The Rise of the Computational Breeder: Rethinking How We Grow Food

What happens when crop science becomes computational

In this episode, Jesse Hirsh sits down with Mohsen Yoosefzadeh Najafabadi, Assistant Professor at the University of Guelph, to explore the emergence of the computational breeder: a new kind of agricultural scientist working at the intersection of plant breeding, data science, and artificial intelligence.

Grounded in his work on dry beans, Mohsen walks through how breeding is evolving from a largely intuitive, experience-driven practice into a high-dimensional process shaped by genomics, phenomics, and multi-omics data. But this isn’t a story of replacement. It’s a story of integration—where traditional knowledge and computational tools begin to inform one another in new ways.

The conversation traces the shift from predictive models to generative and hybrid AI systems, including Mohsen’s development of BeanGPT, a tool designed to make complex agricultural knowledge more accessible to researchers, students, and practitioners alike. Along the way, they examine what it means to translate advanced research into real-world farming decisions—and why accessibility may be as important as innovation itself.

As climate pressures intensify and the demand for resilient crops grows, plant breeding is becoming one of the most critical—and least visible—sites of transformation in the food system. This episode offers a grounded look at how that transformation is unfolding, and who it’s ultimately for.

This is a conversation about seeds, systems, and the emerging intelligence shaping how we grow food.

Hi, I'm Jesse Hirsh.

Welcome to The Future Herd.

Today we get to a talk.

Today, we get to talk about a topic close to my heart.

Artificial intelligence and agriculture, and in particular programmatic breeding,

otherwise known as computational breeding.

That's where my new friend Mosen and one of the pioneers in the field not

only helps us understand the way that AI can empower farmers, researchers.

And policy planners, but also the courage, the learning, the curiosity

required to be a leader, not just in agriculture, but in technology as well.

In today's subject, we're gonna get into the weeds in terms of things

like Bean, GPTA, custom AI for plant breeding, in particular dry beans.

But we're also gonna talk about some of the challenges, some of the

opportunities, some of the developments when it comes to mashing together some

of these powerful tools from data science and machine learning with the wisdom.

And intuition that we find in agriculture and AgriFood.

I've been waiting a long time to talk about AI in the concept of the

future herd, and in that regard, Mosen does not disappoint at all.

I did my best to try to limit, uh, he and my potential to nerd out on

these very interesting subjects, and I think we perhaps accomplished it well.

Maybe giving you some key words that you might search to learn more.

This is really just the tip of the iceberg when it comes to

exploring how this technology is shaping the future of our sector.

And we will definitely have future episodes that get into it.

So think of this as both a kind of introduction and overview as

well as a specific case example when it comes to breeding beans.

Alright, let's listen in.

Mohsen, welcome to the Future Herd.

Thank you so much for having me, Jess.

Now, I gotta say you're our first guest in this podcast who I

can finally talk to AI about, so I'm particularly excited about that today.

Uh, but the first question that I throw out to every guest as a kind

of abstract, what's on your mind, what does the future mean to you?

Oh.

That's a great question.

Basically AI and future, huh?

you, you know, to me the future depends on which part of life we are talking about.

So I can.

Explain future in different aspects of my life.

So let's say in my professional work, the future means planning, improving

and trying to stay one step ahead.

No, not at least, you know, or, or at least not two step behind.

You know, this is something that I can just articulate that.

future in my professional life is about learning from past.

And gain much experience and prepare ourself for the future.

So if I can do this today, I can make my future better and perfect.

Right?

So I like to run and something people like sometimes tell me about

the future and they are sometimes getting scared of these fast paced,

changing that happening these days.

But you know what I love?

I love these kind of fast paced changes and I wanted to embrace it.

So this is.

The meaning of the future in my, uh, actually professional life, but you know.

In a broader sense, like politics or society, if you ask me about it,

you know, I feel that future looks a bit like the past because history

has funny, like, you know, has a funny habit of repeating itself.

So, or even in my personal life, if, if you ask me about what does it

mean, what future means in my personal life, it means opportunity because.

Finally I can lay down and let the future, let the robots and everything,

do my, do my chores and every other things, you know, but yeah.

So that's, that's a different meaning, you know, that I have for the future.

Yeah.

Right on.

Fantastic.

And, and I think you, in that answer kind of gave us hints as to why AI

and robotics is an area of interest.

But let me ask the inverse, uh, why agriculture, why plants?

Uh, what was it about that side of science that not only did you say, Hey, let's

see how AI and this kind of connect.

But these days, AI, people like yourself, people who have that

curiosity are in high demand.

You could have picked any sector, you could have picked anywhere to focus.

Why plants?

Why agriculture?

Why beans?

Wow.

No, that's a great question, Justin.

The point is the time that I decided to work in agriculture was only because

of one reason, and the reason was.

I grew up in a farm and I used to go to the farm with my dad, and I remember,

you know, all the time my dad, you know, we're so happy at the farm and

he tried to just make me happy on the farm and just ask me that, Hey Mo,

do this, do that, help me with that.

And be honest.

On that time, I hated to do that because I was just thinking,

okay, why should I do this?

Right?

I, my passion was like, you know, novel reading, novel books, playing

music, those kind of things was not actually into the agriculture.

gradually I realised that everything that we are doing

are rooted in agriculture, food.

Uh, even our future, uh, depends on the agriculture.

So if I

might,

to survive and read novel book, if I wanted to play music, I need to

have something to eat and this is coming from the agriculture, how can I like

I'm making work.

that there are, there is enough food in the future.

You know what, that was the simple question that I've

tried to answer on that time.

I started to do undergrad on horticultural sciences and.

In the horticulture, we are working with a very small scale rather than like

working on the large scale field area.

