Prepare yourself for a sweet conversation as we welcome Dr. Peter Angeline, Strategic Futurist at the Hershey Company, to give us a little taste of the key technology and innovation initiatives at the chocolate giant.
Peter monitors, researches, develops and incubates current emerging technologies to assess their potential impact on and application to Hershey's business. He also manages Hershey's Digital Innovation Lab at the University of Pennsylvania's Pennovation center.
Hershey has investigated Computer Assisted Innovation, Virtual/Mixed/Augmented Reality, Chatbots, Machine Learning, mobile consumer engagement, Smart Automation, IOT, gamification, nanotechnology, consumer engagement through robotics, Future of Retail, and bleeding-edge technologies.
Transcription:
You've come to the right man on the can the man.
So you're probably wondering why we're playing Sammy Davis is Candyland. But today, we have a guest who comes to us from the Hershey Company. And we had a lot of fun before we warmed up, warming up before we came into the show, talking about all things candy, but it's more than that. So we're pretty excited on our guest today, but I want everyone to understand today's gonna be a little different, a little different is because we are wrapping up four days, four half days of action packed, intense technical conference on big data. So it's just been amazing. I want to thank Monique attack us on our team, who has led that with the support of Taylor, who always has PTC events on her screen. Today, she has it next to her name, but without them, we couldn't have been possible, we had a ton of sponsors with a ton of engagement, really a lot of amazing conversation around AI and machine learning and big data, and you name it, some of much of it was recorded, there'll be opportunities for people to catch up with it. And we're pretty excited. So what's different about today, today, we're gonna go a little longer, because we are gonna have we have a lot to talk about. We have a lot to hear from in terms of our guest, Peter Angeline. And we want to make sure that we sort of wrap up our week with big data sort of relishing in all the things that we've been spending time covering. So of course, I want to give thanks to Huntington bank. They've been partners with us right from the beginning of business as usual. Today, of course, we're blending it. And we're you know, we're joined by many of the participants who have actually been in big data this week as guests or as speakers. So the summit, let me see the summit is tomorrow. Monica, I thought today was the last day, today's the last day. Okay, I'm so excited the days blur. And, you know, I thought it was Friday as well. So anyway, we had amazing sponsors, and I want to give them shout outs in addition to Huntington bank for their, you know, support in doing all this work with business as usual. We had sponsors all week and those sponsors allowed us to make it accessible, make all the information that was being provided all the people that Monica pulled from all around the world, it wouldn't have been possible without our presenting sponsor CGI, they're long term partners of ours, but they have huge presence in our region, as well as in Columbus. And at the end of the show, you'll get a chance to hear from one of their leaders. And that's just going to talk a little bit about their company with keynote sponsor called risk focus. And panel sponsors, which included click, and the Consulate General of Canada. and supporting sponsors included cloud era to QE communications acnc jet store, plus consulting and snowflake, really important for all of you who are on the call join us regularly on ba you. These are these companies that I just talked about that our sponsors are actually an integral part of the technology ecosystem. We can't do much without them and their support and their services. So now I have the opportunity to bring forth. That's why we play Candyland to bring forth Dr. Pete. It's not Peter, Dr. Pete Angeline. And I'm going to talk to him a little bit and then he's going to give a presentation. And we'll have plenty of time for q&a. We're gonna make it as interactive as possible. And he is one cool guy. He has his PhD from Carnegie from you from Ohio State. And from Carnegie Mellon. He's been he's actually a native of he's calls it Latrobe. I never say it the correct way. But he has lots of love for this region knows a lot about Southwestern Pennsylvania and I am just thrilled to bring him to the program. So, Pete, Dr. Candyland How are you? How you doing?
I'm fabulous.
It's great to have you on the show. So I did a little bit of intro. Like who is Pete? I'm not sure I captured everything Of course. And all fairness and the shortness of time. But you are a native of southwestern Pennsylvania even though you've, you know, gone a little bit outside of your region. You seem to be pretty tied back here.
I am I am. My parents were from Darien blairsville. I was born in Latrobe. I went to Greensburg Salem, elementary, middle school and high school Central Catholic for a year or two and ended up at Carnegie Mellon for my undergraduate before moving out into the world and, and going to Ohio State and living in a number of states as my career develops. So
it's great. So here you are. Now you're located in you living in Hershey.
