Latest / The Future Herd / 39: Transitioning Animal Agriculture to Biological Co-Creation with Suresh Neethirajan

Transcript

39: Transitioning Animal Agriculture to Biological Co-Creation with Suresh Neethirajan · The Future Herd

0:01
Jesse Hirsh: Hi, I'm Jesse Hirsh.
0:04
Welcome to the Future Herd.
0:18
As a livestock farmer, I spend a lot of time thinking about how our
0:22
relationship with animals is evolving.
0:26
In fact, I'm happiest when I'm out with my goats on the back part of our property,
0:32
kind of free ranging, and I kind of imagine myself going back in time as if
0:39
I'd been doing this practise for hundreds and hundreds of years for generations.
0:45
Farming has relied on intuition experiences and the five senses
0:50
when stepping into a field or returning to the barn.
0:55
But what happens when you take that deep human wisdom, that kind of ancient
1:00
sensibility that I feel when I'm with my goats and combine it with cutting edge
1:05
ai, bioacoustics, even digital twins?
1:11
Are we looking at a future where technology distances us from our
1:15
livestock, or can it actually help us build a deeper, more empathetic
1:21
understanding of animal wellbeing while still securing the prosperity
1:27
and profitability of the farm?
1:29
To help us deal with this question today, we've got a fantastic guest, CSH Raja,
1:37
a bioengineer and research chair in digital livestock farming from Dalhousie
1:42
University, who splits his time between animal science and computer science.
1:48
CSH is one of those rare researchers who doesn't just look at technology as
1:53
a tool to boost efficiency, but as a bridge to rethink how we care for animals,
2:00
how we manage our operations, and most importantly, protect our food security.
2:07
Again, I feel that we have this real opportunity to use technology
2:13
not as an either or, but a yes.
2:17
And, and there is something really profound in the way in which as
2:22
farmers, we connect with non-human intelligences, AKA animals, and
2:29
lo and behold, we're describing AI as another non-human intelligence,
2:36
which is why I am so enthusiastic, so curious about what happens when AI and
2:43
agriculture intersect and my new friend Ssh, certainly doesn't disappoint.
2:49
In provoking, in moving the conversation forward to not only imagine a a,
2:56
a new world of agriculture, but to actively help make it happen.
3:01
Alright, let's tune in.
3:02
I. Suresh, welcome to the Future Herd.
3:07
Suresh Raja: Thank you, Jesse.
3:09
Jesse Hirsh: Now I have to say I am particularly excited to talk to you today,
3:13
uh, not just because I am an avid learner of ai, but I'm also a livestock farmer.
3:21
And so you're really the first person that I've met who so brilliantly
3:26
connects those two domains.
3:28
But the first question that I ask all guests who come onto the podcast, and
3:35
there's no wrong answer, it's really an intuitive kind of gut question,
3:42
what does the future mean to you?
3:46
Suresh Raja: Like future in the sense of adaptation of, um, uh, artificial
3:52
intelligence or future of the animal agriculture forming here in Canada?
3:57
Uh, yeah.
3:58
It's going to be multifactorial how we can, um, everything is interconnected
4:03
in a way, so it has to be context aware when decisions are being made.
4:10
We cannot just, uh, jump and look through a small window and try
4:13
to make a decision out of it.
4:16
We need to have a little more effort in, um, comprehensively
4:20
understanding the whole landscape from multifactorial multiple point of view.
4:26
And then, um, we, uh, go about it.
4:31
Jesse Hirsh: Right on a, a very technical answer.
4:33
Again, there's no wrong answers.
4:35
It's always curious to see kind of how people respond.
4:39
Now, the second question I like to ask people really gets into their
4:44
kind of origins or backgrounds and, and I try to make it kind specific
4:50
and abstract at the same time.
4:52
And when I researched your background, both kind of intellectual and
4:57
professional, the the question that really came out at me, given the
5:02
range of interests you have, and at the same time I suspect a range of
5:07
opportunities that have come your way.
5:10
Why biology?
5:13
I can understand why agriculture, I can understand why technology or
5:18
engineering or even computer science and artificial intelligence, but why biology?
5:24
What was it about biology that drew you in and kind of kept your interest?
5:30
Suresh Raja: So just to be, uh, you know, um, as a disclaimer, I'm a classically
5:34
trained agricultural engineer, all three degrees masters in PhD in ag engineering.
5:40
Uh, nowadays they call it bioengineering.
5:43
Um, so a bird or an animal, if you look at a cow or a pig or a swine, or a horse
5:51
or a sheep or a poultry, or a duck or a Turkey, it's a biological entity.
5:58
So, naturally biology takes precedences, as I use the word interconnected.
6:04
Um, biology alone cannot provide the solution.
6:08
So we bring in the multidisciplinary field, uh, aspects of engineering,
6:14
aspects of computer science, uh, in biology, along with data analytics to
6:20
provide, uh, ish quality, quality of life for the animals, but at the same time
6:26
offer ways for sustainable profitability for the producers and farmers.
6:32
So it's truly multidisciplinary in that angle.
6:36
Jesse Hirsh: And, and that's what fascinates me.
6:38
Because on the one hand, most of the people in my life that I've
6:41
encountered who are in the engineering side, in the technology side,
6:47
they have no interest in animals.
6:48
They, they often have no interest in human beings.
6:52
but you approach this not just as an interest in animals, but to your point,
6:57
you're looking at their wellbeing, you're looking at the sustainability
7:01
while also looking at prosperity and the profitability of an operation.
7:07
I guess as a follow up still in, in the context of, you know, where you're
7:11
coming from intellectually, did you come to such a well-rounded perspective?
7:17
I mean, often researchers and intellectuals, they tend to
7:21
really specialise and, and kind of hone in on stuff.
7:24
But to your point, you're, you come at this very multidisciplinary, but
7:29
even more so there's an ethics or a philosophy to your work that also
7:34
seems to be coming at things from as many perspectives as possible.
7:39
Suresh Raja: Yeah, indeed.
7:40
Um, you know, uh, if you look at my employment, I'm currently,
7:45
uh, joint appointed between the faculty of agriculture and Truo
7:49
campus, uh, within the department of Animal Science and pop culture.
