AI and the Transformation of Food

March 2019

In March 2019, Gastrograph AI founder and CEO, Jason Cohen, was featured on The Peggy Smedley Show. You can listen to the full episode here and read the transcript below.

 

"AI and the Transformation of Food" Transcript

Guest: Jason Cohen | Host: Peggy Smedley

Speaker 1:

Welcome back to the Peggy Smedley Show. The podcasting voice of IOT and digital transformation.

Peggy Smedley:

Welcome back to the Peggy Smedley show. I'm your host Peggy Smedley. My first guest today was a founder of an interdisciplinary research institute, dedicated to the study and preservation of tea (indistinct). I'm not sure if I say that, and tea culture, he did his original research on sensory science and machine learning and founded his current company after only four years of research. He's sitting right here with me in the studio. Please welcome Jason Cohen, founder and CEO of Analytical Flavor Systems. Jason, welcome.

Jason Cohen:

Thanks for having me.

Peggy Smedley:

So, Jason, I'm really stoked today about having you in the studio. I love when we have guests in the studio because they have a great opportunity to share and tell me all the cool things right here with us. So first of all, welcome.

Jason Cohen:

Thank you.

Peggy Smedley:

And now secondly, have to tell me just after four years, what you are doing because so much is happening in the food industry, but especially food science, and that's changing. And what's really exciting is that you are helping change the way we think about flavor and taste. And you and I talked already before you came into the show about what's happening, but before we even get there, tell me a little bit about your company and what got you so excited about starting it.

Jason Cohen:

Absolutely. So what we do is we have an AI platform that models human sensory perception of flavor, aroma, and texture. And from that, it can predict the distribution to consumer preference. So we can look at different populations at different consumer cohorts, and we can predict the flavors that they will like, the products that they will like. And we use that to help companies develop better, more targeted products that people are going to enjoy more. And I got into that. I was actually a professional tea taster, and I got into it because-

Peggy Smedley:

Now I got to stop you there.

Jason Cohen:

Yes.

Peggy Smedley:

A professional tea taster. Now everyone out there is going, who's a professional tea taster? Come on now. I love tea, but well, who does that for a living? I'm a professional tea taster. What does that mean?

Jason Cohen:

So I wound up heading to China originally to study politics. And I fell into this world of tea. There are teas in China that are almost mythical or legendary things like Da Hong Pao, where there are four bushes growing on the side of a mountain in Wuyi National Park. And there are trees that could be a 100 years old or 500 years old, or even a 1000 years old that you're not allowed to pick anymore down in Hunan. And so I got very into Pu-erh tea, and Pu-erh is a type of aged and compressed tea. That's supposed to get better as it ages like wine, and people were willing to pay astronomical sums for this.

Peggy Smedley:

And is this tea, what does this tea actually do for you? I mean, is it some kind of tea that makes you feel better? Is that the concept behind it?

Jason Cohen:

I believe that tea generally is healthy. I don't really subscribe to any specific health benefits from any specific type of tea.

Peggy Smedley:

Okay. So that, for listeners who are understanding it. Okay, so now continue, go ahead.

Jason Cohen:

Yeah. And so I went to Penn State originally to study political science, and I quickly did not do any political science research. I wound up actually focusing on tea. And so I started looking at what do people taste in these different teas? Can we build models around what they taste? Can we say how this tea is going to change over time? And I started that research in Department of Food Science and Sensory Science. And what I found was that they weren't using sensory science, for the most part, was not using modern tools. They were not using machine learning. They were not looking into avenues around artificial intelligence. And this is starting back in 2009 before AI had become a-

Peggy Smedley:

Before anybody even thought AI was really going to change the way, I mean, beyond playing checkers and chess and things like that.

Jason Cohen:

Exactly.

Peggy Smedley:

Okay.

