The Decibel - How AI influences the food we’re eating

Episode Date: October 20, 2023

Could your next favourite foods be created by a robot? While artificial intelligence has long been used in the production and assembly lines of food, some companies are now turning to AI to create new kinds of food, recipes and combinations that the human brain hasn’t conceived of before.Ann Hui, the Globe’s demographics reporter and former food reporter, joins the show to talk about this new frontier of food creation and the vital question of: does it taste good?Questions? Comments? Ideas? Email us at [e-mail]

In the name of science, today I'm eating muffins. Baked by our freelance producer, Michal Stein. Well, let me tell you, there's something special about both of these muffins. Okay. One of them is from just a regular recipe from a regular recipe writer who I make her recipes all the time. The other was written by an AI. The challenge for me and our producer, Rachel, and our senior producer, Adrian,

was to see if we can tell which muffin recipe came from AI. So is the rule that we're judging which one we think tastes better? We're judging which one we think is AI-generated. Okay. Which recipe? The worst or the best one? Well, so I guess you're kind of assuming. I'm assuming the AI ones are going to be worse, but I mean.

Okay, so there's two different kinds of muffins, and they look kind of different, but let's see. Okay, I want to say that the really dry one is AI. Why? Well, because I feel like this one, if you were like, give me a recipe for apple muffins, it would give you apples. And, you know, they would put apples into a muffin form. They may not, like the AI may not think to add the zucchini, carrots, raisins, etc. Also, it's just so dry. Over the past year, artificial intelligence has become

a bigger part of our lives, thanks to ChatGPT. We even turn to it for food combinations and recipes. I will agree that muffin one, the carrot zucchini one, there's more stuff in it. It seems like what a human

person, a baker, would think that, you know, you should like pile stuff into it versus the second one. It's giving you exactly what you're asking it to give you. And just like looking at the consistency of it, like the first one seems and feels more like a muffin. I'm going to go muffin number one, the darker, browner, zucchini one, as the human recipe and the one that tastes best to me. I'm going to second that. I agree. I like muffin number one. Well, here's the thing. You guys got it. Yay!

Okay, so we got it right this time. But the reality is that AI is becoming more and more common in the culinary world. And that is raising some questions. Globe reporter Anne Huey looked into this. She's thought a lot about the intersection of food and identity and has even written a book about small-town Chinese-Canadian restaurants called Chop Suey Nation. Today, Anne will tell us how the industry is using AI to innovate new food combinations, and how technology has already changed our relationship with what we eat.

I'm Mainika Raman-Wilms, and this is The Decibel from The Globe and Mail. Anne, thank you so much for speaking with me today. Thanks for having me. I want to ask you about this new beer that this company created. It's called Robo Beer, and they used AI, ChatGPT, to actually make this recipe. So can you tell me about this? How did they do that? So this is a company out of Victoria, BC, and they're a small craft brewery, a very small brewery that's very focused on experimentation.

So they're always trying out new recipes, trying out new formulas. And so they did something very cool late last year where they were playing around with chat GPT and their personal lives and and also trying to experiment with, you know, whether it could help them with their work as well. And so they got the idea to ask chat GPT to come up with a recipe for beer. And we're very surprised to find it spat out a recipe very quickly that was not half bad. And like, did it taste good? Like people said it was it was a decent beer? Yeah, yeah. I mean, they had to make a few adjustments, but they were able to produce it,

they were able to sell a bunch of it. And, you know, they consider it more or less a success. Fascinating. I asked you about the taste just because we just did an AI taste test with muffins in the studio here. How did that go? It was very interesting. We could tell the difference between the two muffins and we did find the AI one. We thought it was the muffin that was not as good. So we thought the human recipe was better. But it's interesting just to think about like, OK, so they're using this similar kind of process to come up with beer. And if you didn't know that it was made by an AI, you probably, I guess, may have not. I wouldn't have known the difference with the muffin.

That's wild. All right. So this is the possibility now to go to something like ChatGPT or another AI to find a recipe, like a beer recipe, like a muffin recipe. But what are some of the other ways that we're seeing AI used in food production? You know, there have been a bunch of different ways of using AI across the food industry, just like we've been seeing AI implemented in all kinds of industries around the world. And I think some of the more obvious ones are probably similar to other manufacturing and production industries,

where they're using AI or kind of robotic technology to speed up processes or to kind of find more efficient ways of producing things. And with food companies, we're also seeing AI being used a lot in customer service. So for instance, you know, you have a lot of large food companies, we're also seeing AI being used a lot in customer service. So, for instance, you know, you have a lot of large food companies that are using chatbots as their, you know, first line of customer service. There's also big food companies like Wendy's, for instance, has been experimenting with using AI chatbots to man their drive-thru windows at their restaurants. So there's a lot of that kind of thing as well. But something that I think is particularly interesting has been the use of AI in new product or new recipe generation. So this idea of using AI technology to come up with either new dishes or new combinations of flavors and tastes that,