7:53
And then the 50 percentage of our appointment is in the Halifax main campus
7:57
within the faculty of computer science, uh, artificial intelligence, um, focus.
8:04
So, um, you know, when a farmer, uh, he or she, they are them, they walk and then
8:11
they park their car just maybe half a, uh, 500, 300 metres away from the barn.
8:15
As soon as they park their vehicle, a car, they use all of the five
8:20
senses to yoke and to make a, um, assessment of what's going on.
8:25
Oh, there is something iffy today.
8:26
There is an anomaly.
8:28
Something is not right today.
8:29
So how do they do it?
8:31
They use all their five senses.
8:32
They smell it.
8:33
Sometimes they hear sound.
8:35
Uh, if there is a, a based on the wind flow of it touches their
8:40
skin, they use all of these tools, uh, sometimes their own intuition
8:44
based on decades of experience to say meaningful critical statement.
8:50
Very often those statements are true, and then once they enter, it's exactly
8:55
the way they have, um, uh, perceived.
8:58
But at the same time, maybe a city kid or someone who is not familiar when
9:02
they go, they may make lots of mistakes.
9:05
So there is a lot of, uh, subjectivity and intuition is involved.
9:10
So how do we go about converting the subjective to objective with the help of
9:16
technology through a comprehensive way?
9:19
So we can't really make statements decision making in the livestock,
9:24
agriculture using one modality, one way of approach, because
9:29
everything is interconnected.
9:31
How do we actually move the needle from reactive to predictive through
9:36
proactive approaches with the help of this multidisciplinarity, with the help of a
9:41
lens that can see through the different thread that is connected to be able to
9:48
come to, uh, judgement and assessment and evaluation that can help the lives
9:53
of the animals at the same time, the decision making capacity of the farmer.
9:59
Jesse Hirsh: Absolutely brilliant.
10:01
That, that was such a great answer because it of connects the need to
10:06
have the intuitive level with, of the farmer, with the data-driven needs of
10:12
the researcher and the kind of systemic level awareness that, that, that affords.
10:20
I mean, I, I, I could keep with my questions building up a kind of narrative,
10:24
but I, I kind of want to jump to one of the bigger questions I wanted to talk to
10:28
you about, which is digital twins and, and the rise of digital twin technology.
10:34
Because what you just described in that answer was on the one hand,
10:39
a a lot of really human stuff like senses and intuition, also
10:45
a lot of complicated systems and, and that rely upon a lot of data.
10:50
So assume that many of our listeners right now have never heard of
10:56
the concept of a digital twin.
10:58
So if you could first kind of introduce it and then describe how it applies
11:03
to livestock and, and the way in which you, you use it yourself,
11:07
or you've seen digital twins be applied to livestock agriculture.
11:13
Suresh Raja: So for the livestock forming, digital twin is actually very new.
11:17
There is absolutely no commercial products available yet out there.
11:22
Uh, but when we look at the evolution of digital twin and other sector, it
11:26
was, uh, introduced like 40 years ago by NASA for space related applications.
11:32
Some, some recently in biomedical manufacturing applications.
11:37
So when we look at the word twin, so if we have, you know, twin babies born,
11:43
one is more or less similar to the other, uh, in terms of the behaviour,
11:48
in terms of how they express themselves.
11:51
So it's more like an avatar, uh, or a, uh, simulation.
11:56
Uh, we could say from the um, uh, animal forming animal agriculture perspective,
12:02
when we use the word digital twin.
12:06
So it is, uh, a soluble physical system so we can, uh, interconnect
12:11
it with multiple sensors.
12:13
For example, when we deploy different cameras, different sound recording devices
12:20
and other hardware, and collect these multimodality data, sound data, just like
12:26
audio or vocalisation from the animals, temperature and humidity related data,
12:31
environmental data of the animal husbandry system, and also the physiological data,
12:37
maybe the respiration rate, the blinking and the biological health related data.
12:43
If there are some video cameras that, um, measures movement and motility
12:48
and action of these, um, uh, birds or, um, cows, how do they move around?
12:54
How long they make different steps, how, what tie what side of the body, the cow
12:59
lies is the right side or the left side.
13:01
All of these complex, different data continuously collected.
13:06
Uh, and then we use artificial intelligence to build an avatar,
13:11
a metaverse that can mimic the exact nature of these animals
13:17
or the animal husband resistant.
13:19
And we go about making decisions.
13:22
So right now, dashboards are super common.
13:24
You walk into a barn, you can see multiple dashboards.
13:28
People are also using mobile applications to, uh, in a real time fashion to look
13:33
at, Hey, what's going on in my barn in terms of temperature and humidity?
13:37
Uh, they can even, uh, see the calling of the cows, et cetera.
13:41
That's, uh, most of the time it's what happened before in the past.
13:46
But twin is a slightly different concept.
13:49
The idea here is can we be able to run what if scenarios, what will happen
13:55
in the future if a change something?
13:58
Uh, so we could, uh, do a variety of simulation based experiments
14:02
without subjecting the animal to go through certain external stressor.
14:07
For example, as a country here in Canada, we rely on a different country
14:12
for, um, a inflow of supplement for our feed for the animals, if there
14:18
is a delay or a hiccup in terms of the supply chain and logistics,
14:22
what will happen in about 45 days.
14:25
So I don't have to, um, uh, uh, completely run the scenario and use my physical,
14:31
biological resources to do that.
14:34
Digital tune can do the whole thing with a fraction of cost
14:38
in a smaller amount of time.
14:41
Like that we can run a lot of, um, uh, uh, what if scenarios can be run.
14:46
So there are many advantages to it, even from the sustainability
14:50
net zero perspective.
14:52
Uh, if I am a diary former producing milk, can I be able to go about in
14:58
determining, um, to, for me as a farmer who have a hundred cows, for me to
15:04
pro, to be able to produce one lit of milk, how much of methane I'm emitting,
15:10
what does the methane equivalent emission per lit of milk right now?
15:15
It's a very complex formula.
15:17
It's really a tricky to arrive at that number.
15:21
But digital twin provides a more holistic way of deriving that information.
15:26
So is the case with YAG form for me to produce a dozen yag How much of
15:31
kilogramme of carbon dioxide or methane my particular pal tree or Turkey form is
15:37
emitting so I can run and understand these scenarios in a much more easy manner?
