You may not know it, but artificial intelligence may be responsible for the food on your table. AI is transforming nearly every aspect of our food system, from before a seed is planted up to the moment that food is eaten. AI could even help you decide what food you should eat based on your own health profile. In this episode of Unfold, we take you on AI’s journey from seed to plate. Host Amy Quinton and guest co-host Andy Fell examine the ways “Big Data Comes to Dinner.”
In this episode:
Ilias Tagkopoulos, director of the Artificial Intelligence Institute for Next Generation Food Systems at UC Davis
Christine Diepenbrock, assistant professor, Department of Plant Sciences
Mason Earles, assistant professor, Departments of Biological and Agricultural Engineering and Viticulture and Enology
Christopher Simmons, professor and chair, Department of Food Science and Technology
Danielle Lemay, associate adjunct professor, Department of Nutrition and USDA-ARS Western Health Nutrition Research Center
Dan Vincent, former president and CEO of Pacific Coast Producers
Learn more about Artificial Intelligence Institute for Next Generation Food Systems in the multimedia story written by Andy Fell, “Big Data Comes to Dinner.”
Transcript
Transcribed by AI. May contain errors.
Andy Fell
Hey, Amy.
Amy Quinton
Oh, hey, Andy.
Andy Fell
Are you eating lunch in here?
Amy Quinton
Yeah.
Andy Fell
Don't you have a podcast episode to write?
Amy Quinton
Well, that's why I'm eating and staring at my bowl of pasta with tomato sauce,
Andy Fell
And how is staring at pasta going to help you write a podcast?
Amy Quinton
Well, I'm hoping it's going to spark some idea. I want the next episode of Unfold to be about food.
Andy Fell
Where did you get those chunky tomatoes for your sauce?
Amy Quinton
Don't they look good?
Andy Fell
They do.
Amy Quinton
Out of a can from the store. I'm a great cook.
Andy Fell
A can? You know, tomatoes pair nicely with both pasta and UC Davis.
Amy Quinton
Huh?
Andy Fell
Well, back in the 50s, a UC Davis plant breeder and an engineer walked into a bar . . . no. They teamed up to invent the mechanical tomato harvester. It was the first machine that sorted and loaded tomatoes automatically. The processing tomatoes the ones you use in cans for your sauce.
Amy Quinton
And I recall they also developed a new tomato that would be tough enough to survive that harvester. It was kind of square wasn't it?
Andy Fell
Yeah, the square tomato. Well, not really square but you know tough enough to harvest. Square-ish. Yeah, so perfect pairing, don't you think? Technology and food.
Amy Quinton
Right? Well, you know, it's funny you should say that since part of my job at UC Davis is to tell stories about food and agriculture. And part of your job at UC Davis is to tell stories about engineering.
Andy Fell
Well, then I've got an idea for an episode. Why don't we do something on the newest technology, artificial intelligence.
Amy Quinton
Just said it has to be on food.
Andy Fell
Well, AI is used in agriculture. We have an entire institute at UC Davis devoted to this, Artificial Intelligence Institute for Next Generation Food Systems.
Amy Quinton
Okay, wait, hang on. Let me move my tomato sauce aside for the moment.
Andy Fell
What are you doing your laptop?
Amy Quinton
I'm opening up ChatGPT.
Andy Fell
Are we supposed to be using that?
Amy Quinton
Shh.
Andy Fell
Are you just going to have it write the episode?
Amy Quinton
Oh, let's not go there. Besides, I've tried it and it's not as good as we are. I still have to do the work.
Andy Fell
Ah, then you probably aren't using the right prompts. It's all about the right inputs.
Amy Quinton
You know what it is good at? Coming up with names for a podcast episode. Because calling the episode AI in agriculture is boring. Here's an entire list. From Fields to Feasts: How AI Nourishes Our World. That sounds good.
Andy Fell
Uh, that's no good. I like number 2. Bytes and Bites: Unraveling the AI Menu for Tomorrow's Food. See how it's a play on words of bites with an "i" and bytes with a "y".
Amy Quinton
Yeah, but you just had to explain that to people. So people just listening to a podcast won't get it. Wait, this is it. Number four. Harvesting Intelligence. AI's Journey from Farm to Table.
Andy Fell
and to stomach.
Amy Quinton
We'll figure it out. Coming to you from UC Davis, this is Unfold. I'm Amy Quinton.
Andy Fell
And I'm Andy Fell.
Amy Quinton
it seems almost every day we hear something new about artificial intelligence.