I'm in Lancaster, I'm a little south of Hershey about 20 miles. But it's it's a beautiful area Harrisburg is not far in good proximity to both Pittsburgh and Philadelphia. It's a fun place. That's
great. So we are going to so how can held up? Before we get into your presentation? How have you held up during COVID? Tell us just like a little bit about, you know, Hershey and COVID.
Yeah, it was really interesting. And just as a personal note, excuse me, I was away on my significant other I she went, she and I were in St. Thomas, when everything started shutting down. And I was really hoping that St. Thomas was going to shut down and Hershey was gonna have to put me up in there for however long the pandemic was, but it didn't happen. So we came back and immediately upon coming back, her she instilled very strict regulations on who was allowed to be in work, who were the frontline workers, how they were going to keep their people safe. And I was tasked as one of the people to be on the data analytics team trying to understand the data around COVID. And how Hershey can best maintain their ability to serve the consumers, but keep their people, their salespeople, the frontline workers safe. And it was an amazing synergy of everybody bringing what they could bring to the effort, as the first few months went through. And everybody was able to kind of sort out how Hershey should play in that evolving and dynamic environment. The reality is, though, that in times of trouble, like COVID, people love comfort, even simple comforts, like chocolate, and we had an amazing year. So not only were we able to provide some comfort to people, but we were able to do it in a way that they were maybe hopefully able to help them get through the times as we've gone through so far. And hopefully we want to be in this much longer. And we can all get back to whatever normal is going to mean from here. Right.
That's great. That's pretty Yeah, I mean, I guess that's true chocolate and not alcohol have made. So yeah,
we'll talk about the alcohol chocolate definitely.
Dang, I tried to trip you up there. You've been you've been safe and sound. So what are some of the what what books tell us a little bit about Hershey. There's there's just a little bit of a story there. I mean, it's a massive story. I don't want to I don't want to shrift it. But yeah, a little bit about the Hershey story.
Well, and I don't want to I mean, it's a it's a big story. I mean, Hershey was was a major player in the US commerce. Very interesting life. Both philanthropically as well as in his, in his commercial success. But very briefly, one of the things that a lot of people don't know is that he left behind something called the Hershey trust, which is a holding company. And under that there are three companies. One is Hershey foods, which is the company I work for, and the one that everybody's familiar with. Another is the philanthropic part arm, which is a private company, but they're responsible for all of the philanthropy that Hershey as the trust does, as well as the Hershey School, which many may know about, which is a school for very promising underprivileged children. And then there's also the Hershey entertainment section, which owns the Hershey Park, and a number of different sites, hotels and things around the globe. Hershey himself, you know, he was an interesting guy, he had a couple of companies and he went to a World's Fair in Germany 18 something and came back and while he was there, he saw chocolate and was amazed by it and thought this is going to be something that's important, came back, sold his Carmel company and opened up Hershey chocolate out in Derry township in the middle of pa what reason he went there was because that's where the cows were, and he knew he needed milk.
gotta follow the cows. So we are gone. So how have you been holding up during COVID
I've been great. You know, I'm you can't see me because I strategically have a false background behind me but I'm sitting in the corner of my dining room. And that's actually a new one because I moved here in July. So there's been that interesting, purchasing and getting a new place without really being able to see it because nobody was giving tours. But we're doing fabulous. And I actually just got out for my first weekend in a year went to see my son in Philadelphia. So it's, it's great. We've been very careful. But, you know, now's the time to take strategic risks and try to bring life back up again.
That's great. I've been we've been fine. I'm fine. My team is fine. I've been vaccinated, and I'm ready to hopefully see 70% of the people vaccinated so that we can get back to or just get to something.
start figuring out what normal is going to be.
Yeah, I wouldn't say get back but get to something. Thank you for asking. So let's, we're going now I'm going to, I'm going to tell everyone, here's a little presentation for us. That doesn't mean that we can't ask questions. I'll be monitoring the chat. Jonathan. Christine is here with us, as always vice president of all things media, and we're going to we've muted everyone, this is an opportunity for for you to just focus on our guests. This is not about selling your wares. This is just focusing on our on our guests. So I want to thank you, Pete, you got a little bit of entertaining that you're going to share with us. And if I have some questions that we'll we'll jump in,
please do. And I actually, I think these things go better as a discussion. So if if there is something that I don't cover, or that seems unclear, please feel free to jump in, I am going to give this talk is one, and I'm sorry, I've got screen setup. So if I'm looking away, it's just because I'm trying to make sure everything all the windows are up in the right way. Thank you. Um, this is a talk that I've adapted from a TEDx talk I gave for Pfizer lab middle of last year. And it's it's a developing concept that I've been working on for a few years. And that I think, is starting to catch on more broadly. And it's about this concept called augmented intelligence, which I personally think is going to completely revolutionize the way all businesses operate in the next decade or so maybe even before that. And to maybe make it a little more concrete. What I really think is, is that this concept of augmented intelligence is going to be just as strategic for business as Microsoft offices. So I'm making no small claim here, I really think that the concepts I want to kind of paint for you today is something that's going to be material to everybody's job, or many people's jobs in the coming. Wow,
that's why they call you the Patriots.