15:43
There are also welfare aspects to it.
15:47
For example, you might have heard of the terminology deep fake.
15:52
Uh, so it's a deep fake.
15:54
Some politicians who are not real politicians, they suddenly appear
15:57
on a very big screen in a bigger meeting and then make statements,
16:02
uh, pretend to make statements.
16:04
But those are not real politicians.
16:06
That statements may have very serious consequences.
16:09
So technology has been used for, uh, wrong purposes, but the same
16:14
technology can be put for good use.
16:17
For example, if the mother cow is being sent to slaughter houses for
16:22
some reasons, can we be able to use the defect technology via metaverse
16:28
virtual reality, uh, to ate biologically, physically create that, uh, twinning
16:35
of the cow so that the cow can appear suddenly and help to suit the baby call?
16:43
So there are different elements and different application scenarios possible
16:47
with the help of the new technologies.
16:50
Jesse Hirsh: Well, and, and let me ask two questions that kind of dig a little
16:55
deeper into what you were describing.
16:57
And you know, the first is, where is the data to enable all
17:01
that simulation coming from?
17:03
Because, because you gave a, a, a range of examples that in and of
17:07
themselves are kind of fascinating.
17:09
And I suspect this gets into the research that you've been
17:13
engaged in for your entire career.
17:15
But the other question to that is, are there some animals?
17:19
Are there some livestock that there is more data available than others
17:25
and therefore it's easier to create twins for those kinds of animals?
17:30
I mean, dairy cows for example, strike me as one that would
17:34
have a lot of data available.
17:37
What are the animals that we need to be collecting more data where we need to be
17:41
working more closely, maybe with, with farmers to be collecting that data so
17:46
that we can enable these kinds of twins?
17:50
Suresh Raja: Sure.
17:50
Yeah.
17:50
So that also opens up the question of, uh, the small scale farmers
17:55
versus, uh, large scale farming.
17:58
Uh, you know, somebody who is subscribing to certain, uh, solutions where lely or
18:03
deal level or a bigger company, uh, in terms and all the cows are having some
18:09
wearable sensors, uh, around the neck.
18:12
So they collect these movement related data.
18:15
Uh, so most of the forms nowadays, they have this automated robotic
18:20
systems for milking and diary systems.
18:23
But the small scale forms, no, they still rely on partial manual based milking.
18:29
So the axis and the data availability is much more limited.
18:35
That also decides in terms of the quality of the twinning we can develop.
18:41
Uh, at the same time when we enter into a poultry barn or a broiler or a laying hand
18:45
interest quality system, the model changes because we typically make decisions, not
18:51
peral, it's per floor group of boats.
18:54
And then decisions are made based on those group based intervention strategies,
18:59
uh, on form daily management systems.
19:03
So the, um, uh, access to data also changes dramatically.
19:08
Uh, typically environmental data built conveyor number of ygs, of
19:12
course feed and water are continuously available for all the animals.
19:18
Uh, but to be able to quantify per animal feed even in the diary system and we walk
19:23
into it, we drive a tractor, and we, uh, basically provide feed on both sides.
19:29
There is no way we quantify how much each cow is actually consuming.
19:34
There are different back, uh, calculations based on the chewing capacity, the colour
19:39
sensor, the kind of approximately try to guess, but that's never accurate.
19:45
So there is a trend in terms of adopting what we call a non-invasive.
19:51
Without touching and prodding the animal, can we be able to collect more data?
19:55
So the computer vision systems are gaining a lot of momentum.
19:59
We instal different kinds of cameras at specific locations.
20:03
They capture, um, uh, uh, the movement of the cow.
20:06
They also look at how often the cow consumes based on
20:10
their different, uh, biometric profiling and morphometric future.
20:15
We can actually measure, uh, not only the body weight of the cow, the
20:20
lameness, but also the consumption rate.
20:24
So more of a individual precision diary.
20:27
Uh, consumption and nutrition assessment is possible.
20:30
So the, the availability of data is very heterogeneous.
20:35
It depends on a number of factors, the availability of the sensors and so on.
20:41
So yes, so that is directly proportional to the quality of the twinning we are
20:46
looking at, uh, uh, uh, developing.
20:48
So the more the sensors, the more the heterogeneous data, the quality of the
20:54
data, the better for the twinning model.
20:58
Jesse Hirsh: Well, and, and I'm fascinated by what you said about the
21:01
computer vision, because certainly having cameras or installing cameras,
21:06
WW within operations, large and small is, is becoming increasingly accessible.
21:12
But I I'd also love to hear you elaborate on kind of the, the research process.
21:17
Like the way in which, you know, you work with students, you work with farms.
21:21
I mean, how, how are you able to kind of develop these types of
21:26
methods and gather this kind of data and what are the limitations
21:31
and does open source play a role?
21:33
And, and I'm kind of, uh, I'm, I'm not really coming up with a clear question.
21:37
I'm kind of throwing out thoughts as I go through this.
21:41
But you raised a really interesting point when talking about the difference
21:45
between large and small producers.
21:49
it possible to start empowering small producers to be collecting
21:53
data and being able to participate in some of these larger research
21:58
projects and, and larger simulations?
22:02
Uh, again, to extend that benefit as widely as possible?
22:06
Suresh Raja: Yeah, indeed.
22:07
So how do I go about collecting the data and, you know, manage, supervise,
22:13
uh, the students and mentor them?
22:15
So my primary appointment is research, uh, position.
22:18
Um, so I do a little less teaching and more research based
22:22
on the nature of my employment.
22:25
I also hold what they call as, um, research chair in
22:29
digital livestock forming.
22:30
This means my pri uh, lot of, um, distribution of efforts is
22:34
primarily geared towards research.
22:37
And also personally, I'm very passionate about, um, research.
22:42
So this makes me take, so we have a really nice group here.
22:46
I get to supervise students from computer science background as
22:50
well as animal science background for the past three years.
22:53
Every year, few weeks, we put all of these students in two big vans
22:58
and then drive across Atlantic Canada, stay in multiple regions
23:02
in Sussex County and New Brunswick.
23:04
Some in PEI, we stay for a few days to a week, um, instal different, um,
23:10
hardware components and collect data.