So on this small scale I was working on watermelons and Q Cortes, like,

you know, cucumbers, watermelons, and.

I've, I started to just enjoy actually going to the farm because I, on that

time I had this kind of like, you know, things to answer in my mind and I had this

kind of goal, so I went to the field on a stick, you know, made my hand dirty in

the field, and I just realised, okay, how cool it is to work on the agriculture.

And then later on.

I started thinking, okay, what can I do to speed up this productivity?

What can I do to ensure the food security in the future as a graduate student?

What can I do?

How can I play an important role in this area?

And I, I was just thinking, and I found that, okay, maybe I, we all gathering

all the data, we all collecting everything, how about just the get

the best juice out of whatever we are producing in terms of the data.

And then that was something that I started to dig more into the statistics

and then gradually machine learning.

And then when I get into the PhD and I was started to work on a

larger scale field on soybean.

Which is the agronomy like, you know, crop.

I've realised, okay, now it's time to apply those things that I was just

imagined in my, like, you know what?

Now I can make it applied into the actual agriculture setting.

that was the time that I started to apply these kind of machine

learnings and like AI and more sophisticated statistical algorithms.

Now, before we kind of get into your particular research, tell

me about, for lack of a better phrase, that the ecosystem of researchers that

were alongside you as you did that.

Because on the one hand, you are breaking new ground, right?

Digitally and physically, and, and, and I'm curious.

You know who you were looking for in terms of inspiration, right?

Because you, you're kind of on the one hand inventing it as you go along.

But on the other hand, it is an exciting industry.

It is an exciting research ecosystem.

So where were you drawing inspiration from?

'cause part of you answer was your dad, right?

And like, getting you back on the farm and thinking.

But, but intellectually, scientifically, where, where were you getting clues from?

Where were you getting inspiration from?

Yeah.

You know, before I answer this question, you know what?

that I started to work on machine learning, it was not as easy to

say, Hey, I'm computational plant breeders, because I remember the first

time that I. Remember, I remember the first time that they gave the

presentation, were just telling me, okay, Mo, you are living on your dream.

It's not going to happen.

Like, you know, you are just flying drone.

Take some pictures using neural networks.

Ai, what are you doing?

You know, just do some real work.

You know, do B like a actual plant breeders.

And I remember when I even graduated, people called me, okay, Mosen

is the computational biology.

It's not actual breeder.

Because they had this kind of mindset, and you know what?

They were right.

You know, on that time, you know, no one was predicted what happened in the future.

And you know what?

Sometimes.

What I'm trying to do, I'm just trying to, I, I love to read novels.

I love to read, you know about people's work.

I love to, I love to read and I read, so read so many books, so many

journals, and one of the people who inspired me a lot is Dr. Rex Bernardo.

So he's a plant breeder and I love reading his journal.

He, his, you know, uh, papers, I love to attend his seminar

and he's one of the people I, I usually follow as a plant breeder.

And that was just, you know, I had a chance to talk to Rick Bernardo a

couple of years ago and I asked him that he, if you, if he's available

to meet me in person, and he did so.

Like, you know, sorry, meet in virtual, and he did.

And we talk about so many different subjects and I found

that okay, we are very close.

Sometimes we are trying to adopt new technologies, but sometimes we are

going one step back and thinking, okay, is it really a good technology or not?

If it's good, continue and then.

How about adding this?

How about adding that, right, to make it something better.

right.

So that was something that inspired me to work in this area.

Another one is I love sometimes, you know, to, not sometimes, but most of the

times I'm trying to challenging myself.

And when I'm seeing that, you know, people are just asking me,

Hey Mo, why are you doing this?

And have this kind of like, you know, things in my mind

that, okay, I'm doing right.

I wanted to work and make it applied and finally let people know that,

hey, is something that I told you it works and now it's working.

Right?

So that was, and we call it self inspiration, you know, so just had these

kind of things in my mind to just follow whatever people call it as a dream.

I wanted to make my dream true true.

So, yeah.

And, and the concept of the computational breeder

is revolutionary in of itself.

And, and I empathise with the idea that when you're too far ahead of

the curve, you appear as a fool.

And it's not until after the fact that people respect your wisdom.

So, so.

We are now at that point, thankfully.

Uh, but I imagine it was a bit of a difficult path for you to get there.

But the paradox is, to your point, we seem to be at a moment where

people are starting to accept that our entire world is programmable.

That there are elements to our world that we can engage with on the

level of code and of programming.

So take a moment to unpack what.

Programmable breeding is 'cause I suspect all of the listeners right now, unless

they've already Googled you while they've started the show, have never heard that

phrase before and are still wrapping their head around what that might be.

No, that's a great question.

Just so the point is, uh.

I can just, you know, make an example of what I'm doing

here at University of Guelph.

So I'm working on two different areas, driving in breeding

and computational biology.

These two subjects are very far from each other at the beginning, but

at the end they are super close.

So in the traditional breeding, what we are doing is we are going to the

field and try relying on our breeder's eye, seeing if the plant is good,

they has a good uprightness, you know, it's like, you know, looks healthy.

We are, pick it up.

And then grow next year and a year and a year.

And we try to just, you know, gather more data, more information.

those are usually done manually.

And it's, we are significant.

Rely on our guts, on our feeling.

Right.

Which is perfect.

Which is good.

I love it.

But at some point.

We are adding, we need to add more information.

We need to see plant from different perspectives.

Like how about flying drones?

Take some picture from above and see if we are able to find something that

we couldn't see by the naked eye.

Right?

Or how

Not

you know, dig into the genetics of the plants?

Or how about working on the pure science part?

Of course we are able to do that is perfect.

But what's next?