Jason Cohen:

And so I moved to Department of IST and Computer Science for Machine Learning and Artificial Intelligence and said, let's make, let's bring all of this into the model. Let's actually be able to build predictive models and then, and go and test those predictive models and see if we can predict what someone's going to perceive and how much that's going to affect their preferences for and against those products.

Peggy Smedley:

When you think about that, when you're thinking about people's taste and preferences, because our taste change you and I talked about this, that it's a little different to be able to predict, because as we're younger, we have one taste; as we get older, our tastes change. So how do you use an AI model to actually become a part of that? I mean, it seems like it would be almost difficult to do that.

Jason Cohen:

Absolutely. So one, it is very difficult. On the other side, we have to model that type of transformation explicitly. And so way that we think about it is that consumers in different cohorts, in different eras or age brackets, their preferences are going to change over time through adaptation, as they're exposed to new flavors, as they try new things, as they discover new products that they enjoy. And so this is true on an individual level where we're each going through our own consumer journey. And we're all looking for products that we like, and we're all willing, some of us are more willing than others to sample new things, but it's also true at a population level where entire cohorts of consumers are tasting similar products, are exposed to similar flavors, and are making consumption decisions from a similar basket of available goods.

Jason Cohen:

And so that makes that transformation predictable. And so we can look at, for example, the North American population, we can say right now, sour preferences are near perfect generational shift, where individuals who are above about 35 years old, have a neutral to dispreference for sour and individuals who are 35 years old and below have a neutral to a high preference for sour flavors. Just as an example.

Peggy Smedley:

When we look at that and we're saying different preferences, is this going to change then the food industry overall and the way we think about what our tastes and desires are going to be in food science in general?

Jason Cohen:

Absolutely. So right now, what's been from, say, turn of the century till today is we've had companies, CPG food and beverage companies trying to make generally acceptable products. Products that everyone is going to like. The problem with products that everyone likes is it means that very few people love those products. And so now the market has become more competitive than it's ever been before, right? We're seeing regional brands, we're seeing breakout brands. We're seeing emerging brands. We're seeing brands from foreign markets cross pollinating into new markets.

Jason Cohen:

And what we're seeing is that where there used to be one bestselling, say beer. There used to be one bestselling chocolate. We're now seeing hundreds of products that are each a better fit for a narrower consumer group than those products can be a fit for everyone. And so this is a classic innovators' dilemma, and we look at this and we say, companies need to move from trying to create singular products to needing, to create portfolios of products. Portfolios of products that people can pick the products that they love. And we think that type of better food and better beverages are going to make, better food is better for everyone.

Peggy Smedley:

How do manufacturers, food producers, so to speak, meet the demand of that? That's highly complicated, highly expensive. And the supply chain and logistics of that become very complicated to be able to meet those demands regionally, globally. And then there's food regulations. The FDA comes involved in all this. How do you meet the growing demand of hundreds of different products that you just described and changing taste buds that you just described? Our tastes are changing. So today we think we like it, tomorrow we don't. It's out a new flavor's in. I mean, that's complicated.

Jason Cohen:

Yes, of course. And I'll use the famous Facebook slogan, "Move fast and break things." There was-

Peggy Smedley:

Well, we've seen how they break things. Let's not go there.

Jason Cohen:

Well, it used to be right, that software and the internet had a faster generational time. It was easier to create breakthrough shifts. It was easier to tinker and to change things and test things, right? Because there was nearly no cost of that change, right? Someone can go in and change a few lines of code. And suddenly millions of people around the world are viewing a slightly different website or using a slightly different app, right? And today you can't do that with food and beverage. You can't do that with physical goods, but we're seeing now in other industries where things like that are becoming possible. So 3D printing is making it possible to actually do small scale manufacturing for highly customized things. And to create different versions of them that you can test quickly and cheaply. We're seeing that in the music industry, we're seeing that in content now. Netflix is talking about allowing customized movies and TV shows, right, where you can choose your own ending.