President Biden
Harness the promise of AI to protect us from peril. Ban AI voice impersonations and more. . .
Amy Quinton
Opinions of AI vary tremendously. You love it, you're afraid of it or little both.
Speaker 1
AI is going to help us to be healthier, happier, more productive, and more creative.
Speaker 2
The most important thing to know about AI. It's the first technology in history that can make decisions by itself and can create new ideas by itself.
Andy Fell
Love it or hate it how you feel about it. It's already changing the way we work, live and even eat.
Amy Quinton
And AI has already entered the fields of agriculture.
Andy Fell
Advances in the technology could improve the entire food system from breeding better crops to providing more nutritious food.
Amy Quinton
It could even help you decide what foods you should eat based on your own health.
Ilias Tagkopoulus
This is a really transformative technology. In human history, there are a few times we had technology that was so powerful, and it is such an opportunity for creating a better world.
Andy Fell
That's computer science professor Ilias Tagkopoulus who also directs the AI Institute for Next Generation Food Systems at UC Davis.
Amy Quinton
And our food system is an intricate network of breeders, farmers, producers, distributors, and retailers working to ensure that food gets from farm to fork.
Ilias Tagkopoulus
The food system is really, really complex, the more complex the system becomes, the more the analysis of data and finding the best point of operation becomes difficult.
Andy Fell
It gets even more complex when you factor in other challenges to the food system like unbridgeable weather, droughts and floods, climate change.
Amy Quinton
Or shortages in supplies like fertilizer or an increasing demand for food from a growing population.
Andy Fell
And that's where I can help by analyzing vast amounts of data making predictions and learning from his mistakes. It can play a role from before a seed is planted in the ground up to the moment the food is eaten.
Amy Quinton
To understand this, let's start by talking about the tomato sauce in my bowl Andy.
Andy Fell
No, actually, we should start with the tomato.
Amy Quinton
Actually, we should get more granular than that.
Andy Fell
You mean cellular?
Amy Quinton
Yeah, so let's begin with a plant breeder and geneticist.
Christine Diepenbrock
I am Christine Diepenbrock an Assistant Professor in the Department of Plant Sciences at UC Davis.
Amy Quinton
Christine says when breeding plants for example, a tomato breeders want to select plants with the best traits to cross with each other to create the next generation or new variety.
Andy Fell
So they find the different regions of the genome that are favorable for creating those desired traits.
Amy Quinton
But there are several traits that are important.
Christine Diepenbrock
This includes the marketable yield of the crop, it includes disease and pest resistance, it includes tolerance to drought and high temperatures, resource use efficiency and nutrient concentrations, postharvest quality,
Amy Quinton
That's a lot of traits. And...
Christine Diepenbrock
Those can be you know, separately and together complex trait targets, there can be relationships between those traits, which might make it really hard to breed all of them at once.
Andy Fell
But AI can take that data to predict complex patterns and relationships. So breeders can make sure by selecting traits that increase nutrition, they don't also decrease the plant's ability to tolerate drought, for example,
Christine Diepenbrock
So we can get gains more quickly. If we can accurately predict say how the next generation is going to perform or how a given crop variety might perform in a new environment where we haven't tested it before.
Amy Quinton
So now, let's say we've got what we think is the best tomato seed to plant and let's say in the Sacramento Valley, because that's where most processing tomatoes in the United States are grown.
Andy Fell
And let's say it's been bred to be both drought and heat tolerant.
Amy Quinton
And that's also important in Sacramento Valley,
Andy Fell
And to be highly nutritious.
Amy Quinton
Perfect, until the pests invade the field or bad weather makes it the perfect breeding ground for mold.
Andy Fell
It can be unpredictable but on the farm growers can use AI in all sorts of ways to improve their crop. AI-powered robotics can help growers apply the right amount of pesticide, herbicide or fertilizer to get the highest yields.
Amy Quinton
Assistant Professor Mason Earles leads the agricultural production cluster with the AI Institute.
Mason Earles
We use a lot of cameras and other types of sensors on these robots. And those really need to process in near real time, what they're seeing and be able to do something such as remove a weed that we don't want in the field. Or you could even think of spray a certain amount of nitrogen where it's needed, but not where it's not. So we could be using resources more efficiently or producing foods more effectively. That way.
Amy Quinton
Think this is just really futuristic? Think again. Mason says growers are already using this type of computer vision in the field. John Deere is even advertising it for row crops.