So the next is.

We are having tonnes of data, tonnes of information, and

it's really hard to digest.

what would be next?

is to select the best and appropriate algorithms, right?

So we try to use ai, different types of AI and all those kind of things to digest

all the data that we are collecting.

But what next we get to the result?

Okay, that's perfect.

again, what next?

is we can use this result as to initiate new population and work on the same

pipeline and initiate a new population.

So if you are seeing, this is something that is started to do, like, you

know, as a programme, running as a programme and in each steps is

Is the.

of tradition the innovation.

With the coding and those kind of things, right?

All are linked, chained together this is the future.

I think, you know, to my knowledge, this is the future of the breeding and this is

what we call computational plant breeder.

I published a paper last year on Journal of Plant Physiology.

And I've just, you know, sometimes criticise actual my position because

my position is, uh, driving breather and computational biologist.

And I just mentioned exactly my position and my title in the paper and I said, Hey.

E, everyone can call me computational plant breeder instead of plant

breeding and computational biologists because we are using computational

biology to programme plant breeding.

So how about calling a person computational breeder?

Like you know, myself, I can consider myself as a computational driving breeder.

So anyway, this is the programmable breeding pipeline look like.

I am, uh, to the listeners, I, I am gonna do my best to resist nerding

out today and, and asking technical questions that indulge my curiosity,

but I'm gonna have to ask one right now.

One of the big things going on in the AI space are simulations and, and.

Mass simulations.

To what extent can you simulate some of the cycles that you described?

Because on the one hand, there's the practical, right?

The field tests and seeing how the plant performs in nature and

monitoring that data and repeating it.

But I have to assume you're also running simulations.

To come up with scenarios and hypotheses and other stuff to test.

I, I, for the non nerds who are listening, uh, describe the value that that plays

and the potential that that has for, for agriculture as a whole in terms of how

we've historically been limited to nature cycle, but now we have access to things

like digital twins that allow us to kind of simulate what a field is gonna do.

Yeah, I was going to touch about digital twins as well.

I'm so glad that you touched it.

So basically the point is, you know, uh, let me start it in this way.

What we are, what we used to do is just we grow thousands of individual

plants in the field and which with the hope of selecting one of them,

right, and the one superior of them.

Like let's say I'm growing around 200,000 individual plants and at one stage of

my breeding programme, and then I'm trying to shrinking down these numbers

that at the end, like after eight, 10 years, I'm reaching to one best.

So you can, you can see that how many plans here are sacrificing.

And why such these kind of things happen is because we wanted to

test it on different environments.

We don't have a good simulation analysis in the, in the like indoor to just prove

that what will happen in the field.

Right.

And we can, instead of growing 200,000 individual plants with

like spending so much time, money investing, so many things there.

How about running a good simulation and just shrinking down to a thousand?

Right.

So this is something that people, when

who

designing the jet and Gene, you know, they are doing that.

They

don't.

just, you know, let people just flying with the jet and gene

and see what will happen, right?

They, they are caring about people's life and so we have to care about plans.

Right in this case.

So the terms of the digital twin is something very new.

Very good.

And I think that's the future of the breeding programmes.

They need to run digital twins.

And we have it.

have it at University of Guelph.

What I'm doing is I'm trying to simulate that what if, if these plant with these

gene characters, with this genotype, like, you know, with the genetic background.

What if, if this plant exposed to these type of stresses, how the plants look

like, how the plants perform, right?

how plants perform.

If you are putting this in totally new environments and with

this soil type, these kind of, these are kind of simulation.

So recently I, one of the project that is going on right

now is my, in my programme is.

We try to predict what will happen in the future in different environments,

that they are mostly grown drive-ins and how we are doing that.

We actually have like the good history of growing drive-ins in those

regions, like El in, uh, in Ontario.

We are growing like, you know, in Woodstock in different areas, right?

We are growing drive-ins, different market classes of drive-ins there for many years.

So we have the data.

We have the environmental information like the total precipitation,

maximum temperature, minimum temperature, all those kind of things.

We have it.

what we have is we have the future scenarios.

So there are some anthropogenic climate change scenarios that is

available for scientists to use.

And this is based on different, like, you know, future, like different

scenarios that what if CO2 emission went high and goes up like, you

know, higher than current, right?

What will happen in terms of the temperature?

If the situation is normal?

What will happen if the situation is mild?

What will happen?

have all this environmental information year by year for the

next 50 years different locations.

Of course, it's not accurate, it's not a hundred percent correct,

but it can give us a good hint.

So if we developed a programme, everything based on the history, and then we feed

the future data there, we can predict how much this plant perform in new

environment with the new temperature.

So once we find it, instead of just blindly making cross, not blindly,

just making cross and selecting potential parental lines based on the

history, we are selecting patterns based on history and the future.

You know what?

So that's the main concept of digital twin.

And we can run it, we can test it, and we are doing it at, at the same time.

Yeah.

Right on.

And, and, and the capacity, like just the research questions alone, that, that kind

of thing enables, kind of blows my mind.

But let, let me come back to something we touched upon earlier.

Are you still blazing this trail alone or are there other people catching up like.

Give me a sense of the computational breeding landscape.

Are there other people who are clueing in to what MO'S been up to and saying,

Hey, that looks like fun stuff.

And what is the collaborative space like?

Like who are you able to work with?

Not just at the University of Guelph, but yeah, I assume that the more people who

are kind of ready to play, ready to dance, ready to engage in these types of methods,

the greater the value for all of you.

Oh yeah.

I have a very good colleagues, you know, all around the world.

Those are perfect.

