Mapping the Flavor Genome With Beth Altringer
Hey guys this is Mike Wolf. Welcome to the PODCAST podcast is with Dr Beth alter the head of the Flavor Genome Project also a harvard professor and again crater. She's doing lots of interesting things in the world of food and flavor and dissecting that and so I thought I'd catch up with Beth weekend introduced through some mutual friends and interest a really interesting conversation so I hope you enjoy it as always you could find more podcasts at the spoon were also available. PODCAST is also available on spotify as well as where you get your podcast US apple podcasts other usual suspects before we get started just WANNA mentioned. We're just a couple months out from the circuit to summit. We weren't this week that Adam Brockman the C._E._O.. Of Bright Loom formerly Isa the former digital officer for starbucks will be coming to speak so we just have a a bunch of great speakers coming go to smart some dot com use Discount Code podcast at twenty five percent of tickets and not only that I'm just a little over a week out from heading over to Japan. We're going to our third annual. We'll be having your three in. We'll smirk into some Japan. If you're in Tokyo you love to connect with you. I'm there from August fourth to the twelfth. Just reach out to me through the spoon. Hopefully GonNa be reporting a lot of stories from Japan as well so we can afford that that's for now folks. What's the podcast? Hey One roy said it have Dr Beth alter on the podcast custody how you doing breath. I've been great. How're you good? You are doing lots of interesting things. You are a entrepreneur <hes>. You're also a professor at just a little in this little university. I'm thinking sort of I think it's called what Harvard Yeah and I just learned about you. Through mutual friend and discussion thought apt to talk to Beth because you're doing these things you're working on. This thing called the flavor genome project. You have a game called the chef. If we before we before we get into one kind of here. What about your path and how you got into food in food innovation? Oh Gosh Yeah so I my field is product design but in Grad school I I want to Cambridge University and they have this very competitive. <hes> wine tasting organization and I thought it was interesting. This is a blind interesting training organization ideas store with wine drinking. I guess I'm not surprised totally totally so I was going to this club and then eventually <hes> they have they had the tasting <hes> <hes> like the tryouts and I was like what's and and I tried out for team I made the bottom of the team. There's like a ruthless match with Oxford and but that was where our training started and so so the bottom of the best one taster. You're just like the people who made the team. I was the reservists that's when you know somebody gets a cold that you're on the bench but then the real training started and then the winners about competition compete internationally get to go to competitions and France and whatnot I learned a lot from it and it stuck with me so you weren't to love this idea of tasting flavors and it kinda spark something in your head yeah until I mean the idea that the story of what a wine is and where it was grown and that the sugar and alcohol content would be higher if it's sunny side of the same hill then if it's on the other side is if just kind of blew my mind and <hes> and I didn't really want to become a chef or Somali a didn't really see a career in food but <hes> but it just made me really curious to learn more about a lot of different other types of food and by training and this is where you have your your your in product product design right so while you love at this idea of like flavors in what went into them you are also building and kind of building a career around product design yeah absolutely and I've worked on a lot of different types of products and I think maybe maybe what happened and I got to the point where I felt confident. I think <hes> pursuing some of my own interests you know combining product design an and flavor where I didn't really know how make that work before and at some point you you decided you know you became a professor at Harvard but you also were practising as basically in the commerce side in working and you start at a company in Taco about the origin story of the Flavor Genome Project. Oh Yeah I mean it goes it. It's directly related so I my my wine. Knowledge was atrophying so I wanted to basically create a you know a database ace like an interactive database that preserved what I knew and and kept it up today because I didn't really have the time do that so just started as as nerdy thing show it was like an idea for a personal like fun thing thing rather than to start yeah. It was totally just to preserve the that knowledge and then my interest also changed to be. I really wanted to be able to learn to understand cooking unlike some of my talented friends were able to do they could just kind of whip something up I I can make things but I have to follow recipe and so because I from from <hes> blind wide wine tasting competitions I knew that you like breakdown. The components of taste <hes> and I also knew from some of my other training that you could use you could you could understand a lot about <hes> subjective experience based on how people described things he could do natural language processing on and I thought Gosh I could probably computation only help myself understand food in starting with what I understand Danette wine so it really was pre personal expert exploration at first and <hes> united some friends who are interested in learning from it as well and then I just kept enjoying doing at and then and then I Leo I'm a big nerd so I just kept going. I think that's self self explanatory. You started something called the Flavor Genome Project and yeah it's you as as you know like <hes> are understanding scientifically about flavor and ingredients and how they come together is really rapidly advancing