Jason Cohen:

We believe the same thing is going to happen in food and beverage and CPG. And the way that's going to happen is the large food and beverage companies have the capital equipment. They have the supply chains to do it. They need a shift from being a product focused to being portfolio-focused. And so I don't think it's going to be hundreds of products in one brand. But I do believe that it'll be anywhere from three to 10 products in a brand that can be placed into the regions and into the areas where the individuals who are going to like it, most are going to have access to it and be able to find it.

Peggy Smedley:

Are the big behemoths that we know, these big brand names going to change? And they're going to have a lot of little creative minds coming in to help them think differently, because that's a big mind shift, a big shift of market that they've had, market share, maybe is that going to change them? Because they don't move. They're big ships. They don't move quickly, is that they're going to have to think completely differently then.

Jason Cohen:

Historically, they haven't moved quickly. And I think that many of them have received a wake up call. They received a wake-up call because in the last five years, particularly, they were running an M&A strategy where they were attempting to acquire the breakout brands. They were attempting to acquire these brands that people were going for easy for, and frequently paying 20X, 30X, 50X revenue multiples, even forward-looking revenue multiples for these products and for these brands.

Jason Cohen:

And what they found is that some of them have been able to successfully integrate those into the mothership, right? But really, what they found is that the majority of those brands, those were the brand of the moment. That didn't mean that those are going to be the brand of the future. And so if they hired, if they acquired them at the height of their popularity, that doesn't mean that it's going to sustain that level of growth or even that level of popularity. And so they really need to be able to bring the technology and bring the ability to incubate and create their own brands that are targeted for the right people at the right time.

Peggy Smedley:

Are there a lot of hurdles for them right now? Because it sounds like this isn't something that can happen overnight. We've seen the internet of things take a very long time for manufacturers. We've seen it take a very long time for the food industry. They're starting to understand the internet of things in AI, but they're not there. They've been very reticent to embrace this. Is it going to happen all of a sudden, or is it still a slow moving tide that's starting to permeate them a little bit, well, what's happening here?

Jason Cohen:

Yeah. I think, what it is, is we are seeing the cracks in the dike and we're going to see a little trickle of this type of thing at first. Companies are going to dip their toes in. They're going to run a bunch of experiments. And then, when they see that it works, and when they see that another company, one of their competitors' done it, and it works, that dam is going to break.

Peggy Smedley:

So then it's going to be the me too, is going to really take hold because they're going to say, look, we've got to be like the Jones' and really say, we've got to do that, or we're going to be left behind. And we're starting to understand, because if artificial intelligence, and that's what we were seeing; Amazon showed us with Alexa and we saw Google Home, that our consumers are embracing that. And now, from a business standpoint, we're seeing the benefits of it. It's not something that's going to harm us. It's going to give us intelligence-

Jason Cohen:

Exactly.

Peggy Smedley:

... we never saw before. So when we think about that happening, and we see how machine learning is going to take it to the next level, are we going to see some leap frogging of really embracing this technology faster than what we have at this point?

Jason Cohen:

Absolutely. Absolutely. Think about it this way. When companies, when the internet used to come out would say things like, "We're an internet company," right. Everyone used to say amazon.com, right?

Peggy Smedley:

Absolutely.

Jason Cohen:

And then when cell phones came out, suddenly you had come companies that said, "We're a mobile company. We're a mobile first company," right? Now, every company is a mobile company. Now every company is an internet company. And now, right with AI, some companies be like we call ourselves, "An AI company," right? But eventually every company's going to be an AI company. This is going to be a technology that's going to permeate the different functional groups and the different capabilities, and it's going to empower a lot of capabilities that are not possible now. And I don't want to oversell machine learning and AI, I think that there's, one, I think there's a lot of companies that are attempting to do things that there's not enough data for. There's not enough predictive power for, or that existing systems continue to work better. I think that there's a lot of, if you look at the big data trends where people are like, we have big data, we're going to do big data things with the big data because we have it. Right. And then-

Peggy Smedley:

They don't have it yet.