You know, I, I'm proud of the community that I'm working in terms of the academia,

I don't want to talk only about University of Guelph, but you know, that, you

know, university of Guelph fostering collaboration, like, you know, and they

are fostering collaboration environment.

So we are having a highest collaborations between different

departments and others, but.

Aside from University of Guelph, I'm right now collaborating with different

researchers in Canada, like University of Laval, Ottawa, like, you know, I'm working

with Saskatchewan's, but, and also us like, you know, entire north, uh, America.

I'm collaborating mostly with some, there are, there are tonnes

of, you know, good breeders that they are working in this area.

And you know what?

All of them, like, you know, I, I had a chance to talk to a couple of them

and I know that, you know, we are somehow complete complement our jobs.

Like they are working, I'm working on these, uh, developing new algorithms

and everything, and they are using all the algorithms, like most of the

algorithms that we are developing here to test them on different locations.

And I'm doing the same thing.

So we are just trying to validate our algorithms.

We are trying to validate our pipeline, right?

And.

We are not only looking for the very small scale of the regions that we

are working, we wanted to expand it.

Right.

And right now I'm collaborating with some universities in France,

some universities in the uk, right?

Those peoples are also in interested, and you know what?

remember when I developed some of the algorithms and I published

these algorithms, I received very good and positive feedback.

Said, Hey, how you did that?

We are, we wanted to test it in our genuss.

We are working on grape.

Do you think that the algorithms that you develop into ripens with working grape?

And I said that, oh yeah, why not?

Let's check it.

Right?

And we check it.

We fine tuning all those algorithms and now it's work.

Right?

It works.

So basically, yeah, they are very collaborative.

Uh, environment all around the world.

We are, I, I, I can say there are tonnes of us, you know, to working in

this area, but there are a couple of us, some of us you know, are working

in this area in the computation as a computational plant breeder.

Yeah, I.

And at the risk of asking another nerd question, which, you

know, we will do our best to try to answer as accessibly as possible.

Talk to me.

You know, one of the, the research interests you have on your profile talks

about computational tools and AI platforms in that the, the subtext of everything

we're describing is there, you know, you, you need technology to gather data.

You need technology to analyse data.

Course the digital twin stuff is a whole other scale.

Um, without getting into the details of it, how accessible are these tools?

How available are these tools and how rapidly are they being developed

to your point of, you know, there's some computational breeders, but

how do we scale that number up?

Uh, in the sense that part of it is just access to the tools and access

to the literacy to use those tools.

Uh, no, that's, that's a very great question.

So basically I remember, uh, like, you know.

Eight to nine years ago, I was just focusing on developing different

algorithms to use whatever data that we are producing to make a prediction

and can help us to screen large number of genotypes, not large number of

actual plants, you know, in the field.

And we are able to detect diseases at the earliest stages, those kind of things.

And that was all limited to the prediction ability.

So we call them a predictive AI algorithms and there are tonnes of them, right?

Gradually we shifted from only working on the predictive to also

working on the generative ai.

And what does this generative AI means?

It means that, how about we are asking question and we are

getting an answer response by ai.

So AI can generate response for us.

So how AI can generate response.

It can be based on tonnes of things that we are feeding ai, right?

Right,

And I

and I.

thinking, okay, what kind of information do we have that I can

develop this kind of generative ai?

And I had a little search, I just wanted to share it with you.

So in on the year of 1970, there were around 12 to 15 publications on drive-ins.

Right now in 2026, actually 2025, like last year, was around 2,900 paper

that were published in Drive-ins.

you'll see that you know how much.

Scale were large, you know, right now is large.

The same for the conferences I've searched, or even the same for the word,

the actual driving word over the internet.

On 1970, it was 120 times the word of driving was repeated over the internet.

Right.

it's more than 6 million times.

So why this word is getting repeated?

And I've searched and I found out, oh yeah, there are tonnes of valuable

information there over the internet.

Even without there is a need for us to produce more data.

have tonnes of data.

So the challenge from the past last eight years, I think it's changed.

Changed from do we have the data to how we can find reliable and

relevant answers quickly, right?

Those type of things.

So on that time, I was just searching to see if there are any

like pipeline established that I can use it or if there is any Interac

inter integration, I can make it.

I couldn't find it.

I couldn't find it, started to do it and build everything from scratch.

So I started to develop something like a app called right now

is called it being GPT, right?

So this app being GPT, we collected more than 300 thousands indivi, like, you

know, literature all over the world.

And we indexed all of them.

We collected all the genetics.

All the phenotypes, all the agronomy, all the notes, everything that

was published in Drive-Ins and put it, put it in the being GPT.

And you know what, when you are right now asking question being GPT

search and finding relevant answer based on only these information, not

based on the general, like the chat GPT or the Clause and others, right?

So.

So

How

how we.

information.

It was really hard because, you know, so, so many of this information

were not publicly available.

We had to ask, you know, and so many things.

But right now I'm, I'm, I'm seeing that, you know, these type of retrieving in

information and gathering information is getting facilitated because

people right now understand that.

Importance of sharing these type of information and building

such as kind of apps, right?

So that was one of the app, the predictive AI was another app, right?

And now the challenge is how to make it hybrid, how to have the prediction

and actually validate whatever we are predicting by tonnes of information

out there through let's say Bing, GPT.

And this is the hybrid ai, right?

So.

I don't want it to go through the details and talk about all these type of things.

You know, I don't want it to be nerded here, but, you know, I, I, I'm getting,

I'm getting fascinated every time that I'm talking about AI and I can

talk for hours, but you know what?

Right now, the future of, uh, I think the future of, uh, ai, not ai, actually

AI in agriculture is how to make a hybrid ai, like the quantum hybrid ai.

And this is something that we are trying to go to toward

it and move toward it, right?