right now and so it's just such an exciting space to bring <hes> data together from a lot of different sources de. Let's talk about that right so so we have had people who are just really good at creating food. They understand it to some personal level. Maybe this have heightened senses around kind of identifying those things you talked about in wine elements but as you just indicated. We're we're marching forward fairly rapidly and being able to apply technology computational technology or the understanding of flavor and I think that's one of the things you're tapping into yeah absolutely but I'm also but not not in a way that is separate separate from human taste right and the human taste perception of flavor. Bring those two things together and and when you say you you WanNa bring those together is it identifying how the person and describes the the the flavor in the components and then being able to put that into some sort of data set that we can then use the long-term goal is to to to develop new ways to automatically understand flavor based on what we understand scientifically about it based on what we <hes> can learn from how people talk about it both novices and experts and then use different methods to identify patterns that are Kinda like new <hes> data driven genres of flavor rate right. Now we organize flavor by like you know Italian food or Greek food like does actually don't really make a lot of sense like if if we could organize it based on <hes> the regional patterns acidity preferences you know there's all kinds of things that we could be doing that would be smarter approaches to grouping what people like and how they pretend they perceive even the same dish when he's talking about the flavor genome project is ultimately. What is it now? People are going to heart people gonNA use it for example in product development. Yeah I mean it's it's very very experimental research software at this point so what I would like to do as connect <hes> you know what we're learning in a in flavor science to <hes> how people bowl experience flavor and flavor is actually really complicated in terms of human experience not only do you have some genetic differences and perception but you also <hes> have taste preference differences than those are linked to <hes> emotions to culture to history. It's very complicated space so instead of trying to answer you know what is the right thing to serve you right now what we're doing as just <hes> combining a lot of different types of data to make better guesses and would that ultimately help say C._p._G.. Product companies develop a flavor. That's more highly tailored towards like a certain subset upset or or kind of a Denver demographic etc based on what you've learned from the data yeah I mean. I think there are a lot of things that we could ultimately do with us. This is why it's pretty experimental at this stage but yeah yeah I think you could use it for product development. I personally very interested in in more kind of naturally driven products than <hes> synthetically driven products and so I'm interested in how we could develop new flavors that are also relatively healthy <hes> and the other side of it is how can we help <hes> consumers make choices that will be a little bit more likely to lead them to what they would find delightful <hes> while also meeting some of the other goals they might have that are not really delight driven like nutrition driven goals so in the labs but also ooh practitioner in it the kitchen level both maybe from a company standpoint but even a consumer standpoint knowing like if they have a flavor profile or certain aspects like you can maybe kind of point them in the right direction and creating something different than they've had they've experienced orange for yeah absolutely and I think that <hes> so I'm I'm very consumer driven in terms of that that perspective but I think a lot of us <hes> I was listening earlier podcast of yours and someone said I can't remember exact phrase but something like you know people struggled to name ten recipes and that we you know we tend to eat the same things over and over again. We tend to choose based unfamiliarity <hes> and I think that's partially because we don't really have something that is the coolant of flavor literacy right like we consume <hes> flavor all the time but we don't have educational feedback loops around so you don't really <hes> learn the equivalent of <hes> reading and writing in flavor that make sense and I think that if we could help people understand in a really accessible away how flavors interact and maybe like if you like Guacamole we'll guacamole as a is kind of a certain combination of fatty nece and acidity and texture and if you like that so you probably if you eat fish would lake severe J. as long as you don't mind a little bit of a runny texture right like we could organize and serve options to you based on some of those preferences. I really love that idea a expanding the flavor or seen kind of helping people better understand because you're right like we don't have the A._B._C.'s and one two threes of flavor wooder see I myself and I write about food all the time but I'm Mike Sane person to go. I use a teen the same ten things and most of us show like that so yeah giving consumers tools and I think also giving even providers of food products tools that they can help consumers. I make decisions and better decisions can be really interesting absolutely and then there's there's a practical side to to where the cost of you choosing something that you don't like so much. Is You know that's kind of wasted time and money eight to some degree but it's also the opportunity costs of something that you would have chosen that you liked. What better why I mean we eat? We we've had this conversation. We all eat three or four times a day. Have you thought about why we don't better trained trained consumers or why they have. There hasn't been the tools to help them better understand flavors better well. It's in this aisle actually