Jason Cohen:

Yeah.

Peggy Smedley:

They talk a big game, but it's not there. So that's what we were saying. You've got to have the information to really make it work.

Jason Cohen:

Right. And so, I'm not claiming that AI is going to come in and solve all of our problems, but I think that there are very specific problems that AI will make a big impact on. And we are seeing it in content, we're seeing it in music. We're seeing it in non-food and beverage products. And I think that we're going to start to see a lot more of it in the food and beverage industry.

Peggy Smedley:

We talked about that AI is going to take away mundane tasks. It's going to give you the information to be able to make better predictive decisions and do things faster, quicker. Is this what we're talking about? Because that's the analytical side of things that you're talking about with enough information, you can make better decisions.

Jason Cohen:

Right.

Peggy Smedley:

And those decisions are going to replace the poor decisions that you've made up to this point that you didn't need. Is that what we're talking about?

Jason Cohen:

That is what we're talking about. And one of my favorite examples of this is many companies decide if they're going to release a food and beverage product, using something called a 60/40 test, that's where they take, say, their competitor products and their new product. And they say, we're going to get a panel of about a hundred people. And 60 of them have to like our product more than any of the competition, right? So 60/40, 60% have to prefer us. And if we can beat that test, we're going to launch it. If we can't beat that test. We're not going to launch it, right? The problem, there is, one, that's binary. That's a yes or no answer; that gives you no direction over what's good, or what's bad about the product. It's predictive over whether that product was actually going to succeed when it goes out to the marketplace. The panelists that you bring in to tell you if they liked it or not, they're not a stratified representative sample, right, of a homogenous population.

Jason Cohen:

And so that's not a predictive test. That's a bad decision metric. It's not analytical, it's binary and it's not predictive. And so what I think AI is going to do, what we've seen AI do, is we've seen AI take tests like that, and actually be able to instead give you a predictive result. You can see what percentage of the population is going to like this. You're going to see how much they're going to like it. And you're going to see who exactly in the population is going to like it, right? There could be a man versus woman split. There could be an age split. There could be a generational split, right? These are things that are just going to fall out of the ability to make these types of predictions. And that's going to lead to better products for everyone.

Peggy Smedley:

When we talk about binary being so important, when we then start looking at technologies emerging, we start seeing all of these new things. The next thing that comes is everybody talks about when you're in the food industry, you have the ability to start doing a blockchain. You have the ability to do transactions. Now you have a lot of very strong emotion about blockchain. There are benefits in a transaction through a blockchain. Why are you... You cringe, even when someone brings up a blockchain, you could just see how this upsets you. Are there not benefits to a blockchain?

Jason Cohen:

So, so I am a blockchain skeptic, and my skepticism arises from the technology and from the use cases of technology.

Peggy Smedley:

So you don't think that exchanging of transactions through a blockchain, in the food industry and food sciences is a good idea.

Jason Cohen:

I think that there are use cases for well designed immutable distributed trustless data stores. I think that the blockchain is a poorly designed immutable distributed trustless data store.

Peggy Smedley:

Because you don't know what's going on there. It's simply, you don't know who's there, what they're going to do with it and everything that's behind it. That's that trustless part of it that you're talking about.

Jason Cohen:

Right.

Peggy Smedley:

That people believe it's going to happen. We've already seen bad things happen globally behind these blockchains.

Jason Cohen:

Yes. And there's a nice exam where companies are now starting to put food and beverage information or sourcing information onto the blockchain. And my questions, they are, one, if it's your company and your company alone, right? Then why does it need to be trustless, right? Either, consumers are going to trust your brand or they're not going to trust your brand. And if you need a resort to some trustless technology to promote your brand, I think your brand has other problems that it needs to solve. That's one.