Commercial
See & Spray ultimate uses proprietary camera technology that was developed by Blue River Technology, a subsidiary of John Deere. So it uses computer vision and machine learning to scan the field constantly for weeds from crop and then sprays only the weed.
Amy Quinton
Mason says adapting this AI technology to work in different types of specialty crops. The fruits, nuts and vegetables that we grow in California is the next step.
Andy Fell
So I can take different kinds of data in the field and combine them. For example, video cameras on tractors can capture images of plants.
Amy Quinton
Satellites can gather data on soil moisture. And then there are drones. Armed with the right technology, they can capture height, area and volume of each plant and capture spectrums of light that humans can't see.
Andy Fell
And all of that data could be combined with weather data and historical yield.
Amy Quinton
And AI can then help growers better predict yield while minimizing water and fertilizer use. Mason says AI can give growers a step up.
Mason Earles
The challenge that they're dealing with is there's a lot of uncertainty out there in any given year, right. And so it's sort of giving them a superhuman suit to be able to go out and do the things they're doing better.
Amy Quinton
So now that our highly nutritious drought and heat resistant tomatoes have been harvested, they head to the processing plant.
Andy Fell
Well hold on Amy, because AI can be used before the tomatoes even get through the doors of the plant.
Amy Quinton
You're right.
Andy Fell
Professor Chris Simmons, the chair of our Food Science and Technology Department has used AI to try to predict the likelihood that a loaded tomatoes could get rejected at the gate.
Amy Quinton
He says sometimes those tomatoes can get bruised, moldy or squished. And that's a waste.
Chris Simmons
It's a concern for tomato processors. And it's a concern environmentally because each time a load of tomatoes is rejected, it's not just an economic loss. It's also a loss of all the resources that were used to produce and become embedded in those tomatoes. So all the water, the energy, the fertilizer, other inputs are lost.
Andy Fell
Chris and other researchers at UC Davis and UC Berkeley looked at years of data on every load of tomatoes including things like variety or type where it was grown and what time of season.
Amy Quinton
They combined it with weather and climate data to train different AI algorithms to try to predict when a load of tomatoes would be rejected for mold.
Andy Fell
And they did find some weak predictors.
Chris Simmons
Those would be the time of the season and the acidity or the pH of the tomatoes themselves.
Amy Quinton
So the models worked. But Chris says not in an ideal way.
Chris Simmons
They were predictive, but not with a specificity that would really allow anybody to say, well, this load coming from this location at this time of the season is really at a very high risk of having mold and maybe we should not even bother,
Amy Quinton
Which just proves your point, Andy, that you made at the beginning of this episode about ChatGPT.
Andy Fell
It's all about the right inputs?
Amy Quinton
Yeah, or in this case, more and better data.
Andy Fell
So now these tomatoes have made it to the processor where the enter a high tech, highly automated environment.
Amy Quinton
That sound you're hearing are tomatoes, going through a flume system up an elevator to be steam peeled, and eventually through a color sorting machine at Pacific Coast Producers cannery and Woodland, California.
Andy Fell
So the color sorting machine uses a line of cameras and sensors that can reject green tomatoes with little mechanical fingers.
Amy Quinton
It's an AI that's been around for decades, says former CEO Dan Vincent, who's also a UC Davis alum.
Dan Vincent
And you can't see the product going through it's going so fast, and it's making decisions based on density based on size based on color in a split second.
Andy Fell
Pacific Coast Producers sales about 200 different product types, not just canned tomatoes and paste at the Woodland plant. They also sell sliced peaches, pears, fruit cocktail, maraschino cherries to every retail and food service distributor in North America.
Amy Quinton
And all those products get labeled at the company's plant in Lodi about an hour away.
Dan Vincent
But that adds a huge amount of complexity. So you can say okay, we have dozens and dozens of customers and 100 or 200 SKUs.
Andy Fell
An SKU is a stock keeping unit in industry speak. They label the same products for many different customers.
Dan Vincent
So you can take that and say what are my data intersections. And for a company like ours, it's 14,000 theoretically unique items you have to manage.
Amy Quinton
Dan sees the next frontier of AI helping producers label all the different products, while minimizing the number of times they must pause the equipment for a new label. Because that's where you lose efficiency.
Dan Vincent
Those 14,000 unique items, I can look at history, I could lay on analytics or algorithms for seasonality, for promotions, for new items, for store count addition, and apply all that to a forecast of demand. We could then use those tools at that level to be more intelligent when we case and label to get bigger runs.