23:12
We have been to more than I would say, probably a dozen, uh,
23:15
forms, um, this past year alone.
23:18
So we do a lot of interactions, engagements in terms of field data
23:23
collection, uh, and then, um, through different data sharing agreements
23:28
with milk testing companies, uh, and also these bigger, uh,
23:33
companies which provide variables.
23:35
Uh, we li uh, uh, share and collect data via their own API application programming
23:43
interface or a PA as they call it.
23:45
And we collect the data when we enter the bond.
23:48
Sometimes there are multiple providers.
23:50
They provide different kinds of, uh, services and systems.
23:53
So it's not only one particular company.
23:55
Milk testing is provided by a different company.
23:58
The variables entirely a different company.
24:00
Uh, the robotic milking system, they belong to typically two or
24:04
three companies in the world.
24:06
Those are the bigger companies, uh, rotary Milking Parlour or the
24:10
robotic milking, uh, missions.
24:13
So like that.
24:14
And then if there is a veterinary record or wet health, those services
24:18
are provided through a different system.
24:21
So the data itself is very multifaceted and, um, complex and heterogeneous.
24:27
So we make an effort to, um, uh, collect all of those data.
24:32
So, um, irrespective of a small scale to large scale farmers, we collect them, even
24:37
if it's a small scale farmer, depending on the interaction and even a small
24:41
whiteboard based records they write.
24:44
And, uh, we take snapshots and some of those handwritten
24:48
records, we collect those data.
24:50
So all of these data are pooled into to be able to develop the twinning.
24:56
Uh, almost all of our research, if you look at, uh, we publish all the
25:01
mathematical models, algorithms, codings, everything is published for free.
25:06
Uh, because we believe strongly in open source.
25:10
How do we bring down the cost?
25:11
Can we be, for example, a few years ago, uh, uh, both the ca Canadian
25:16
and the American friends, US and European, uh, they passed a lot.
25:20
Hey, you have multiple different mobile phone companies.
25:23
You have USBC type.
25:25
There is a different way of charging your phone.
25:27
Why there are so many cables, you are confusing the public now they made it.
25:32
US BBC is the only way all the phones can be charged.
25:37
So that's the idea behind it.
25:39
Can we be able to create a common standards rather than
25:42
everybody beat their own drums?
25:44
Uh, is there a common way we can create a consensus so that there is a unified way
25:50
of providing solutions for the farmers?
25:53
Can we be able to narrow down the adoption pathway for the producers and farmers?
25:59
Can we be able to bring down the cost of technology?
26:02
It does not matter whether you have 60 cows or 300 cows, but still you
26:06
pay proportionally less based on the number of animals you have.
26:10
You don't have to sign up for a larger licencing fees for the
26:14
specific technology you want to adapt.
26:17
All of these considerations go into in terms of the technology
26:21
development as well as the validation.
26:24
So we pay enormous attention and care in the, uh, design development as well
26:31
as the validation of the technology.
26:33
And then when we go to the, we also have a nice form in the campus.
26:37
We have about 65 cows within in the Truro campus.
26:40
We work almost on a daily basis.
26:43
So we do a smaller scale, um, trial validation, and then we do the
26:48
second scalability at the commercial forms, which is about 250 cows.
26:54
So, um, this is the nature and the scope in terms of the way we work.
26:59
We get things done in technology development for the adoption pathways.
27:05
Jesse Hirsh: Right on.
27:05
And, and two thoughts that kind of, or two questions that kind
27:08
of directly come out of that.
27:10
One is, you know, you mentioned the different kinds of data that when you come
27:14
into a barn or you come into an operation, is interoperability a growing concern?
27:21
And, and I say this, 'cause one assumes that those sources of
27:25
data will continue to increase.
27:27
I mean, equipment is increasingly a source of data.
27:30
You mentioned the environment.
27:31
There might be multiple operations or multiple facilities at a single operation.
27:37
And then the, the second issue is, is kind of the notion of the
27:41
commons itself or, or a data sharing.
27:45
'cause on the one hand we're talking about using data to create a digital twin.
27:51
on the other hand, maybe different people have different twins.
27:55
Maybe a different advisory company or a different financing company,
28:02
or a different insurance company has a different twin and therefore
28:07
a different view on the marketplace or a different view of risk.
28:12
I mean, to what extent should we all be sort of going off the same page and to
28:18
what extent do we need to be anticipating the interoperability, the ease by
28:25
which not just researchers, but farmers and governments and regulators being
28:31
able to kind of see the big picture.
28:34
Or am I getting ahead of myself?
28:36
Are, are these concerns we should be thinking about?
28:39
Or is, is it premature to be concerned about this sort of thing?
28:43
Suresh Raja: Yeah.
28:44
Yeah.
28:44
Um, just so you are right, because we have already thought through
28:47
this, there are many, uh, folks, um, have, uh, given a clear thought.
28:52
There is also an agency in the EU icar, uh, specifically related to
28:57
direct cloud data sharing forum.
29:00
So, um, um, when we create a standardised protocol, for example, you go to an
29:06
electronic store, let's say Best Buy, and then you buy a laptop, and then it comes
29:11
with what we call as figures of merit, which means this laptop will work from
29:16
this temperature to this temperature.
29:18
If the temperature goes below plus five degrees CEL Celsius,
29:21
automatically the battery will drain.
29:23
This is the normal operating conditions.
29:26
So this figures of merit, irrespective of the brand, uh, we
29:30
buy, they are pretty much the same.
29:32
Same like cars.
29:33
It does not matter what model car we buy.
29:35
They have a specific unique standards with which they have to operate.
29:39
There is a regulating body that oversees that.
29:43
IP uh, like a international electrical electronics, uh, authority.
29:48
They oversee some of those similarly for the diary cow,
29:51
as well as for the twinning.
29:53
Um, right now, as of now, there is really no biological or,
29:58
uh, uh, animal forming related standardisation protocols available.
30:02
There is nothing available, but people are wrapping around their head, coming
30:07
together to be able to create a consensus.
30:10
So, again, as you rightly, um, put it, you know, if there is a finance
30:14
company or an insurance company or a veterinary health service provider,
30:18
if you have everybody beat their own drum, then it'll create more chaos and
30:24
unnecessary burden for the, um, farmers.