And I am gonna have to have you back on the show because I

too can talk about AI for hours and I think we're gonna at some point

organise a panel in which we together talk about AI and agriculture.

But I wanna a ask an important follow-up question around being GPT

from, from a kind of user perspective.

Because one of the, I, I think the appeals of like chat, GPT and clawed.

Is it makes complicated stuff accessible, right?

It makes it easier for people to talk about law or other complicated

areas and learn the terminology.

Does Bean GPT enable, say, bean genetics to be more accessible to a

student agronomist or a bean farmer who maybe has mixed scientific knowledge?

Does it enable that kind of, uh, accessibility in the kind

of learning curve to get people into computational breeding?

Absolutely.

Absolutely.

Jess.

You know, the point is that the main point of developing

BGPT is to make it accessible.

For a variety of people with different background, with different

knowledge, and it's not only designed for myself or like, you know, and

any other person specifically.

The point is when you are asking any question in being GPT, being GPT, try

to answer fully in a very simple way.

Let's say you are interested in a specific, like, you know.

Agronomic practise in dry pins, and you are asking this question,

okay, what would be the best herbicide in this case and when I

can apply this herbicide in my field?

So being GPT is trying to find relevant research and then try to make it like,

you know, uh, seem not the simple, but just make it more understandable and.

Provide the answer to you.

I can tell you that, you know, being GPT is not one app, is not one algorithm.

It's a kind of committee that consists of eight different algorithms.

We call it rag models, like we call it retrieval, augmented, uh,

generative, like, you know, models.

So these kind of models, each of them are trained for a specific.

Data, one of them trained only for literatures.

One of them is going to only search for the data.

One of them is going to only search for the agronomy.

One of them only going to search for the environmental data.

So many information when you are asking question, all these

people are discussing together and

And they.

threw out some of the words, some of the information, and

who collect all this information.

Is the open AI is the G-P-T-G-P-T collect all this information, make it simple,

make it ready and give it to the user.

So no matter if you are expert in the BG beam area, no matter to what kind

of knowledge, to what knowledge you have about being, no matter that what

you are going to search like in terms of, and you wanted to get BGBT, try

to provide you as simple as possible, as an understandable as possible.

Now I, I keep wanting to go to the environmental scan

questions, but do you know of other, uh, uh, uh, similar plant-based GPTs or

similar kind of agri centric, uh, gpt.

I've, I've seen a couple, actually, you know, they just recently

released a couple of GPTs, but.

And, but you know what?

I haven't had a chance to dig into all of them in detail, but what

I found that, you know, they are specialising in couple of data databases.

Like one of them are only looking for the literature, right?

Another gpt is only looking for some, like, you know, reports and

data and I couldn't find that.

You know, there is one, uh, like, you know, large language model.

it's not one, actually, it's a couple that each of them are specialising

on different algorithms, different database, and then when you are asking

question, your question is going to circulate among all these apps, all

these like, you know, algorithms.

So it's not one, and I couldn't find the same and the similar things in others.

Right.

I mean, this is where I will nerd out for a moment and say, this is

partly the revolution of open claw and the harnesses, right, of creating a software

layer that can access multiple models and combine that with a growing memory.

We're digressing.

We'll talk about that in the advanced episode.

But o on a related point, uh, as someone who I assume also

teaches a and also engages with students, tell me what's that like?

Like what are the students who are coming to you, what are their research interests?

And as a teacher, what are you trying to.

Get those students to focus on.

'cause so far I think I, I, my mind has been blown just thinking

about computational breeding and the kind of research you're

doing and the tools you're doing.

But part of what we do here on the future Herd is think about the future.

And of course, that's where the students come in.

So give us a glimpse, get a, give us a sense of where their

interests are and, and what you're trying to get them to look at.

Yeah, so actually, you know, every winter semester I'm, I actually

developed a course at the plant agriculture department at University of

Guelph, and the course is mostly about, you know, in intro to computational

biology and agricultural science.

So I'm trying to teach different s we call it like different.

Words like, you know, phenomic genomics, transcriptomics, proteomics, all

this kind of word, like big words.

And how we can integrate all of them through different AI and through different

algorithms and how we can, uh, find the best, like, you know, the actual, uh,

background behind any traits of interest that we are interested to work with.

So.

This is one of the course that I developed for the graduate students and also for

the undergrad, specifically for the second year undergraduate students.

I was just thinking, okay, how about letting students to make their hand

more dirty and dirty into the data?

So I develop bio data for people, plant, planet, and I'm going to teach it for

the first time in this coming fall.

I'm not sure that you know what will happen, but I'm really looking forward

to that and I'm thinking that, you know, it would be a good source for a

student to make them ready for the future courses that they are going to take.

Right?

So, so far I'm like, you know, last year I was able to teach, applied

Bioformatics and how we are using bioinformatic in agriculture.

So all of them, I found that, you know, students are very

much interested use trends.

Like they are interested to work with ai and every time they're telling

me, okay, Mosen, we love to work ai, how we are developing this ai,

how we are working with that ai.

And the first question that I'm asking them is why you wanted to work with ai.

Is it, do you think that you know AI is really right tools for you?

What type of data you have?

Do you think your data is big enough?

Or your data is large, or your data is streaming data or the complex data,

each of them has different meaning.

So don't get interested.

Don't get excited and just brag that, Hey, I'm working with ai.

Of course, working with AI is good, but have you had this kind of

like, you know, good understanding why you wanted to do that.

So this is something that I'm going to teach analytical thinking.

think before selecting the right or the like, you know, whatever algorithms,

whatever path you wanted to choose in your future, just think before that.

Don't just follow because it's trend.

Don't get excited of everything that comes new.