Jason Cohen:

Two, if you are going to be sharing this, let's say, that you need all of your vendors and all of your suppliers on it. Why does it need to be distributed? Are you sharing all of your information with all of your vendors and suppliers across each other? Or are you just localizing all of that information internally, right? This would be much cheaper and much easier to run if you just had a normal database that they could write to. You could make a right only database, give right access only to certain companies. And three, why does it need to be immutable? People make mistakes all the time, right? If someone enters in the wrong information, why should that be stored and distributed to everyone forever? So I haven't seen a killer use case that has dissuaded me from the skepticism.

Peggy Smedley:

So if you think about that, immutable decision making right now. Why isn't it others seeing the same thing? Why isn't that thought of a blockchain, getting more people to say, "We need to rethink this," because there's a lot of big companies out there, banks on the financial side, putting their thoughts and beliefs behind the blockchain. But there's others who aren't, there's other big financial institutions saying, "I am not jumping on this blockchain bandwagon." So I'm both sides of the fence. I'm seeing it right now.

Jason Cohen:

Definitely. On the financial side, there could be transactions. There could be things that make more sense. I remain skeptic mostly because I think that proof of stake and proof of work are both frequently misapplied. So, for example, proof of stake means that, that would be voting by percentage of ownership. Proof of work means being able to solve a hashing algorithm faster than others. I don't think that either of those solves problems for distributed decision making, and that's where things like Ethereum stepped in with smart contracts. But when Ethereum got hacked and someone stole that, they said, "Right, if the code is reality, well, then we just had a bunch of money stolen. So we're going to go hard fork that, revert that," right? So you actually needed a human intermediary to step in. And so even in the financial world, right, this has not yet become a de facto standard. Not everyone has bought into this. And so that's one side. On the food and beverage side, I see fewer applications than on the financial side.

Peggy Smedley:

So you can have a blockchain. The idea of a blockchain can be there, but we still need to have a, I hate to say, a governing body that really, that takes charge of this-

Jason Cohen:

Exactly.

Peggy Smedley:

... is where you're saying. And I think that's where the industry maybe is losing sight of this. And so when people see that, don't throw the baby out with the bath water, but there's some benefit if we do it in the right way, is what you're saying with this. And so if people are looking at this saying, if we do it the right way, there may be some hope for this with the way you're looking at it.

Jason Cohen:

Well, I think, if the blockchain has done anything well, it's made boring back-end, back office technology exciting, right? And it's gotten executives in these large companies to re-look at things like their supply chain, their sourcing, their data storage-

Peggy Smedley:

Their security.

Jason Cohen:

... their sata sharing, their security, right? And it's got them saying, "Okay, now I see that we need to make improvements there." The improvement should not be, right, a CIO of a company saying, "Get me the blockchain." But if it's gotten them to look right, it's done one good thing.

Peggy Smedley:

Paying attention to things that they haven't done, right?

Jason Cohen:

Exactly.

Peggy Smedley:

And one of them, going to the security, we're not doing enough across the board to look at security, right. I mean, even though we have governments saying, we've got to do things we're still not doing, especially when we connect everything, right?

Jason Cohen:

I 100% agree.

Peggy Smedley:

So when we look at this right now, when you look at, what do you think are the biggest trends, then? Taking this all with AI and machine learning. If we talk about blockchain. The next thing we could say is quantum's coming. There's a whole lot of things that everybody's thinking about because we always tell companies, "If you think you're there, you're never there." You're always looking to the next thing. As you just said, an AI company, everybody's going to be an AI company. Well, what's the next thing that's coming. You have to constantly stay ahead of it. What would you tell food companies, food science companies that are in the industry that are looking at the food industry and saying, what do you got to be thinking about? Because you got to keep moving forward?

Jason Cohen:

Right. So I think that the Christian Anderson innovators dilemma talked about sustaining innovations. And he talked about disruptive innovations. The companies have to stay on top of the sustaining innovations. They have to be able to produce a better product. They have to be able to respond to competitor products. They have to have the abilities internally to create product leaders. Sometimes, those abilities to out-compete where you are today to out-compete companies that are encroaching on your products and your territory, and your sales. And in order to develop new products that are going to create new markets, just look at something like Kombucha, which carved a market for itself, right?