30:28
Farmers does not need a lot of, um, very shiny dashboards.
30:33
They want to know insights.
30:37
Can it help to bring down the time to results?
30:40
Um, uh, can they make decisions in a very much more, uh, sophisticated
30:45
manner without spending lot of effort from their side?
30:49
Can the needle be moved from reactive to predictive through a proactive way?
30:54
For example, on a Tuesday morning, if there is an anomaly, uh, if it
30:57
takes two to three hours to come to a conclusion, can they bring down
31:02
the time to maybe within 10 minutes, all of the input are being fed into
31:06
it with a very limited energy spent and time spent from the farmers?
31:12
Can they arrive at the conclusion within, uh, a reasonable amount of time?
31:17
If the technology can offer that, then there is more of a pull of
31:22
the technology from the sector.
31:24
If not, it's basically a push.
31:26
I developed this tool, that platform, and we try to push to
31:29
the farmers and make them adopt.
31:31
So in terms of technology development, that need to be a clearer consensus,
31:35
clearer understanding, uh, of, of the unified standardisation protocols
31:41
among multiple stakeholders.
31:43
That also opens up, Hey, who owns the data?
31:46
Eventually the animals are being managed by the producers.
31:50
Farmers.
31:51
Farmers, uh, pay certain services.
31:54
Sometimes they have subscription model.
31:56
At the end of the day, some companies basically allow the
31:59
farmers to provide certain data.
32:02
They may withhold a large amount of data.
32:04
So that's also being thought of from the government perspective, from the, um,
32:09
farm, like, uh, we have Diary Farmers of Canada and different, uh, marketing boards
32:14
and the co-op societies here in Canada.
32:16
So from their perspective also, uh, the end of the day, the data
32:20
belongs to the farmer and of course the animals themselves.
32:23
So, uh, uh, we are, as a researcher, we act as a liars and we act as a bridge,
32:28
as a catalyst to provide solutions.
32:31
But it has to be a genuinely impact creation.
32:36
Genuinely value is added at a reasonable manner.
32:40
It's not like a particular company makes money, um, uh, without
32:46
specific, um, uh, a value addition.
32:49
So those are some of the guards that have been thought through in
32:53
terms of how do we enable this?
32:55
What are the things we need to take care of in actually bringing the technology?
32:59
So again, some of these boundary conditions may not be applicable, uh,
33:02
universally between different species.
33:05
So those would vary depending on the context and the type of,
33:09
uh, uh, livestock or the, uh, bird based farming that happens.
33:14
Jesse Hirsh: Well, and I appreciate your point about sort of push and
33:17
pull that often there's different dynamics in terms of whether it's
33:20
the industry itself wanting a technology or external pressures.
33:25
And, and for me a great example of that of course is disease transmission.
33:30
And, you know, avian influenza, for example, is something that
33:35
I know the, the poultry industry has, has been very concerned with
33:40
Suresh Raja: Yeah.
33:41
Jesse Hirsh: has this technology, where has disease mitigation, disease,
33:46
uh, tracking and, and traceability.
33:50
Where has that played in fostering both, uh, industry adoption, but
33:56
also the evolution of the technology?
33:58
'cause a few times you've sort of alluded to the ways like, you know, the point
34:02
about the, the using video and, and, and imaging, to what extent is the need to
34:10
understand the health of the animals and the larger health of society, how much of
34:15
that is starting to drive, uh, uh, some of this smarter and, and some of the data
34:20
collection that we've been describing?
34:22
Suresh Raja: So, yeah, it's a very important, uh, uh, focus.
34:25
Uh, I was a faculty at the University of Gulf many years ago for about eight years.
34:30
So we developed few technologies for point of care diagnostics
34:34
of our avian influenza.
34:36
So we take a swab, put it in a little, um, small, um, uh, tube, and then
34:42
we know the results right away then and there within, uh, five minutes.
34:46
So as of now, this gold standard is the PCR.
34:49
We send a sample, either bled or faecal or oral swab to a centralised laboratory,
34:55
do PCR, and then we know the results via email or phone call, uh, if there is a
35:00
high pathogenic, um, uh, hemagglutinin present in the virus, uh, uh, samples.
35:07
So the goal for, at that time our project was how do I bring
35:11
down the time to results?
35:13
Can I avoid, uh, shipping the samples so we know the results then and there?
35:18
So that was the focus.
35:19
We also got a couple of us patented technologies and
35:22
signed some licencing agreement.
35:24
Right now, with the help of artificial intelligence and also in
35:28
the, um, realm of digital twinning, it becomes more and more clear.
35:33
For example, uh, we have access to, um, different weather patterns.
35:38
We also know what will happen in the next, uh, uh, week or so, in term
35:43
more or less, in a much more accurate manner, the changes in the, uh, weather
35:48
conditioning of that particular region.
35:50
So that is known, uh, uh, from the epidemiological perspective.
35:55
Also from the pro, we have this pro action, uh, within Canada, we, uh,
36:01
have a certain way of, um, doing trick and trays and track where
36:05
the animals are moving, when it is expected to move from a governance
36:10
perspective, from the, uh, Canadian Food Inspection Agency, agriculture,
36:15
Greenfoot, Canada, uh, perspective, we have a, a satellite eye view of
36:20
what's going on, what kind of movement of, uh, different animals happening.
36:24
So there is a different set of data.
36:27
We also have certain records constantly or periodically, if not in a, a
36:32
continuous fashion, um, um, with some discreet data points available between
36:37
multiple forms across the country.
36:40
So those data can also be taken into account.
36:43
So all of these data, when we pool together, even though there are some
36:48
missing data point with the help of deep learning, we may be able to, um,
36:53
um, say with the much more clearer probability that, you know, this is
36:58
the incidents level in terms of your epidemiological, uh, decision making
37:03
and also, uh, making statements in terms of the possibility of, uh, that
37:09
going in epidemic or a pandemic nature.
37:12
So the ai digital twinning really comes in handy in terms of predicting
37:20
the particular incident or even, even before that happened with the help
37:25
of multi-data, uh, in a, because of the ability to compress data and also
37:31
the, uh, irrespective of the data volume just a decade ago was very
37:36
difficult because the data volume is humongous To be able to process all
37:40
of the data we need few weeks now, the few weeks have become few minutes.