Well, you know, someone's got to do it. And it's one of those fun things, right? I mean, that's what they pay me to do. They tell me go sit in a room, look at all these cool things that are happening and tell us what you think is gonna happen. 10 years
weren't happy. All right, it's all yours.
So let me find the right thing to click. There we go. So the name of the talk is data driven innovation, using augmented intelligence. So let's start off by just thinking a little bit about what innovation is. Innovation as the discoverer of vitamin C's quipped, is it seeing what everybody has seen, but thinking, what nobody has thought, related to that, right. And that's something that has been very natural through the years, you know, depending on what you've been exposed to, and the things that you're doing in your daily life, you see the certain things that everybody else has seen, but every now and then you get that spark, that idea of something that takes you in a different direction. And then that opens up a whole new door of inquiry or thought, or discussion or a new product that could make people more interested, or more engaged with your company. And these sorts of things. This innovation requires very uniquely human attributes like intuition, and analogy, and inference, and it's really important, but we've got to realize that the world is changing around us. And the reality is, is that the sources of inspiration are coming at us in a very different way now than they used to Data, Data, Data, Data data to quote, The Brady Bunch. It's the collection of things coming out through the phones through all of the things people collect online. It's whole sources of new information of new sources of inspiration that we haven't had before. And there's magnitudes of it. And I want to give you a sense, and I'm not sure this is exactly the the an audience who may not have seen this, but in case you haven't, let me give you a sense of how much data is actually flowing through. And I want to do this in terms of rice grains. So let's start off first by saying one bite one character that we would have four data one character and a z. An h for Hershey. One character is one piece of Greek is one grain of rice. All right. So if I multiply that by 1000, and I have 1000 grams, of rice, that's about enough to fill a bowl. Okay, so we call that a kilobyte. So that's a kilobyte a rice is roughly equivalent to a bowl of rice. Now, if I get to eight a megabyte, which is 1000 kilobytes, now I have about eight bags of rice, roughly, that's about a megabyte of rice. If I go up 1001 more time, then I get a gigabyte of rice and a gigabyte of rice is enough to fill three trucks. Now. Now we're in a place where it should be familiar, your phone probably has a couple of gigabytes of memory on it of data. Right? So one gigabyte of data on your phone is three tractor trailers. So if you have 32 gigabytes, right, that's, that's 96 truckloads of rice of data that you have on your phone. Let's keep going up a terabyte 1000 gigabytes will feel to containerships with grains of rice, let's keep going. a petabyte, which is 1000 terabytes covers the entire island of Manhattan, in rice, one layer thick. So we're still going up, let's go up one more an exabyte, is 1000 petabytes and it covers the entire West Coast of the United States. And you probably see where I'm going. The next one is zettabyte. And zettabyte. is enough rice grains to fill the entire Pacific Ocean. Now, why am I going through all this? Because we do somewhere around nine zettabytes of data every year right now and it only is growing. We're talking about Pacific oceans worth of data at our disposal. So what's in this data, it's some of the stuff that I showed you, it's got some some information about how our products are doing, it's got the information about how people are using our products, it's got information about all kinds of stuff. And I think I can just get that information out of that pacific ocean of data, then I'd be able to do a much better job serving consumers. But that's not an easy task. And the reality of it is that I can't do that myself.
