorly understood uh and in general learning algorithms that are also black boxes are something I'm very uncomfortable with um because the flaws which have to exist in a black box system are things that you can't gain access to except through enormous experience um however uh from a kind of smoothing the edges perspective um and also a solving closed systems perspective uh things like the alphago you know Leela chess projects uh and the folding the the alpha folding project demonstrate that there are very important questions that are actually based on you know physics which which is at the end of the day not such a big system that we really can throw a lot of computer power at and get some very valuable insights out of um my system is in fact in one sense a an artificial intelligence system it's a learning approach uh however I'm very consciously because of the game theory aspect integrating not just machine perspectives but human perspectives as well uh and I think that creating super Minds um is something that's both common I think markets have been around for centuries and that they functionally operate as supermines so I think that artificial super intelligence if you will or sort of assisted super intelligence is something that is is common that we've actually built our societies around at this point um and something that we can do better with using the technology that we have uh and so I think it's something that you should be thinking about how to pursue because if you're in a competitive industry and your competition figures out how to do it you know if if there's a bunch of farmers and one of them's got a John Deere and the rest of them are standing behind an ox pretty soon the guy with the John Deere is going to own all the farms uh and and that's that's kind of where we are right now is that while there's there's some skull sweat required uh if we can the technology is there to be able to make things that work better than than what we have and so that means that if you're not doing it uh then then your lunch is likely to be about to be eaten by somebody who's going to figure it out you mentioned Alpha zero and I mentioned earlier that I'm a chess player and like that entire the way they taught that computer software to play chess it's ingenious first of all um but it had a huge it's already had a huge impact on the game itself I mean there's strategies that would have never been employed before that are now um but the I think yeah I think it's actually quite fascinating that so computers that are already much much stronger than people at chess but because the way that they were stronger than people at chess was so foreign to the way that we're actually even capable of playing chess that while people were getting better we weren't really learning much from the computers about Chess whereas the the alpha Lila you know those types of things are playing chess not just at a much much higher level and in far better ways but in ways that human experts can sort of watch and extract principles from and so so people have learned these new principles about Chess that human beings had never figured out before and and yeah I think that's that's a well-made point well I mean like even the you know Magnus Carlson is the current world champion and um he won't be for much longer because he decided not to defend his title I personally think he's bored um well he he literally told them that unless this one particular young Grandmaster was his opponent that he wasn't interested in defending his title which kind of tells me that he thinks that that's his only real competition but he's hmm rightfully so he's kind of cocky but at the same time he's humble uh which is kind of a weird combination so he'll never come out and say oh the rest of you guys are in competition for me but it's kind of inferred but what I'm getting at here is he even he's employed like um they called upon all the way on the right side the H Pawn and he's employed like using that pawn to attack the opponent's king exactly like Alpha zero does um yeah yeah you know and it's changed the game right yeah because it really does work better and once you watch the computer destroy other computers that are vastly better than you dozens of times you start noticing hey this thing happens over and over again and it keeps working and because the computers trained to make moves based on probability of Victory it does build up these patterns of things where it's doing it's making the move that looks like it's going to lead it towards Victory and so it it has its own special patterns of you know maybe I don't need to Castle as often because I can just win because I've got an extra move or the H Pawn or or other patterns like the I'm not I'm not much of a chess expert but I've I follow this from sort of the opposite angle of being interested in in these techniques that are being used to have the computers Now teach themselves how to be even better one on the in on the the chess in the Chess World the AI is being used to help teach other people how to play like there's a there's a startup right now called aim chess aim chess and you can go into aim chess get an account um It's Perfectly free at the moment of course they'll find a way to monetize it which I hope they do um It's Perfectly free and you give it you basically connect any of the online chess playing locations you connect your account and then it as you play games on those places it's pulling your game data in and then it's presenting you with lessons around your own games in your own positions I mean I've used it and I'll say to myself how does this