Jason Cohen:

Sometimes that actually comes from disruptive innovation. And that's what we're going to see. Right now, that's the transformation that we're going through with AI. And with the ability, giving companies the ability to predict perception and preference of highly targeted, different demographic groups and consumer cohorts is going to be one of the things that helps those companies compete. What comes after that, I think, is going to be another mix of sustaining and disruptive innovation. And I hope it's quantum because quantum computing will break blockchain.

Peggy Smedley:

Well, and break a lot of things, right? We're all in trouble. I mean, we're all thinking future. But looking at it from a flavor perspective, like you're looking at it right now, what do you see as really exciting? You're talking about what you are doing. Are there others, like you just mentioned, that are really game changing? That really excite you; that's saying, "Here's a world that's really changing," and you've seen things that you said, "I'm inspired by what I'm seeing out there."

Jason Cohen:

Yes, definitely. So many companies now are attending a great conference that we were both at the Smart Kitchen Summit. And I think that one of the most interesting trends that's going to take place; that's two years, three years out in the food and beverage industry, is moving production to point of sale or consumption. And so we're seeing that a little bit with things like Bevi and SodaStream, but there's others like Our Ocean. And then there's big players like Coke freestyle and Pepsi Next. Those machines have the ability to create what's right now, customized products, but they don't have the ability right now to create personalized products. And the difference is that customization is saying, I don't want onions on my hamburger, right?

Jason Cohen:

Personalization is getting a product that's perfectly crafted to meet your preferences. And the reason that they don't have it is because those machines, while they're capable, well-built machines that can blend different flavors together, on an individual serving basis. They don't know what those preferences are. And so I think that when you cross the capabilities of AI and the capabilities of this edge manufacturing for food and beverage, you can get personalized products which opens up a whole new world for where companies can compete and what it means to be a brand

Peggy Smedley:

And how far away are we from that? And I mean, talk about it, are we there?

Jason Cohen:

The technology is here today. You know what they say, "We're already living in the future. It's just not evenly distributed," but we have not seen that out on the market yet, but the technology is here.

Peggy Smedley:

And, but that's the exciting part. It's here, but how long before we can all enjoy it. We can say that, wow, I'm getting exactly what I want. And I'm living in a world that makes me feel and taste what I want.

Jason Cohen:

Yes. I think we could be as early as two years out. And as far as five years out, I think that the first ones that we're going to see are going to be much larger machines. They're probably going to be in something like a bar or a supermarket or a restaurant. And eventually, we're going to start seeing them at home.

Peggy Smedley:

It's exciting times. What do you think, we, as an industry need to do as consumers to appreciate this, but as businesses to understand where we're going to be? I mean, these are of the things right now, we're in changing times and it's constantly evolving that we talk about. So is there things now that we have to say, look, enjoy the times because it's not slowing down, appreciate what's coming?

Jason Cohen:

Right. And I think that the best way for these companies to take advantage of this and to enjoy this now is to make larger investments into exploring and applying new technology. They don't all have to be successful. You don't have to win every pilot. You don't have to come back and say, "Everything has always worked for us," but you need to set aside the budget. You need to set aside the time and the resources to start piloting and understanding these technologies and how you are going to interact with them.

Peggy Smedley:

Well, Jason Cohen, founder, and CEO of Analytical Flavor Systems. I really enjoyed this conversation. What's your URL so our listeners can go up and learn more about you guys?

Jason Cohen:

Our product is called gastro graph AI, and our URL is www.gastrograph, G-A-S-T-R-O-G-R-A-P-H.com.

Peggy Smedley:

All right. Thank you so much, Jason.

Jason Cohen:

Thank you.

Create a better-tasting world

Get in touch to speak with our team so we can learn about the challenges you're facing.