Andy Fell
AI Institute for Next Generation Food Systems at UC Davis is hoping to work with companies like Pacific Coast Producers to research how AI could help improve food processing.
Amy Quinton
So now our tomatoes have been canned and are on my plate. But we're not done. You might be asking should I eat this? Is it healthy? Well advances in AI will make knowing what is nutritious much easier.
Andy Fell
Right now nutritionists track what people eat by making them fill out questionnaires or keep food diaries. But that's pretty tedious, says UC Davis Adjunct Professor Danielle Lemay. She's also with the USDA's Western Human Nutrition Research Center.
Danielle Lemay
And not particularly accurate either, um, memories fail. And with the food frequency questionnaire, you kind of have to do math in your head to figure out well, how often do I really eat that over the course of a month.
Andy Fell
So she thought maybe AI and computer vision could help us track nutrition better.
Danielle Lemay
Wouldn't it be great if we could just use our phones to capture what we're eating in order to better inform dietary assessment.
Amy Quinton
Danielle is working on an AI app capable of predicting health effects, just from a food photo.
Andy Fell
There are already AI apps that can estimate calories and nutrients of a meal from a food photo. But this will go much further.
Danielle Lemay
We want to know even more about your food. What the nutrients are, is just a piece of it. But there's this whole world of thousands of molecules in each food that we're starting to map with analytical chemistry.
Amy Quinton
The goal is to eventually come up with a detailed list of ingredients and micronutrients, their health effects and interactions. It would mean combining and analyzing databases about food to databases about molecules, and how those molecules impact our health. It's a bit dizzying.
Andy Fell
Just consider all the molecules that make up the ingredients of our food. Most of us them are still unknown.
Amy Quinton
Ilias Tagkopoulus says we only know about 2% of what's in our food. It's referred to as the dark matter of food. And it can influence metabolism and physiological responses in the body.
Andy Fell
And then just consider how different we are as individuals with age, weight, genetics.
Amy Quinton
It's hard for me to get my head around this.
Andy Fell
But it's just the kind of big data problem that AI is good at solving. And if we get there, it'd be like having a personalized dietitian by your side, helping you make better healthier food choices for your needs and goals.
Amy Quinton
There is one last and very important piece of the AI in food systems puzzle, and that's integration.
Andy Fell
Yes, as AI becomes routine at different steps of the food system, these steps can themselves be integrated by AI. For example, information gathered about crops growing in the field can be combined with AI for plant breeding.
Amy Quinton
Integration is a key goal of AI Institute for Next Generation Food Systems. Again, Danielle Lemay.
Danielle Lemay
Within different parts of the food system, there's many interconnecting loops. And if you only think about one of them, and you're siloed, you'll not be able to solve the whole food system issues.
Andy Fell
And that's important because AI could solve many of the food systems most challenging problems. Ilias Tagkopoulus says AI holds promise for a healthier, more sustainable planet.
Ilias Tagkopoulus
It will allow us to be more efficient in the way that we grow food and we consume food. And that means lower cost, lower resources and better, healthier food for us.
Amy Quinton
That's not to say AI comes without issues. Ilias says regulation but not overregulation is key to making sure AI is ethical, trustworthy and equitable.
Andy Fell
Sure, but you know what Amy? We'll probably find in a few years that AI fades into the background.
Amy Quinton
Huh?
Andy Fell
Yeah, AI will just be crunching big data in the background, helping breeders and farmers and processors and consumers to do what they want to do more efficiently. We won't think about it any more than we think about why there's a label on a kind of tomatoes.
Amy Quinton
I bet this can of tomatoes I used in my pasta came from that Woodland plant.
Andy Fell
Quite probably.
Amy Quinton
Does it say it on the label?
Andy Fell
I don't think so. I don't know.
Amy Quinton
See, just thought about a label on a can of tomatoes.
Andy Fell
Well, that wraps up this episode. You can learn more about Artificial Intelligence Institute for Next Generation Food Systems on our website at ucdavis.edu/unfold.
Amy Quinton
And there you'll find links to a longer story that Andy did on AI in the food system. Plus cool AI generated photos and videos, and even this episode transcribed by AI. I'm Amy Quinton.
Andy Fell
And I'm Andy Fell. Thanks for listening.
Amy Quinton
This entire episode was written by ChatGPT.
Andy Fell
Kidding. Unfold is a production of UC Davis. It's edited by Marianne Russ Sharp. Original music for Unfold comes from Damian Verrett and Curtis Jerome Haynes. Additional music comes from Blue Dot Sessions.