you know, human brains might not otherwise be able to come up with. And so it sounds like we're often talking about like creating new recipes here, too, or new food combinations. I guess, traditionally, how would this be done? Like if we're talking about a chef or a recipe developer, what is the process of them usually coming up with a new recipe or new food combinations? So when we're talking about historically how food and dishes have been created, you know, it's been mostly through tradition and knowledge that's passed on through many, many, many generations. That's essentially what a recipe is, right? It's somebody came across a combination of foods at one point, and then they pass it on to others and they document it and that gets passed on. And then

when you're talking about kind of cutting edge chefs or even food companies in more recent history, when they're trying to come up with food, a lot of it is really trial and error. Cooking is trial and error. It's trying a lot of things. It's making a lot of mistakes. And then every once in a while hitting upon something that's really great and something that tastes really good. And then you replicate it and then you fine tune it and you get to a place where it's, you know, something that you want to share. Okay. So, and when we're talking about that trial and error process, how would AI change that process? That's something that AI proponents say is really the main benefit of AI. It's in speeding up that whole trial and error process, really,

hopefully helping us be able to fast forward through, you know, the first 99 tries where we were initially making mistake after mistake after mistake, and then getting to that 100th try that's, you know, reasonably informed. So that where say a food product in the past might take weeks, or even months or years to get to a place where it was really, really good, you could now get to that final version in the time span of like weeks or even sometimes days. And so what is the AI actually doing here to find the combination that works? Some of the kinds of things that I've seen include, there's this company called food pairing in belgium for instance and what their technology does is it scans through a database of thousands of ingredients analyzes them based on their flavor

aroma and texture profiles and in doing that, they're able to come up with combinations or pairings, as they call it, of ingredients that, you know, you and I may not have thought could work well together or could taste good together. But that, you know, based on the science, they found actually share certain characteristics that might lend themselves to tasting good together. Wow. So that's really interesting. So it's essentially, it's kind of looking at all these different kinds of food and it's breaking it down. It's like the flavor profiles of each food essentially, and then seeing like what could match with that. So they've worked, for instance, with some high-end fine dining chefs, Michelin-starred chefs. In coming up with ingredient combinations, one of the combos that they're well-known for is the pairing of kiwi with oysters,

which I think most of us probably wouldn't have thought would taste good together. No, we might not go to that, yeah. I mean, I'm keen to try it, but again, it's not something that, you know, we might initially think to do. They've also worked with like big food companies to come up with new, say, potato chip flavor combinations. You know, so it's this technology that they're using that can potentially be used for kind of all different purposes. There's also a lot of AI companies trying to create recipes for fake meat. This is a big thing now too, right? And people have been working on plant-based meats for a long time. But I guess what changes with AI? Like what does this bring to the equation now?

Yeah. So one of the companies that I looked at for this story is called NotCo and they're based in Chile and they use, again, this AI technology to try to find plant-based ingredients in this world that we may not otherwise know can actually bring characteristics of meat or animal products to food. has helped them discover that, for instance, the combination of cabbage and pineapple can replicate somehow the taste and texture and mouthfeel of dairy milk. Fascinating. I mean, aside from being very cool, just from like a science and geeky standpoint, and also, you know, being potentially commercially lucrative for not co, what they also say is the promise of this technology is that if it can a speed up product development and bring to shelves quicker, that that could potentially bring down the cost of that food.

And obviously, again, if they can potentially make some of these plant-based foods taste better, that that is good for the overall mission of more sustainable eating and having more people eat in a plant-based way. Of course, we've seen in the past with some of these plant-based meat products, a major stumbling block has been two things. It's been the cost of the products and also taste. A lot of people still have concerns about the taste or the quality of some of these products. And so if you're able to make improvements for either of those things, it could potentially be huge for the sector. We'll be back after this message. All right, so we've talked a little bit about, you know, how this is done.

I wonder, like, how good is AI at making these kind of recipes right now? Is it as good as people made recipes? I mean, you guys had your experience with the muffins. The muffin taste test. Yeah. Which we can tell the AI one just wasn't as good in this specific example, right? Right. Others have done similar experiments as well. I remember the New York Times did a story, I think last year, where they tried to cook an

entire Thanksgiving meal entirely made out of AI generated recipes. And the outcome there was not so great either. I think that a lot of these are still newer technologies. They're still early in their evolution and still need a lot of tweaking. One thing that's been repeated to me over and over and over again from everybody that I've spoken with is that no matter how sophisticated these technologies can get anytime in the near future, they're still going to need that human element, whether it be along the way, kind of guiding the process and making sure that things that, you know, the machines and robots are never going to quite understand about kind of humans and people and consumer preferences and that kind