37:45
Actually within few minutes a large volume of data can be processed,
37:50
competitionally, uh, efficient to uh, provide critical insightful
37:55
predictive values based on that.
37:58
So there is also a way of incorporating much more, what I
38:02
would call as non-tangible data.
38:05
Maybe there is a certain, um, uh, somebody is tweeting a certain, uh,
38:10
keyword and that is kinda creating an anomaly that can be picked up by
38:14
a web browser or a, um, robot that's constantly scanning the internet.
38:20
Oh, there's a certain fluke.
38:21
Maybe there is a possibility.
38:23
Oh, it's a 0.5 percentage possibility.
38:25
Then we can factor that into it.
38:27
So like that we can create much more complex way of modelling to be able
38:32
to come up with more precise and accurate way of, uh, track and trace
38:37
and offering value, uh, from the governance and safety perspective.
38:42
Jesse Hirsh: Well, and, and let me kind of take two questions and, and
38:46
fold them into one, which the natural response to that is, you know, how do you
38:51
communicate that to the decision makers?
38:54
How do you communicate that to the farmers in the sense that if you
38:58
know a pandemic or if you know an epidemic is going to happen a week
39:03
away or even days away, that may not be enough to reach some people, right?
39:08
It, it, it can be difficult to, you know, uh, get people to pay
39:13
attention to something that they may not want to pay attention to.
39:17
then the, the second part of that that I'm, I'm kind of curious about is,
39:23
know, the, the broader notion of, of health and wellbeing of the animals.
39:28
Because on the one hand we're talking about reacting to a bad situation,
39:33
but a lot of your research is kind of on the preventative side.
39:37
It it's looking at how do we kind of create better welfare, better
39:41
environmental conditions, a better life for these animals, not just because that
39:47
ends up benefiting the producer as a whole, but because it ends up creating
39:51
all sorts of better solutions right.
39:53
Through the value chain.
39:55
How, how have you been approaching that?
39:57
What, what, what is your argument here and, and what role does the technology
40:02
and data play in either helping to do the knowledge translation, explain to
40:07
people what may be coming, or on the flip side, help, you know, improve
40:12
the, the, the life, how improve the wherewithal of the animals as a whole?
40:18
Suresh Raja: So, uh, in the year 2025, just a year ago, we
40:21
launched four mobile applications.
40:24
Those are freely available in both the iOS Apple store and
40:27
also in the Google Play Store.
40:30
Two of them for diary sec, uh, sector one is called Diary Ya or Diary at Canada.
40:37
The second one is Moog.
40:39
The two are for PAL tree sector, CL talk and cl.
40:43
So, uh, these out of these four mobile applications, three are based
40:48
on what we call as bioacoustics.
40:51
We listen to the vocalisations, the sound of the animals based on the sound.
40:56
We try to assess the mental makeup of these, um, animals.
41:00
Um, um, are they, if is the cow going through heat stress?
41:05
Uh, is the bonding between the mother cow and the cough good enough so that they
41:09
are happy, um, uh, in terms of welfare?
41:13
Is there a specific delay in access to the feed when they queueing
41:18
up in front of the automated, uh, milking machine for milking?
41:22
Um, um, or is it too much queue?
41:24
Is the waiting time much more longer than actually it's necessary?
41:29
Uh, if they are standing for longer time, sometimes even I have seen personally
41:33
the, um, of the cow starts losing milk and the milk goes on the ground.
41:39
Um, so, uh, uh, that's a clear, uh, uh, you know, uh, impediment
41:43
in terms of the welfare.
41:45
So some of these technologies look at these smaller telltale signs.
41:49
Sometimes we brush it off and we look at from a larger perspective, small little
41:54
changes can provide, uh, a bigger impact in terms of the wellbeing of the animal.
42:00
So by optimising, uh, we can actually provide the energy quality of life, uh,
42:06
in terms of the resource optimization, net zero and carbon credit pathway.
42:11
If I go and tell the farmer, Hey, you instal the solar panel in all
42:16
your roof, they will push me back.
42:18
Who is going to pay the capital cost?
42:20
Uh, that's an expensive proposition idea.
42:23
I am, uh, giving to them.
42:25
But at the same time, if I tell a poultry farmer, Hey, you know what?
42:29
You do the cleaning of your poultry farm every 53 weeks.
42:33
All I am suggesting is that move from 53 to 47 weeks.
42:38
Because based on the number of birds you have in the poultry barn, the y
42:43
production, the amount of the different volatile organic components in the
42:47
gas, um, the vocalisation sound and the movement of the bolts, uh, these are
42:53
the certain anomalies that's happening.
42:56
I'm not asking you to do every quarter, I'm just asking you to
43:00
move the needle from 53 to just six or seven weeks in front.
43:05
It.
43:05
It may not really cause a bit, uh, bigger impact, but it'll have a drastic changes
43:12
in the quality of the life of the boats.
43:14
The more you keep the boats happy and healthy, the obviously you will
43:18
see a production, uh, increase.
43:20
And also if they fall into a certain sickness or a health related challenges,
43:25
the recovery would be much quicker.
43:27
We call this is resilience or the adaptation capacity, and
43:31
they go to a certain disease.
43:33
How it much, how long it takes for the animal to go to the ground
43:37
state, the zero health state.
43:39
Normal health state, the faster, the better, the longer, which
43:43
means we need to tune and to tweak and optimise the operating.
43:48
So by paying.
43:50
Uh, and when we look at the ventilation system in a pal barn,
43:53
or even the fan that, uh, kicks off when there is a heat related,
43:57
uh, incident, heat days, hot days.
44:00
So we provide suggestions.
44:02
Okay, you do it exactly 30 minutes and this is the exact PPM for
44:06
your conditions, um, in terms of rotation of the blades of the fan.
44:11
So small little changes, not much.
44:13
There is a screwdriver change here, there, there, small make tiny little
44:17
changes, but it gives a drastic, a bigger impact down the road.
44:22
So it is not practically possible to look at these smaller changes,
44:27
uh, when we look through the lenses of traditional research.
44:31
So artificial intelligence and twinning can also provide value for small
44:36
scale farmers in terms of enabling the quality of, uh, enrichment of
44:41
the life of these animals at the same time without sacrificing the
44:46
profitability or creating burdensome in terms of operations of the farm.