If you are seeing, okay, this is the huge potential, even, you know, the

reason just that's, that's interesting.

The reason that I started to work with machine learning and those ai,

like you know, many years ago was not because on that time it was not

trendy, I was just searching among all the literature and I found that,

okay, there may be a good use of these kind of algorithms in agriculture.

I can't test them.

So I started to

Not test them, and even in home,

that time no one buy it.

Right.

You know, they told me that.

Okay, yeah.

You are living in your dream and you are not an actual breeder.

I

I never gave.

give up.

Gave up because you know, I just wanted to work on that area and I did.

And now you are seeing that, okay, here is the real things right, and

we are working on everyday life.

So if I wanted to just only use whatever is getting trend,

I'm not going to be successful.

And this is something that I wanted to teach my students and I like,

you know, they are at the end of the semester, they are all getting

appreciated and said that, hey, right now we have a good understanding of

whatever data we have, or maybe we can use the simple statistical analysis.

we can get the body where we want it, right?

It's not just only using the fancy ai.

We can use a simple, having a good discussion and understand why and

what is the data and how we can use this data in a better way.

Okay, let me see if I can use a metaphor to set up my next

question, because one of the recurring themes of our conversation today, which I

think you, you just evoked again, is the idea that there's this learning curve.

And to be a good researcher, you wanna push yourself up that learning curve.

You, you wanna challenge yourself and to your point, not

be content to be in the trends.

To really try to go further and anticipate either your own curiosity or where the

world is headed, that's clearly what you're encouraging your students to do.

Where does industry fall in that learning curve?

Right?

In your interactions, not just with big industry, but even farmers,

even PE, agronomists, people who are curious about this stuff.

Where are they on that learning curve and, and how do you do what you do with the

students, which is encourage them to be courageous, encourage them to really try

to imagine where we need to be to make the most of this technology and these tools

Oh

I.

so you know what this is.

This is actually one of the things that I am working every day, and at

the same time, I'm trying to improve my day every day by any interaction that

I have with farmers, growers, industry partners, and also government bodies.

So each of them, you need to make sure that you have, you know, their own

Stack picture,

how to talk with them, right?

It's

it's

something that

that I can.

the rag models and some of the statistical analysis

behind everything to the farmers.

They all just saying, okay, what would be the most benefit of it?

And what I found is farmers are really, is real.

Uh, they are really appreciate something that I can just show them,

Hey, here is something that would work.

Right.

Test them and see if it works.

You can just, you guys can also dig into it and learn more about

it and we can work together.

And you know what, I'm proud to say that I have a very good

supporters in terms of the farmers.

I have Ontario bin Growers.

They are supporting me in different aspects of my programme.

And of the times they invited me to give a talk to actual, to their

a GM, to their research, day, but.

They are not asking me to talk about the actual bean seeds or anything.

Last time they asked me to talk about bean GPT.

Because they are also interested to see what is this being GPT look

like, how they can use this AI in their actual day-to-day work.

Right.

And it was last Friday that I was, that I was talking to a couple of

them and I told them that, yeah, here is the ai here is we are doing, and they

told me that they are advocating for AI in different committee on, on my behalf.

You know what, this is the interaction.

This is the relation that I really appreciate.

I really admire that and I've seen, I've seen such this kind of

openness in different places, but it depends on how you approach them.

Another reason, another example that I can give you about like some of the industry

partners that I'm working with them.

So they are actually appreciated ai, they are actually appreciated.

These kind of trends, things and everything.

to what extent, what, what is the benefits?

If you can show the real benefit to them, that's perfect.

If you couldn't prove it.

couldn't prove it.

Okay, so what's the, what's the point?

Right?

And this is the gap that

that.

I think it's in the, like, you know, research areas that we are

doing so many advanced technologies and advancing things, right?

But when we wanted to translate to the actual, to the applied settings.

We are sometimes like, you know, lagging behind.

So this is something that I am actually learning every day and

it's a learning curve for me.

And I'm also trying to, um, provide this kind of experience that I had, you

know, to my students and everyone who like to work as an industry partners,

like, you know, indu work in the industry or academia in the future.

also the last part is in this case is we are training, whoever

we are training right now.

They are going to build the future.

So if we are able to tell them, okay, here is the research and here is the way

that we are interacting, they can also realise, okay, this is the way that the

academia interact with the farmers, with the growers, with the industry partners.

Once they are, became, become, you know, a farmer, become a grower, become industry.

in the future.

They are realising this kind of interaction.

They are already this kind of like, you know, relation has

already been built, right?

So it can facilitate.

So whatever that we are doing is not only benefiting us for today, it's also for

the future and the future generation.

And this is something that I'm.

So l let, let me throw a relevant follow up at you.

And, and I say this only 'cause I, I've encountered this myself and I'm kind

of curious what your advice would be or, or what your approach would be.

But I think a lot of the farmers that you're describing are kind of

self-selected, curious, and on the one end, you're right, they want see the

tangible, they wanna see how it works.

But I, I think who we should acknowledge, 'cause they're

the people who don't come up.

Are those who are afraid because there are people in our society

for various different reasons who are kind of scared of ai.

I encounter that all the time.

I try to foster their curiosity.

I try to help them see past all the nonsense to look at it

as a tool that can help them.

But how do you approach that?

You know, how do you deal with people who, whether they're in the agricultural

sector or in your family, or people you just encounter on the street and they

know Mosson is a guy who's into ai, how, how do you address that kind of fear?

Because I, I think for some people it's like the obstacle to them learning all the

stuff that we're talking about that I, I think we both think is really exciting.

Oh, that's actually, this is, this is one of the most important

things that we are facing these days, to just make sure, like, you know.