of thing, making sure that the process is, you know, kind of has guardrails along the way. And then also just at that final stage of tasting, of testing, of looking at and touching and tasting. I think that human step is always going to be absolutely crucial to the process. Can you give me an example of like the importance of, I guess, human guidance in this way? Like what, I guess, what would the AI miss that we might catch? Well, one example with NotCo, for instance, that's the company that makes the plant-based milk. In one of their early versions of trying to come up with that dairy substitute product, the AI technology actually thought to use dill. They thought that it'd be a good idea to use dill,

except for the fact that it actually turned the drink green. So there are those things where humans are still having to intervene and still having to get involved and kind of teach the technology. Okay, so it sounds like there's a lot of really cool stuff happening here, Anne. But I imagine with all things AI, right, there's got to be concerns on the other side. So what are the kinds of hesitations, I guess, that people have about using AI to make our food? There are, of course, a pretty broad range of concerns, just like I think there are with all uses of AI right now. It's still a new technology. So I think there's a lot of questions and a lot of fears. Of course, there are the concerns that, you know, will this technology or will these programs

potentially replace human jobs? I think that's a very real concern that we hear about AI in general, and especially in a country like Canada, where so many jobs are in food production, are in the food industry? I think it's a fair question to ask, for sure. I think ultimately, though, a lot of the discomfort comes from just the importance that food holds in most of our lives. I think that for all of us, food is an important expression of our cultures. Food is an important part of our relationships. For many of us, it's, you know, when we think of food, we think about our mothers feeding us or our grandmothers feeding us. and I think that food has always been an expression of sort of who we are as a community and who we are as humans and so if you think about the saying we are what we

eat what then does it mean when what we eat is no longer being produced by something or someone who is human? Yeah, it's an interesting actually thing to start thinking about when we're having this conversation. And I'd actually like to take it a step further here. And too, if we can just step back a bit and look at how not just AI, but technology has already changed our relationship with food. What are the kinds of ways that it's shaping things differently now? It's interesting because food and technology have always been linked, right? So the way that we eat and the kinds of food that we eat

has always been shaped in large part by the technology that we have available to us. With Instagram, for instance, it became all about visually attractive food, right? About food that you could photograph, food that you could, you know, take that overhead beauty shot of on the plate in the restaurant. Restaurants with really good lighting all of a sudden became very important. So you can see these examples where food and technology, you know, are really going hand in hand with one another. I think TikTok has been really interesting, both in terms of like the platform being a video platform and also it really becoming popular around the same time as the pandemic meant that a lot of people suddenly, the focus has been very much more

on home cooking. So Anne, I mean, we've talked about a lot of different ways that AI is being used here, how companies are using it to innovate in food and food production. But I guess I wonder about the personal level too, though. Like what might it look like to have these tools, to have AI in our kitchen when we're cooking? A lot of these tools are already in the works. There are big grocery stores out there who are using this technology to try to help consumers as they're making their way through the grocery store, figuring out what to buy and put into their carts. There are kitchen appliance manufacturers, people who make refrigerators and stoves and that kind of thing, who are working on technology that can figure out what we have in our refrigerator and

based on what day it is, tell us, you know, spit out a whole meal plan for us complete with recipes or, you know, put in our next grocery order for us. I think that there have been a lot of arguments over the last many decades about humans stepping further and further away from their food and becoming, you know, increasingly detached from their food and the production of their food. And, you know, most of us go to a grocery store and we buy food wrapped in, you know, nice packages and we never touch kind of a farm or we don't necessarily have

any hand in growing our food. Many of us pick up a cell phone and press a bunch of buttons and then food magically is delivered to our houses. Like, I think that is a very valid point. And if you are in that camp, then this idea of like AI and robots producing our food, you could potentially see that as, you know, the next step in us being increasingly alienated from our food. But I think that because AI also means that, you know, you or I could go to one of these tools and punch in, well, this is what I have in my refrigerator and I only have 20 minutes to cook something and this is what I feel like cooking and it's going to spit out something that's directly tailored to me and it's going to

be delicious and hopefully hopefully it's going to be delicious and hopefully it's going to be something that I can actually make. I think that there's also an argument to be made that this kind of technology might bring more of us to cooking, might make, you know, cooking and food feel and seem more accessible. And thank you so much for taking the time to speak with me today. Thank you, Menaka. That's it for today. I'm Menaka Raman-Wilms. Michal Stein helped produce this episode. Our producers are Madeline White, Cheryl Sutherland, and Rachel Levy-McLaughlin.

David Crosby edits the show. Adrian Chung is our senior producer. And Angela Pachenza is our executive editor. Thanks so much for listening. And I'll talk to you next week.