44:52
Jesse Hirsh: Well, and you just gave a great example of sort of the livestock
44:56
equivalent of precision agriculture, of, of understanding how to fine tune
45:01
processes, operations and facilities.
45:05
I, I'm curious from your perspective as a researcher, uh, when a lot of people talk
45:10
about adoption of precision agriculture, they mostly are looking at crops.
45:17
To what extent are livestock producers, not just in Canada, but around the
45:22
world, starting to adopt precision practises when it comes to managing their
45:29
herds, managing their, their livestock?
45:32
Suresh Raja: Yeah, in, uh, I used to be a faculty in for four and a half years.
45:36
I was a professor in Netherlands, so I have visited many forms.
45:41
Uh, recently we were in the United Kingdom travelling crisscrossing,
45:45
uh, different parts of the UK as part of the Canada UK Livestock,
45:50
uh, innovation Exchange programme.
45:52
A bunch of us, they put us in a bus in uk and we were really, uh,
45:56
travelling the full country, had a better understanding of the landscape
46:00
of a diary innovation ecosystem.
46:03
So in terms of the precision livestock forming, uh, it's
46:07
very similar UK and, uh, Canada.
46:10
But, uh, some of the western, uh, European countries like Germany, a little bit
46:14
of France and Netherlands, uh, they are very quick in terms of adopting this
46:20
precision animal forming or precision livestock forming in terms of the, um,
46:26
uh, deployment of different technologies.
46:29
For example, there are some companies, uh, now we can even actually assess
46:34
the weight of the broiler without touching the broiler just by scanning
46:38
and collecting videos and images.
46:41
Um, there are also different ways of assessing, uh, uh, the swine industry
46:46
health, uh, using camera based, uh, technologies, uh, here in North America.
46:52
Um, the way, uh, because it's very big, huge, uh, in terms of the, uh,
46:57
number of animals we handle, uh, um, by land also the distance with
47:02
the forms are also very, very big.
47:04
So, um, our way of forming is slightly different compared to,
47:08
um, some of the western countries.
47:10
Although in terms of breed, in terms of some of the feeding
47:13
conditions, it's pretty much similar.
47:15
We follow same style and same protocols in rearing those animals
47:20
in terms of production, et cetera.
47:23
So there are certain, uh, uh, compare and contrasts and analogies we can do.
47:29
Uh, but uh, the technology slowly evolving.
47:33
Um, uh, you are very much correct.
47:35
The uptake of precision agriculture is very high in the crop side,
47:40
but it is, uh, picking up in the animal side, animal forming.
47:45
Uh, but it'll continue to grow.
47:47
I would say.
47:48
Jesse Hirsh: Right on.
47:49
And you know, one of the things we try to do here on the Future Herd Podcast is
47:55
get into the realm of policy and kind of ask guests the kind of policies or the
48:00
kind of regulations that they think will, uh, uh, enable innovation and prosperity.
48:07
And ironically, one of the reasons I encountered your work was I, uh, saw
48:12
a post, an article you wrote about Canada's AI strategy, kind of criticising
48:19
it for being too urban focused.
48:22
So I'd love to hear you elaborate on that and talk about potentially the way which,
48:27
not just AI policy, but agricultural policy could kind of work together to
48:33
enable some of the, uh, outcomes that you've been describing in today's episode.
48:39
Suresh Raja: Yeah, sometimes I get a bit frustrated when I see those policies that
48:43
are mostly geared towards city folks.
48:45
And, for example, even cybersecurity.
48:48
Uh, people think that, um, uh, cybersecurity means, uh, only
48:53
finance or bank or hospitals.
48:56
But cybersecurity can also happen in a poultry bond or a diary bond
49:01
when a certain attack happens in a, uh, you know, food production
49:05
system or a milk processing system, uh, it can have a HAC in terms of
49:11
supply chain in terms of accessing to the food security of our country.
49:16
Um, so, uh, when we look at the dual use technologies, I think that for the past
49:22
few weeks, that's kind of hot buzzword.
49:24
Everybody talks about dual use technologies, defence and et cetera.
49:28
I say agro defence comes first because we need to, if there is an attack
49:34
in agriculture, it is an attack in the country's food security attack
49:38
in the country's infrastructure.
49:41
So please do not treat agriculture as an afterthought or a second
49:46
thought, give it a priority along with the major infrastructure such
49:51
as finance and hospitals and human health, et cetera, has to receive the
49:56
same level of deserving attention.
49:59
So that's what, that's what triggered me to write this, uh, article in policy
50:04
options and, um, asking for the attention of the policy makers, Hey, you know.
50:10
Uh, bit more, uh, focus is needed in agriculture area.
50:14
Uh, especially, you know, when we, we have a very good supply chain management
50:18
for both in diary and the poultry system.
50:21
How do we create a resilience?
50:23
We get hit from all the time from other countries and so on,
50:26
but how do we provide a value?
50:28
How do we create a resilience in the system?
50:31
Are there things that we could do so that there is a sustainable, um, uh,
50:36
shared abundance and prosperity among all the stakeholders within the country?
50:41
Can technology play a role from the policy perspective?
50:45
What are the things we could do to, um, enable make that happen?
50:50
So even, for example, data sharing, when artificial intelligence is being,
50:55
uh, constantly used for development of tools, platforms, and technologies, I
51:00
always propose, hey, do not develop a black box model at the end of the day.
51:05
Former has to have a say.
51:07
You know, it has to be a grey model.
51:09
So farmers can say, you know, over supersede the AI model.
51:14
No, I don't need you to perform this action.
51:16
They should have that, um, freedom, independence to completely, um, uh,
51:21
um, be the final decision making authority in terms of the adoption
51:26
and deployment of the technology.
51:28
So during the whole technology development adoption process, how do we infuse and
51:34
make the farmers as the owners, uh, uh, that need to be considered when
51:39
we develop these governance models and technology development pathways
51:44
from the government perspective?
51:46
Jesse Hirsh: Right on.
51:46
Absolutely.
51:47
E especially your point about the governance, just on the user level of the
51:51
user being empowered to override what, what, what the system's recommending.