There's something that the economist Herbert Simon coined called bounded rationality. And what it basically says is that people are inherently limited by a few factors in how they can be rational, they have a limit based on their specific brain anatomy, my wetware in my head can only allow me to take in so many facts at a time, my personal experience is going to cloud the way that I choose. And the way I'm biased. The the my knowledge that I had due to the fact that I went to this school and somebody went to a different school, that difference in knowledge may make us do different things under similar conditions. So what it basically says is that I can't know everything, and therefore my ability to be rational, is limited. So if I've gotten the Pacific oceans worth of data, I can't get all of that data in my head and leverage it and use it in the way that's optimal. I'm bounded initially, from the start by my ability to be rational, to be rational on that huge amount of data. Right? So what am I going to do about this? Because all of that data, there's there's gold hidden inside that data, we used to call this, we used to have a practice I did when I was a consultant called data mining, because you go through the data, and you mined out the things that are valuable out of it, how do I get that value out of it? Well, this is where AI comes in. And this is the where we're going to start with the augmented intelligence definition. Now I realized that AI can conjure up all kinds of scary scenarios of robotic doom and human subjugation and everything else from all the wonderful media and and books that people are writing both. Suppose that AI specialists and non and I don't want you to worry about that. And part of my hope is that some of the things I'm going to show you briefly here will convince you that maybe some of that's a little, not as accurate as it could be, let's say, to be nice. So, before we get into augmented intelligence, I want to change the conversation just slightly, because I think this will help us understand the way AI works in its place. And what I'm talking about with augmented intelligence little a little better. I don't want us to think about that pacific ocean of data as bytes anymore. Think about it as individual physics experiments, right? Think about there being some little sensor somewhere out in the world. And every second it flashes and takes in whatever information is there and puts it puts it somewhere and says that was an experiment. So now I have zettabytes of experiments out there. And if if I'm looking at it from a physics standpoint, what a physicists do, they don't see all of this data, and then they generalize it into a law that covers a whole bunch of that data, but in a very simple way, in a very mathematical sometimes But sometimes in a way that just summarizes maybe zettabytes of information, but in something that's at a higher level that we, as humans can better manipulate, to do the things we want to do based on those regularities. So what we want to do is, we know we can't, because our bounded rationality get down into that data, we've got to find a way to abstract it up into physical laws or data laws, so that we can operate at a level that makes sense for us. And that's where AI comes in, specifically, the machine learning and data science side. But we're gonna let them do that work for us. Because it's hard and are bounded rationality is not going to let us do it, we need something that's going to be able to do it for us. But there's a catch.
Ai, the only thing that AI can do is work with the data it has, if it doesn't have the data, it can't do anything, it doesn't, there's no magic here, it doesn't magically pull in some new idea that it doesn't have access to it derives laws from the data it has. And so the problem is, is that AI sits inside its own data bubble, if I don't give it the data, the AI can't do anything about a topic, if I'm giving you stuff on chocolate, it's not gonna be able to tell me anything about gum and mint. So if it's not, if I don't have data about gum and mint, it's not going to be able to tell me anything about it. Does this look familiar? Yeah, AI is also bounded, rational, rationally bounded. It has bounded rationality, just like we do in a different way. And this is a bit of a clue to where I'm going to go with his augmented intelligence. But first, and I think this is the most important point I want to make is this idea that AI is inherently limited in the things it can do. And I'm gonna give you a couple of examples. So that you have a good feeling about why this is and what that form it takes, and hopefully, maybe settle some of your concerns about robots coming to take over your life. So I'm going to give you two examples. The first was an AI researcher who decided they wanted to create an AI system that would identify animals from an image. And so they took a bunch of images of animals, they trained up what was called a neural network with this. And then they showed it an image of this image of a panda that was one that hadn't seen before. And the neural network said, Oh, yeah, this is a panda. And I'm 60% confident that that's what it is not bad. I mean, you know, it's something that four year old can do panda. But the reality is, is that to get an AI system, a computer to do it, without actually explicitly programming everything in, it's kind of impressive. But then the devious researcher did something to kind of shake it up, it took some noise, the researcher took some noise, and added it to the image. And when we look at that image, we say You know what? That image looks like a panda. But when that same image just modified by the noise was given back to the AI system, the AI said, Oh, this is a Gibbon. And boy, am I sure that this is a Gibbon. And you know what happens when you confuse and a panda with a Gibbon, you get a really ticked off given. So the reality here is that the AI built for this system is showing its bounded rationality, it had bounded past panda rationality, because the set of images that was that were used to train it weren't sufficient to separate pandas and Gibbons in all photographs, it probably found some tricks in order to do that. And it's those tricks we need to be very wary of, and very aware of as we look at AI and the way it's going to impact us. So I'm going to show you one more example. And I'm hopeful that this one will maybe scare you a little bit. There was some research done in by some ai, ai folks looking at researchers looking at trying to identify who was a criminal and who wasn't just from shots, photos of the faces. So they took two sets of data, they had samples of people who had mug shots, and they put those in as one set. And then they took off the socials shown social media, pictures of folks who were more than likely not criminals. And then they trained up the a system and the AI system did fabulous on saying this person was a criminal, and this person wasn't a criminal. Except it was wrong. In fact, it cheated. Now, it's hard to see. But there is something in these photographs that will allow anybody to look at this. and separate the two sets has nothing to do with whether or not the person was a criminal. Can you see it? Let me help you out. While it's not completely clear, that the people in the bottom are smiling, there is enough of a difference between the mouths of the people who are getting a mugshot taken which let's face it, they're going to jail. They're not happy people. There was enough of a frown for for that to be different enough. Then the class of folks who were not having their mug shots. So did this really tell a criminal from a non criminal? Absolutely not? What if they used it that way? What if they concluded that everyone who found was a criminal? What kind of problems would the people who were maybe just unhappy that they have been in? Because of that? And I don't know about you, I'm a little bit of a science fiction fan. But if you remember, a minority report, and the whole idea about precog, this looks a lot like precog. This, we're trained to say you're going to eventually be a criminal. And where does that lead us then if we have something like that? So hopefully,
you know, the other thing is that all the people on your bottom that were smiling, all have REM, you can see that they're wearing suits.