position look so familiar and then I realize that looks familiar because I just played that yesterday um and it pulls it and it teaches you where you've made mistakes and it has like multiple different drills and that's where it's going they're trying to take Ai and have it teach humans how to play better that's that's the obvious end game there what I wonder is like with things like chat GPT and like Jasper um they're they're they have to be gaining a lot of data and knowledge about what people are interested in just by the questions that are asked what the inputs that are given to it and I'm wondering like what's the end game there what are they going to do with all of this massive data that they compile about people's requests to these you know AI softwares and I mean like chat GPT has gone so crazy that if you try to go to their site and use it you're you're lucky right now if you can because they they didn't expect it to be so popular right away so they didn't build in good scaling for it but I mean it's what I'm finding interesting is that people for instance that people use a lot of AI tools for writing content for their website and they're saying that Google is working on a way to identify that and then penalize them and I say to myself I don't understand why you'd want to penalize them because you know let's say I'm a copywriter and you're building a website so you hire me to write your copy you didn't write it I did so why are why it's no different than if I let the AI write it for me well another part and parcel of that is that that that process will actually create a a darwinian reinforcement thing to improve the algorithms themselves um so uh another AI inspired thing is the Deep fake uh which has some impressive things but uh you know occasionally really is pretty ugly and and there's some famous cases uh was it the uh the Luke Skywalker cameo in uh in whatever it was where some random person on the internet was able to actually make it look like Mark Hamill when the actual Pros doing the thing at the time couldn't um but one of the one of the things that that happened in that Community was people built detectors you know like I'm going to I'm gonna I'm gonna build something that can look at this thing and find out whether or not something bad is happening here you know is there a fake in there well then once you've got a trained detector you've got the exact tool you need to train a better version you know can you make a fake that gets past the detector well it'll look better if you can so then you can train a new generation of better fakes um I uh with my sort of Journey in the financial world I was running I ran into a woman um who was in the reg Tech space and she had worked with a startup that had uh trained up an AI to detect money laundering behavior and so they had taken Market places like stock markets where known money laundering had happened historically um and then sort of other random days where that wasn't going on and and done some you know uh distinction training and so then it was able to to say okay this you know here this day looks suspicious check out these sorts of these sorts of you know players and then go after money laundering uh however they had a stumbling block they uh they went after commodity markets um to to do their thing and their algorithm was incapable of distinguishing uh money laundering from non-money laundering uh days and the the basically the two possibilities they came up with um are one because the interest of commodity markets are so different from the interests of of stock and bond type markets that it was just too difficult to actually find these types of things or two um because money laundering uh and manipulation is so trivial and Easy in commodity markets it might be so common that there's no such thing as a clean day to actually train again right right so we can't figure out the patterns that it's supposed to be figuring out because they're kind of muddied up yeah so like you're trying to you're trying to train a facial recognition algorithm to like tell the difference between uh you know uh some some movie star and other people but you train it where every face it ever sees is the movie star and it can't tell the difference between between Tom Cruise and anybody else because everyone it's ever seen is Tom Cruise well you know what this has been a great conversation I've really enjoyed it but why don't you tell my listeners like where they could find you if they want to reach out and and get more information about what you're up to uh yeah absolutely well the easiest way to reach out to me is my email address uh Noah P Healey yahoo.com um you can find me on LinkedIn I'm the Noah Healy there and I have a website at corddisc.com that's c-o-o-r-d-i-s-c uh where you can find out more about my Marketplace ideas all right well we'll make sure that gets into the description for the video and in the show notes for the audio that way people can find you easier and they don't have to rapidly grab a pen and try to write it all down that's always not never fun but but again thank you so much for being on um this was our first episode and I couldn't have thought of a more interesting way to kick it off actually it kind of worked out nicely I think you're uh some of the Insight that you've had has kind of blown my mind a bit so really appreciate it so this has been a the first episode of gaining the technology leadership Edge and um this show is going to come out every week um and um just come on back and and let's um learn some things from the smart people that are going to be on the show foreign
2023-02-08