It's at the same time we are certain, certain about some sort of things, and

at the same time, we are uncertain about some sort of things that happening.

So I can tell you an example.

I, I've, uh, so many people ask me the same question, will AI replace our job?

What do we need to do?

Like, you know, these ai uh, I, I can't rely, I can't trust ai how you are

like building everything on the ai.

Not everything.

Most of the things that you are working is, has been built, you know, in ai.

How you do that are, you are not afraid when you are developing

a co breeder ai co breather.

When you are not afraid that these co breather will in the future

is getting to an actual breather.

Breather, and they cannot replace your jaw.

I received these tonnes of questions and you know what?

Up to a couple of months ago, I had not a clear answer for them.

I was

I.

to say, okay, yeah, here's the ai.

Here's the AI would be good.

I don't know what will happen in the future, but if you have the good

idea, if you have the good, like, you know, working in this area, there

are always, you know, job for you.

There are always, you are not just AI is going to help you.

So I was just ask, answering these kind of things until at

one night, over the weekend just got bored, of the watching tv.

So I

If I.

to start back to the working and it, I re exactly remember that

it was like 11:00 AM It was somehow at the midnight asking Claude.

To develop an app that is currently exist, but adding these new informations

and these new techniques there.

I asked Claude that, Hey, Claude, can you develop.

ODM.

There, there, there is a app, like, call it web, ODM.

I don't want to go through the details, but this is for the, uh, stitching

all the drone images together to get the actual orthomosaic picture.

So I asked this, oh, like, you know, Claude, that can

you develop web ODM for me?

And Claude for the, at the first glance, responded, yes, here is the one.

So I check it and I found it.

No, it's not.

I replied

back.

not.

you do this again?

The same thing?

No, that's not, can you redo this?

Redo, redo, redo.

And it was around 5:00 AM after tonnes of back and forth.

I got so frustrated and I asked Claude, Hey Claude, I told

you to create this app for me.

Why you are doing this?

And you know what, what?

Claude responded back to me.

That was interesting.

I respond?

But I understand your frustration, but just to let you know, that web

ODM was built based on human, and it was based built for like, you know,

it's around 500,000 lines of code.

There I am AI and I'm only able to develop 4,000, 5,000 lines of code.

I'm not able to do that.

So you know what?

00 AM I was just thinking, okay, I got the answer.

Will AI replace our job?

It depends how many lines of code we are worth it.

If you are worth 5,000 lines of code, of course you can be replaced.

it's 500,000 lines of code with the novel idea, with tonnes of work behind it.

No, it's not going to happen.

And you are the one who are going to train ai.

You are the one who are going to just, you know, build ai.

So that's the answer I found that.

And it was like, you know, it cost me six hours, not atop like back and forth

with Claude, but finally I found the right answer to like, you know, respond.

Like anyone who are asking me right now, I have something to say.

Yeah, it totally depends on you.

I can't tell you that.

Will AI replace your job or not?

Check, see how many lines of code you are worth it.

Well, and, and the brilliance of that answer is that with

each research paper that you read, the lines of code in your software increases.

Every research paper you write exponentially increases the lines

of code in your software stack.

And this is where I will say to any farmers or drone operators listening,

if you don't know what web ODM is.

Give it a Google, because if you do have a drone, you could do a

lot of fun stuff with Web ODM.

We're digressing.

Uh, we're, we're just about out of time, Mosen, and the, one of the last

questions I ask our guests, this is usually the second last I ask, is

there anything that we haven't talked about today that we should talk about?

And this could be research that you're currently excited about, or

concerns or thoughts that you want.

Other leaders in the sector to be thinking about when it comes to these tools and

it comes to the potential of these tools.

Oh yeah, so you know what I appreciate.

Just, you know, we, we've covered so many different things and

I really enjoy talking to you.

Something that I just wanted to put emphasise on it is.

Whatever we are developing, whatever we are advancing, we should appreciate what

has been built for many years in the past.

We cannot replace our tradition.

We cannot replace whatever have been built.

Right?

I'm seeing that, you know, most of the times that I'm submitting any proposals

or developing new ideas and I'm just talking about it, people often, often

ask me that, Hey, Moss, are you.

Traditional breeder, or you are the new breeder or whatever

you wanted to just use.

You are not using any traditional things.

You are using like new and innovative tools.

And I wanted to say it's a mixture of both.

I have to respect traditional breeding.

Traditional breeding is something that I cannot replace it.

No one can replace it with any new technologies.

We can compliment whatever has been built for many years.

it's the science, it's not practical, it's not based on probability.

It's not so many.

It's, it's pure science.

And you know what we have in the definition of the plant breeding as the

art and science, and it's totally true.

I've just right now realised it with my own, like, you know, full,

like, you know, the point is.

We have to make a best decision our eye, by our feeling, and we have to have, be an

artist to develop and design a good plan.

At the same time, we need to follow science to see what's going

on on the biological background.

What is the biological background behind any traits, how this plan look like,

how this plan perform, get the DNA, other things This should be acknowledged

that plant breeding is actual science.

It's not the practical things that we are doing.

Right, for any plant breeder, our job is not finding the best.

Our job is to find real assets in everything that we are working.

Is, is, is something that I. trying to move toward it.

I'm trying to just realise even a little and tiny data that we are collecting

a small plant, that we are living in the field, what we can best deal

and do, you know, do from like, you know, how we can use of them, right?

So this is something that I wanted to emphasise more about it and just,

you know, let you know that, that, you know, this kind of breeding.

Yeah, like I can't, I can talk on behalf of all plant breeders and

I, and I know that, you know, they all would be appreciated that I'm

telling this, for every dollar that is spending in this area, it's around

35 to $40 we are receiving, like, you know, any government receiving back.