Yeah. Yeah. And this and I agree, that could be another thing that the network picked up on. In this case, when they did the analysis, this is what they found that separated the sets there probably other little cues like that as well. But you're exactly right, you never know what they're going to be looking what these systems are going to pull out as being relevant. It's not always, or even most of the time, what we would consider relevant and making the decision. So we have to be very vigilant about what these are doing. And this is why there's so much concern about bias and data sets, and, you know, racial inequality based on the way these algorithms are working, etc, etc. And it's an important topic. And I can't stress that enough.
Okay,
so So now, we've seen that there are two types of intelligences that I'm thrown out here. One is human intelligence. But we're bounded, you know, we have bounded rationality, there's only so much we can do, we need these high level descriptions, so that we can have use our mental agility in our mental skills, the right way to solve issues, and do innovation. And then we have artificial intelligence, which has this ability to take a bunch of really molecular data and build it up into those rules that work well for us. But it has its own bounded rationality, it can only see things that are within the data that's prevented. And if it's not in there, it can't help you find the pattern, plain and simple. Now, there is an overlap between the two. And some people who are cognitive scientists, or pure AI people, that's where they like to play. I'm interested in the parts that don't overlap, I want to know how humans work and how artificial intelligence works, and what they can do for each other. And that's what I call augmented intelligence, I want to take the pluses of both and find a way of gluing them together in a way that it makes. We have artificial intelligence that's better in every sense of the word. And we can perform our tasks that we do better in every sense of the word. Now, what does that mean? It means that I'm saying AI in the end has to be a tool on our tool belt that we pull off at the right time, and it does, whatever that task is, and that's well suited to its bounded rationality. If we know when to use that tool, and where it's applicable, then we can use it effectively. Now, I want to be clear, I'm not saying we need to augment our intelligence by putting a chip in our head, I'm really talking about adding a new tool to our toolset so we can do our jobs a little bit better. So I always get people hungry when I put this one up. But I blathered on long enough about AI and computers, let's talk about some chocolate because, hey, we play Candy Man, before we came in, so might as well tease you with a little bit of what augmented intelligence means to chocolate. And I'm sorry, you know, if I were there, and we were all together, I'd be handing out kisses or something along those lines, you'll have to invite me back after the pandemic and promise to bring to two helpings for everyone. So what I want to talk about is some work that we did over the last couple of years. And I want to set up the problem first, and kind of explain the opacity of this slide and maybe make it a little more understandable. What if we had Hershey could understand the way people perceive things when they taste them. Let me let me say why that's important. We design products, we typically do it with a specific collection of consumers in mind, generally, we're looking at broad based products, we're going to put out something that a lot of people are going to enjoy, they're going to get moments of happiness, and, and it'll be something that they'll enjoy at the times when they when they need to have that kind of lift. The reality is, is that I can get the same product at five different people. And I'll get five different responses in terms of how they like it. And the reason is, is because even though it has the same flavor, all of these different instances of the product, people perceive those flavors very differently. So the consequence of that is that somebody who is let's say a chubby white, balding male, 50 year old Pittsburgh boy isn't necessarily the best model of Successful Hershey product knew what we have to do is we have to, we want to be able to understand how people taste, you know what differences people taste? Can we accommodate that and come up with products that have a broad a high broad appeal for everybody, but yet still carry that Hershey essence and culture and things that people expect from it. So the idea is this, we want to be able to take a product and be able to predict using AI, how much a particular group of people are going to like that product. Okay, so I'm going to tell you about a very briefly about some of the work we've done that. And let me just say, the reason we would want to do this is it allows us to innovate a new product that will allow, that'll be either we can either target to a specific demographic, or we can target to a broad collection of demographics, and live for another day, so we can make more chocolate and make more people happy.