goes back to that line training so the way that you get trained is really different than anything. I've experienced that you know and I spend a lot of time in universities so I see a lot of different educational models but this is you know the way I would describe it as like a sensory logic puzzles so you know like a logic puzzle where you are figuring out what things cannot not be based on observable characteristics and then you also have to have the sensory training so they train you to be able to do is estimate you know the acidity level alcohol level residual. Sugar <HES> Tannin level and they train you to be able to record your nose observations at time one two and three because the wine is changing a little bit over time and all of this is recorded right Zulic on the chemical answers you can be you. You know you're right or wrong on the <hes> knows answers. It's really a kind of within subject design in the sense that you are internally consistent. When you smell you you know <hes> ave or whatever but <hes> you and I would probably very different in terms of our our <hes> smell notes and you bring up point rarely each person has has their own context they bring their own like genetic makeup thrown ethnicity and all that would imagine feeds into something might taste very different to me the same exact thing <hes> to you and then there's just place right? If I'm at a different elevation we know science. Does things taste different different elevations exactly yeah I mean it's taste is really really complex to understand <hes> but back to your question about why haven't we done this well. So far I think part of it is because we computational we didn't have a system of tools and <hes> we are in this moment where that's getting easier like. I still think it's it's it's going to be wild before. This is easy to do but we're seeing. We're seeing that something has genuinely a friend about what's happening today. Now you're teaching at the highest session ones right. You're teaching at Harvard but could we see this basic flavor <hes> kind of education this kind of still down into a more primary levels of education. I guess it's something that could be the next shop class and ten years. Oh my Gosh I love that idea. I'm so I don't know if you know we at Harvard. We have this science and cooking educational program. We actually have a free online class. We teach <hes> we teach science through food and through cooking and it's it's a really fantastic way to teach all kinds of things you know. I think thank you could teach history. Classes in social studies classes through food. Some people are doing that but yeah so I think you're on something that we could. Have you know very food based curriculum in some ways in the future. I'm betting practical way like <hes> I would like this type of education to be available at scale and I've I've been building a game called called chef league that is really designed with that in mind as just a fun casual way to access really high quality flavor sites. Yeah and I'm glad you brought it up. I wanted to talk about chef league in what I weren't talking to you before the podcast that product is real your first consumer facing product built on top of your project right on top of the data set. There's your building for the flavor genome yeah absolutely so I'm really excited about it. Because I've wanted from the start. I wanted to create a consumer facing product and I wanted it to be you know not <hes> <hes> not snobby away right where this this product is not trying to say that this is good and this is bad. It's let's just pulling from an enormous amount of data and how different types of people judge recipes and so tell us about it right so it's chef league and what is the goal the game well the the goal is I think first and foremost this is a fun game and some people probably wouldn't even see it as a game at just the fun apps that you might interact with <hes> but you know the the mean reason that you would play this is because you want to <hes> you WANNA learn to be able to invent recipes improvise recipes <hes> do what type of prototyping for flavor that a chef can do in their head or that. I think we all have that friend who can whip up anything from you know so called nothing in the cupboards right and the aim is really designed to just be a fun way to do that so you get these challenges and it's like <hes> you know. This tuna steak is a day old. <hes> what can you do with it and you're competing against two other players to make the <hes> tuna steak is delicious as possible it had <hes> we have these little <hes> judges that are based on real real chefs and and they give you a score and nine if there's not an immediate score <hes> like if those A._i.. Critics can't agree than <hes> your idea goes to public voting so other players will vote on whether you're by adding pomegranate seeds to Guacamole as a better edition than adding tomato which is less creative so I'm competing against two other live people were kind of remote for me is like we're just friends right. Don't know the people see them. Yeah exactly so <hes> you're there. You don't know who they are but you're always competing as to people and then there's the live judging aspects as game player. Can I be one of those people that votes on others yeah absolutely you have to be actually there's a you know in order to earn tokens to enter the <hes> the recipe challenges you you have to do a certain amount about voting and so this is your first game. It's your first product and what else is on the road map. You have other ideas I mean that's. I'm really excited about this game so I'm going to be in this for awhile and Games are not that easy you to to to make an <hes> an like actually getting the game out. The door is really kind of the starting line so we just launched about it. You know few weeks ago we ever hands full but what's pretty exciting. Exciting is like the we're just GONNA keep pushing boundaries with <hes> the game as one product. Obviously I hope it successful <hes> but I want to just keep pushing forward. You know our understanding of how we can stitch stitch together the scientific data with <hes> with human experience of flavor data and the human experience of flavor data that we have today is actually pre poor <hes> I I think that's the lowest quality data that we have in our understanding of flavor so we're just gonNA keep working on improving that and connecting it to the great work that everybody else's doing. No you're a data you build products with data. And how did you get the consumers to input all that data run flavor so we've done a lot of different things so it actually started with you know because I had in you know all of my original <hes> wine tasting training materials. I built the backbone of it based on <hes> the way that we breakdown flavor understanding of wine and then I would go sort of one vertical at time so say cheese and and FI and learn how experts who are the equivalent of a Somalia in cheese. How do they break down the understanding of you know what cheese tastes like what are the key descriptors and some of those would map onto wine and might might be descriptors that we can stitch these two verticals together with and then they're also going to be a lot of new descriptors so so then we would add those attributes and then you might go to bread to beer and like each one is it? It's a slow process but the backbone of of it all is is how an how experts arts who developed that capacity to you know what what you might call flavor literacy to accurately described to one another a flavor experience it's based on that and then after that we added large volumes of of data that could be more thoughtfully attached to that backbone. Is that make sense. I think so I mean Somali as like that's a well understood like expert in a particular area that has has been <hes> attracting experts in celebrating that expertise for a long time cheese. I'm curious what a smalley cheese stamos but like there are just a number of different proactive better word verticals are areas <hes> some of which probably they have like more rigorous. I guess framing and freeze all and metrics around lake. I know that in beer they're certainly that and there's even like a Berry Berry X._M._l.. For example so did you tap some of those existing frameworks to help you kind of build it yeah absolutely and here you know that was our experiences. Wine is is kind of by far the most <hes> developed in many ways and beers pretty close behind <hes> but a lot of times like something like a croissant in like what does croissants simile like but those people do exist a person who can you could give if this person you know twenty different croissants and they could like really distinguish what makes them different and and that person would have you know of sample size of one opinion on what is the best Chris on <hes> where sort sort of less interested in that where more interested in what is an effective categorization system for Khorasan's and those people can be very hard to find but you know so far we you find that they they exist every category I wanna be for for Croissant donuts myself. I gotta get but exactly we're seriously like flavor. There is a thing but also texture mouthful is a thing but I imagine you're also quantifying trying to get a tat right. We are absolutely and so I mean the type of expert I'm talking about is somebody has to so with croissants for example bull run starbucks bought <hes> legal belong main Frenchaccent but <hes> you know they had to provide a consistent croissant product in every starbucks and like somebody committee has has has developed the language to get that differentiate what is acceptable quality and what is less acceptable level quality in terms of taste texture and Mike experience so we try to find it find those people and must have been half the fun. Did you get to actually go out some of them in their locales. I can go to other places I didn't my my husband was the one who it helps me with the croissant with John Descriptors and he actually met you know the C._e._o.. Of that company Liberal belonged Jerry and Oh fun side note he said that one of the key factors and croissant quality across America was how how much pride the person making the croissant took in their work which is just like a fun detail disease. It somehow actually taste that that's interesting. How do you quantify that fast yeah? We didn't put that in our framework but I think it's that's I think it's interesting and I think it's it's. I would love for all of us to kind of take more pride in our our flavor choices but yeah that's that's another problem. That's a get but it gets out at an important point right. There's so many intangibles around and flavor like in pride and <hes> you know dedication to the crap is something is really hard to quantify but it's it's certainly probably there. It is yeah and I think for what we're doing. We don't necessarily like it's not the key variable but it just just one second actually through I thank take up to Redo that little bit that we were talking about the Altea up on the pride part. Will you bring up a good point. I mean there's so many intangibles right so pride and craft and expertise that's hard to quantify but it certainly matters it totally matters and and I would argue that you could see it in something like a croissant like the the textual component of signs very important and it's very time consuming to get that right and if you don't have the you know if you don't have the pride in the craft the likelihood of being there's is lower but we're really only kind of looking at how do you define the textual component. So where can people find out more about what you're doing and particularly working. They find your game. Oh Yeah so the the game is only for now <hes> but you can find it live on the APP APP store <hes> it's chef league like sports league but of chefs <hes> but it's really meant for you know everyone from amateurs to <hes> people who are a little bit more advanced.