Right?

So it's, you are spending $1 and you are receiving 35 to $40 back.

So these supports need to be continued.

If you are stopping supporting anything like, you know, plant

breeders, it's not showing tomorrow.

It's showing in the next 10 years and that time, there is no way that

you can get it back and turning back.

Right.

I just wanted to give a like $1 to $35 to Doug Miller, the former

executive director of Canadian Soybean.

Uh, um.

CSGA, Canadian Society of, uh, CSGA.

It's, uh, anyway,

Canadian Soybean Growers Association.

No, it is.

Uh, is, uh, I can just search.

CSGA.

It's Canadian Seed Grower Association.

Right on.

Right on.

Is it Canadian Seed Grower Association?

Yeah.

In fact, you kind of previewed what perhaps we should talk

about the next time I have you on.

'cause we've alluded today the entire time to food security.

But due to my fault, we've spent a lot of time focusing on the technology

and the role technology plays.

But we can talk about the role that technology plays around food security.

To your point about the payback that comes from spending money

on researching plant breeding and researching in particular beans.

Um, on the last question, which again, you almost just gotta answered right there.

The last question I throw to guess is who do you look up to?

Who are the leaders that we should be listening to?

It's kind of the shout out section of the show, and it's meant to be spontaneous.

It's meant to catch you off guard in the sense, what's the first

name that comes to your head?

In the plant breeding area or in the computational biology or like,

Anywhere.

This is like the, this is like the first question.

What is the future?

This is the the last question.

Who do you look to for inspiration?

Oh, I'm just looking like, you know, the first name that is coming to

my mind is, Dr. Ricko is coming to my mind and I have, I have tonnes of, tonnes

of people who I really admire them and I just wanted, I just following their

work and everything, Dr. Seth Moray, I'm just, you know, following their

work yeah, there are tonnes I can't really, you know, tell you about it.

And it's not only breeders, they are also.

Some of the farmers, growers, like, you know, industry partners, they

are all people who are following.

Yeah,

And that's why I throw it as the last question 'cause it's not

like it's, you know, the Oscars or the Academy Awards where you can thank all

the people who helped make the movie.

This is about one or two names and you did it spectacularly most.

And speaking of which, this episode, like many of our episodes, has completely

blown my mind, partly because of the enthusiasm you bring to your research.

But also 'cause I really feel you make this stuff accessible.

Like you really help, to your point, farmers have kind of forced you to talk

about this stuff in a way that that gets to the point, that speaks to kinda what's

at stake and what the opportunity is.

So.

Thank you very much.

I mean, I, I really feel that this is, uh, the start, uh, of some conversations

between us to really help people see what you see, which is the potential

for this technology to really not just revolutionise agriculture, but address,

uh, food security at a global scale.

Thank you very much.

Oh.

Thank you so much for having me.

Just the point is, you know, I'm moving toward it to make everything

accessible, to make it everything, like, you know, facilitated for learning

and I'm at the same time learning.

And you know what?

Something that I wanted to say at the end is none of the things that I'm

right now doing would've been possible without a great support my mentors, Dr.

Isman Chens, you know, and also like, you know, uh, some of the

supporters that I have, like the.

Ontario being growers.

Right.

So, and that I just, you know, support, I received the support and crich those

like, you know, supporters are just very like, you know, helpful to build up my

research programme and they are trusting me on just doing all these kind of things.

I would, I, I would like to also thanks them at the end as well.

Right on.

for having me and I'm really looking forward to having

another conversation with you about more exciting stuff.

Right.

Right on.

I hope you enjoyed that half as much as Mosin and I did.

Uh, it's no exaggeration to say that sometimes talking about AI leads

to a kind of time blindness where you don't even realise how long,

uh, uh, how much time has passed.

Uh, while I do make an effort to keep our episodes on time.

Clearly Mosen and I are gonna have to continue this conversation if only

to explore the kind of Venn diagram where traditional agriculture and

some of these new practises overlap.

Because what I found so affirming about Moss's Worldview is it's not an either or.

It's not about embracing the future at the expense of the past.

Rather, it's about using the future to reinforce our values

from the past, to use data and intuition together and in concert.

And that's why accessibility matters.

That's why when it comes to talking about technology on the farm, it's not about.

Companies, it's not about experts, it's about farmers playing with these tools,

seeing how they can bend the tools to their practises rather than expecting

their practises to bend to the tools.

Now before we close, I, I do wanna point out that we've got

an exciting new pilot project.

Called Commons do the future, her.ca.

We're currently solicitating your views about food security, and we

did kind of touch upon that today in terms of AI and food security.

But we are gonna be expanding the topics that we are focusing on, uh, when it

comes to employing the commons, one of which is gonna be the 2050 process itself.

What do you see in the future?

But another is gonna be about AI and agriculture, which includes ethics, which

includes governance, which quite frankly includes any views you may have on ai.

So if you are still listening, that means you enjoy this episode enough to at least

share it with your friends, but also to go check out our website and see what

we've going on, some of the features we've got, so that moving forward.

You can see yourself as part of the herd, even when I press the intro button and

instead I want the outro button so I can evoke that kind of 1970s television vibe

that I'm currently using to set the tone, the motif, the culture of the future herd.

If you have any questions about the podcast, about the process.

About the age agentic software we're billing to facilitate

podcast production, drop us a line.

Uh, hello@thefutureheard.ca or complain or cove about us on social media.

Uh, what did they used to say?

Good news, bad news.

It's all news.

Good attention.

Bad attention.

It's all attention.

And if there's anything we want, it's yours.

I'm Cheezy Hirsch.

See ya.