So the way we did it is as follows. And this is a little bit of an opaque slide, I'm going to take my time, and please feel free to ask questions, if I don't quite get there. We found a partner called analytic flavor systems who had a huge collection of tasting data. This data was over multiple products over multiple demographics over multiple countries, 10s of 1000s of different tastings of different products all over the place. It was basically a data analysts data scientists dream, I've got all this data to play with, what can I do? Well, what we started off as we'd went and did data science, and we dug into this data and we said, Is there a way that if I know that a bald, chubby white Pittsburg boy tastes something and this is what he perceives? Can I predict from that how much a middle aged Japanese woman have to living in Tokyo will perceive it. And it turns out, you can it's not easy, but you can. So we started off by building all of those associations so that I could take one tasting, and be able to say something about the likeability of that product from all other demographic perspectives related to that. And that's what those graphs at the bottom are showing. that's a that's a representation of taste. So I'm in what I'm saying, sugar is higher, bitter, is lower, etc. So that's just our representation of taste. So taking all of that we can pull that all together. And now what do we have after we have all of these models? Well, we can simulate flavor preference. If I come up with one of these graphs, and I hand it to the simulator, it'll tell me for each demographic that I'm interested in how much that demographic is predicted to like that flavor. And that, again, is related to how they perceive the flavor, as opposed to excuse me any other factors that might be coming down, down the pike. I'll mention one in a minute. And so I would be able to take something like say, I want to create a new product for young urban Hispanic non smokers. Okay. And then I could go through, and I could say, here's a here is a product, here's a regular Hershey Kiss, how much does a young urban Hispanic nonsmoker like this, and it should tell me, it will tell me, it's their preference for that they're liking of that compared to other products. So now I've gotten a way of actually doing some work here. If I add to that, a search engine, meaning I have the ability to then try different flavors over time, and based on what the simulator tells me get the better and better flavors. Now I can play this game. So first, I start with the search engine, and it says, Okay, here's a random flavor, I throw it down to the simulator, the simulator says, Okay, this random flavor is gonna be liked this much. And then the that information goes back up to the search in search engine says, Okay, I'm going to keep iterating on this going round and round and round, until I get to one that maximizes the liking from the candidate consumers that I'm interested in. So once again, we started in the simulator, or sorry, we started in the search engine, a candidate flavor comes down. The simulator, ranks it and says this is how much it's going to be liked by this demographic that goes back up to the search engine, the search engine then decides which new flavor probably related to the last one it's going to look at next. And we lather rinse repeat around this cycle, until at the very end of flavor that's well liked by the demographics we set up pops out. Now, the question is, where was the hard work in this? Who did the heavy lifting? Was it the AI I mean, think about it. With the AI correlated all the amounts of flavor data, we had it search through millions and millions of possible flavor combinations, trying to narrow down to a couple of things that maybe our tiny brains could figure out. It went through a number of calculations and went and found pertinent patterns that we were able to look at at a higher level. And it narrowed down the discovery process, it kicked out a few things that we would be able to look at and say, yeah, we could make that or no, that'd be too expensive or something along those lines. On the other hand, we had to ask the right questions, the humans, we had to determine what which of the AI tools made sense how to organize that statistics. That's that simulation in that search in a way that would actually answer the question for us. We designed those algorithms, we implemented those algorithms, and then coming out the AI propose things to us. But we had to then bring other knowledge to bear and say, yeah, that that might be okay to manufacturer, no, people aren't going to really like that, because the AI system doesn't know about this.
So in the end, what we're getting where we're at is that the AI is the crane, it's the tool that allows us to lift things heavier than we would normally be able to lift. But we're both the crane operator and the crane designer. So I put it to you who's really doing the hard work here. So with that, let me just close out and say, This combination of artificial intelligence and human intelligence merged together in a sympathetic way where we leverage the the strength of both and avoid the bounded rationality of both, I think is going to be material to creating new innovative products and new innovative business practices, and in general new innovations of all types in the coming years. In fact, I don't see how it can't be. This type of a tool, as I said, is, I think on par with something like Microsoft Excel, and being able to get the right information to the right person at the right time. So they can make better business decisions, and move things forward in a way that's better for the company, and more enjoyable to them, because they're focused on the things that they care about the questions they care about, and don't have to do all the boring stuff with the data and trying to correlate things that they would otherwise not be enjoyed doing.