51:58
I think what you've demonstrated, uh, so brilliantly today is the
52:03
way in which agriculture is at the forefront of a lot of, not just
52:08
the development of AI technology, but its deployment, its iteration.
52:14
what I'd like to do as a kind of closing for guests is encourage
52:18
them to essentially do a shout out, like who are the researchers?
52:23
Who are the people that you're following, either in Canada or around the world
52:28
that you would like to see the rest of us kind of plug into or pay attention to?
52:34
You know, where are you getting your inspiration from?
52:37
Suresh Raja: Um, I read quite a bit.
52:39
I also follow, um, different disciplines, not necessarily in agriculture.
52:45
Um, um, you know, I, um, look at some folks from Oxford University,
52:50
Marion Dawkins, who does a little bit of welfare re related research.
52:55
Some people in the United States, my fellow faculty colleagues, uh, some
53:00
policy makers in, uh, Western Europe.
53:03
But they also, um, when we meet in conferences and we have
53:07
conversations, I derive inspirations from them and get new ideas and
53:12
new directions for my own research.
53:16
Jesse Hirsh: Right on.
53:17
And you know, a final question that I kind of have to ask for my own selfish
53:21
purposes, your kind of next interest?
53:26
I mean, on the one hand, I suspect that you are absolutely up to your eyeballs
53:31
in the current research projects you have all of which are fascinating.
53:37
But I'm kind of curious as to whether someone like yourself is anticipating
53:42
what comes next and what that might be.
53:44
Uh, what would be the kind of research interests or research
53:48
projects that you see on your kind of medium to long-term horizon?
53:53
Suresh Raja: Yeah, so we have a plan for five years.
53:56
So for the next five years, we know exactly what we are going to do.
53:59
Uh, so it's more of a computational interconnected systems.
54:02
So for, uh, modern animal forming, um, how do we bring down the time to results?
54:07
How do we improve the quality of life?
54:10
Can we be able to enhance the interactions between the animal and
54:15
the animal caretaker and the farmer in terms of bonding, in terms of empathy?
54:21
Can we be able to move the needle from transaction based relationship to
54:26
co-creation based relationship more often?
54:29
Sometimes when we look animal forming, we like, Hey, you know what?
54:32
I give you food or feed.
54:34
I give you shelter, you give me back yag, you gimme milk, you gimme meat.
54:38
So it's more of a transaction based.
54:41
Yes, it's indeed it's needed.
54:42
We are 9 billion people growing and growing.
54:45
We need to feed protein and all the people, but at the same time
54:49
can be infused the values of human, uh, in terms of this co-creation
54:55
possibility with the animal.
54:57
Uh, I belong to a human kingdom.
54:59
You belong to animal kingdom.
55:01
We have both have short lives.
55:03
Can we create a harmony synergy through which we can use the
55:08
technology as a bonding way, as a, um, bridging way to coexist and
55:13
co-create and travel in this journey?
55:17
So those are some of the elements we have been looking at.
55:20
How do we create shared prosperity and abundance for everyone, um, um,
55:26
as we navigate this tumultuous times.
55:30
Jesse Hirsh: Wonderful.
55:31
What, what a great way, uh, to end what has been a, a really,
55:35
uh, uh, enjoyable and in my case, educational episode, uh, Resh.
55:42
Thank you very much and I, I hope you can come back.
55:45
We're only just scratching the surface of what's possible, especially to
55:49
your point of thinking about it as co-creation, that we are working with
55:53
our animals to help feed the planet.
55:56
Thank you so much, Siresh.
55:59
Suresh Raja: Thank you, Jesse.
56:02
Jesse Hirsh: One of the things I find so frustrating about
56:05
contemporary public discourse is everything seems to be an either and.
56:15
One of the things I find so frustrating about contemporary public
56:19
discourse is everything seems to be an either or rather than a yes.
56:24
And what I absolutely enjoyed about my conversation with Suresh today is
56:30
it not only adds a level of nuance to conversations around animal farming
56:37
conversations around the use of ai, but it also suggests that there's an
56:43
incredibly important role for empathy, for centering the concept of care, of
56:48
co-creation, of really thinking about not just our relationship with animals,
56:54
but our relationship with each other, with nature, with the world as a whole.
56:59
Too often we focus on the conflicts.
57:02
We focus on our differences rather than finding our commonality, rather
57:08
than finding our passion, our empathy, our desire to make the world a better
57:13
place, and that is one of the main motivators I have for the future herd
57:18
is food is one of those places where people tend to open their minds.
57:24
They tend to indulge in a certain level of curiosity, try new
57:28
things, entertain new ideas.
57:31
And that's where I do find that there's this interesting intersection between
57:37
ai, animal farming, animal welfare, and the concept of intelligence itself,
57:44
let alone its most, uh, pressing application, which is intuition, which
57:50
brings me back to me and my goats.
57:53
One of the things I love experiencing and exploring when I take them on
57:58
our walks is our combined intuition.
58:02
On the one hand, I'm trying to read the group, 'cause
58:05
fundamentally I don't control them.
58:07
We're out beyond the fences, out at the edge of the property where
58:11
they could just run onto the road and get hit by a truck or something.
58:16
And at the same time, they're enjoying themselves.
58:19
They're enjoying the experience.
58:22
So there is both a level of communication and a level of intuition that if I could
58:28
combine that with the kind of technology, uh, that CSH described, and I sort of
58:35
already am in the sense of the videos and the observations and the, uh, kind of
58:40
anthropological, is that the right word?
58:42
No zoological observations that I'm engaged in.
58:47
But again, to actually turn this into research, to turn this into
58:51
data would allow me to take my skills, my herd, to the next level.
58:56
And that is really the opportunity, the potential that some of this
59:00
technology offers, as long as it's accessible, as long as it's open source.
59:06
And in that sense, uh, uh, uh, we will be returning here on the future herd
59:12
to the subject of ai, to the subject of animal welfare, farm welfare, farmer
59:19
welfare, plant welfare, and open source.
59:24
Uh, a preview of the future.
59:26
Yes.
59:26
And a opportunity for us to end this episode.
59:30
Uh, thanks again for listening.
59:32
The future, her.ca is the website, uh, your friends are
59:37
who you should share this with.
59:39
And until next time, I'm Jesse Hirsch.
59:42
We'll talk to you soon.