Wow, this is great. You know, there, a lot of people have made some statements, but not that, you know, there's one question is like, how have you applied this if you have on the manufacturing line?
Yeah, um, so this was, this was sort of my initial shot across the bow of Hershey saying, this is really important. And I'm going to prove it to you. This was the first thing we did. But in my attempts to take over the world with AI, I'm looking at other places to take this concept. And I'm looking at probably not manufacturing first, but package design, marketing design, you know, how do I communicate? How do I take all the data I have about people, and put it in a way that a marketing person can create an ad that will say something to a consumer in the way that they need to hear it that fits the way they think, and they listen, and they hear? So it really is a question of how do we make sense of this data and put it in the terms of the people who really need to leverage it to do their job? manufacturing eventually, but I think higher ups probably the place to start.
So how does, how do you work across the organization? Do you have your own team?
So I'm in the, yeah, I'm in the applied Digital Innovation Group. And I think Andy hunt, who's one of my teammates is on the call here, we sit within the chief growth officers department in digital transformation. And our job is multifaceted. We have a we have a process where we identify emerging technologies, we look at them and curate them for the company, and say, these ones are important for us to worry about these ones aren't. And then we create and look for use cases in the company. And we go to demonstrations, which what I just showed was one of those demonstrations. And that way we can communicate the importance of these things, then we partner with the businesses to bring those relevant technologies into their field into their part of the business so that they can get return on investment for the pre work that we do. So really, we're bound only we're sort of technology first and then enterprise second. And when we see a technology, we know enough Andy and I know enough and other people in the team know enough to say I need to go talk to supply chain about that, because this is something they're gonna want to know about.
And just speaking real quickly about supply chain. That was one of my questions. And then there are a couple of points out here. How How was you? How are you impacted through any issues on the supply chain side?
We we I'm not gonna say we were lucky because this was work that paid off. years in the making. We were ready for the pandemic, surprisingly so because we certainly didn't conceive of it, much like everyone else. But we had no issues with our supply chain. We were agile enough, people put in many long hours in March, April, May, June, you know, Andy and I being two of those people, but people in supply chain and on the floor of the manufacturing the of the of the plants, they put in long hours. But we had the infrastructure sufficiently agile enough to change things around quickly and be able to continue to keep stuff on the shelves. And a lot of our competitors did not do that they were not able to keep things on the shelves. And, you know, that benefited us. But it was a testament to the forward looking and the adaptability of the supply chain team that they had enough agility built into their system already that they could be able to react in such a quick way and keep Hershey you know, keep customers happy and keep her she happy.
And keep the shelves packed. Exactly. And so and so there's, there's, we have a few minutes. But there's a question here that that is really interesting that Seth asked about a product that was created and rejected for manufacturer and marketing. Can you give an example of that?
So in this particular case, the first thing we tested was, we came up with a flavor that was we dubbed it Mexican hot chocolate kiss. So it was a kiss that had a little bit of a cayenne pepper flavor. And we were specifically looking at both general population and urban Hispanic population, a lot kind of a kiss. And I gotta tell you, we tried that. And this is something I forgot to mention. Once we know the flavor, the challenge then becomes how do I make that flavor? How do I make it from a concept into a physical kiss. And you can never match the flavor in the concept with the formulation that you put out. And we missed the formulation terribly. So it didn't go any further than that. We still have to understand how to take it from concept from this way of doing concept to to a finished product. That's still something we're learning.
And so what about blockchain in any emerging tech?
I have to say I think blockchain is incredibly important, but I think it's still a little soon in some cases. blockchain for sustainability and for transparency, I think is a growing need. We are definitely looking at that. I think other applications of blockchain that I've heard for supporting IoT and other things are maybe a little wrongheaded. So blockchain is very important. But I think we have to be very careful about exactly what applications of blockchain we go to first, you know, the thing that's going to be most important as we move forward is security of data security of plants. We've seen that with all of the hacking on any number of things right now, the thing that scares me the most is somebody driving up next to one of my plants, logging into one of my manufacturing mines, and bringing it down just from the laptop in their car. Security is going to be the most important thing we worry about in blockchain. Still some questions along those lines,
okay. I mean, to the rest of the world, there might not be just by what's happening, right?
certain applications, certain applications,
right. So, before we wrap up, this has just been incredible. I said, What if someone asked a question, What are you reading? Or what's your favorite book? What are you reading?
Well, my favorite book is in sciences of the artificial which is, which is Herbert Simon, right. And
were you trained under her plan?
No, but he was he was a Carnegie Mellon when I was an undergrad. Yeah, I I did sneak over and see a couple of his lectures. And I do I have been reading. I my dissertation was on using genetic programming, which is a method of, of machine learning that harkens back to Darwin Darwinism. So Daniel Dennett book, Darwin's dangerous idea I've been reading lately, and then I, I'm in the middle of a book called calling bullshit, excuse the French. But it's it's a it's basically about how to look at somebody who's presenting data and say, Whoa, wait a second. And it's really quite a good book. It's, it's, it's nice to see that other people are concerned about data visualization, and how you do it correctly, and how people try and use statistics to obfuscate instead of enlightened
Well, you've been amazing. There's a ton of questions, we'll capture them somehow. I mean, what one of the questions just about candy is, the world seems to be migrating towards dark chocolate is Hershey doing the same.
Oh, we have some dark chocolate offerings. There will be more coming up. We have a number of offerings that will be let's say Typical for Hershey coming out in the next few years. And I think I think it's gonna be an interesting collection of of new products coming out in the next two years. I think people are going to be quite pleased and quite curious.
I'm, I'm already salivating. I was salivating right at the beginning. So we're gonna take some of these questions and we're gonna shoot them to you later. Okay. Okay, cuz I want people who are really interested. I want to bring up one person, first of all keys. It's been amazing. Thank you. What a treat. I didn't feel like skinnypop was owned by Hershey. I removed it too. Right. Yeah. Yeah. I mean, there's a lot of acquisitions that I probably missed out on. So this has been recorded. You've been amazing. I can't thank you so much for
joining. And I appreciate the honor and I appreciate to get in front of my fellow Pittsburghers.
Yeah, we're doing some amazing things here in Pittsburgh. Okay, there's, I know, we'll stop going on. So it's glad to have like to pull you in into the work that we're doing. So mnos nice, writes he here. I'd like him to close out for us. And then before we do, Jonathan, what's on for tomorrow? Are with the British Consulate stopping by to talk about the upcoming EU and climate conference. It's happening in Glasgow in November. Great, that's great. We have a great relationship with them. So I'm going to pass the baton over to mnos. And he's gonna close us out. So thank you, man, as
well. Thank you, Adrienne. Pete. What a session again, as I said, if you like chocolate now, but I think the last four days have been fun. I am going to quote another future espied right in another American futurist who recently said artificial intelligence will reach human levels by 2029. Follow that out further, say 2025, we would have multiplied the intelligence, the human biological machine intelligence of our civilization by a billion fold. I guess that's what Dr. Pete was referring as augmented intelligence. And the best part is, we will get to witness all of this in our lifetime. We are truly living in an exciting world. And this beyond Big Data summit is a living proof of that. Thank you, Pittsburgh Technology Council for organizing this exciting event. As a leading technology organization, making big bets on AI. CGI is very proud, an honor to be associated with this even for those who do. For those of you who do not know CGI, we are a global company, large presence in Pittsburgh. And we are a trusted artificial intelligence expert, helping clients demystify and deliver responsible AI. We combine our end to end capabilities in data science, machine learning, deep knowledge, deep domain knowledge and technology, engineering skills, to generate new insights, experience business models, all powered by AI. I must acknowledge when I walked into this summit, as with I actually walked into the summit, a lot of excitement and expectation. And I'm even more excited today than I was on Monday. content and conversation has been so relevant. I am looking forward to continuing this conversation beyond this even Thank you PDC. Once again, back to you, Audrey.
Well, thank you, everyone. Thanks for being with us. Again, Pete. I can't wait to meet you in real time. And that you can come and we can just be bathing and all the products that Hershey has, has under their portfolio because I'm looking here, you know, the whole conversation about peanut butter and, and chocolate just makes me insane. Jonathan can tell you I want to devour like a whole bag just sitting in one session. So anyway, I'm glad you're healthy. I'm glad that you are that we know you now. And we really look forward to keeping in touch
so you can feel pleasure. I appreciate the opportunity and the fun.
It was great. Thank you so much. Thanks, everyone. Thanks. Stay safe.
Transcribed by https://otter.ai