Empowering Choices: Animal Welfare, Sustainability, & Innovations in Farming, Pet Care, & Food - Part 2: AI and Machine Learning to Improve Poultry & Dairy Industries & Pet Health

Dr. Fritha Langford 25:02
Absolutely. So, I think in terms of the second risk, it's important to realize that we don't know what the downstream impacts might be. And because we've only looked at a few generations, we might not have come up with all of the environmental threats that perhaps that gene is protecting against. And so, we may inadvertently reduce an animal's ability to cope with some other aspect of its environment, not just PRRS.

Erin McCann 25:15
Thank you for clarifying that.

Dr. Fritha Langford 25:17
So, there's all sorts of ways we could think about gene editing. Another one would be for another major disease that we see both in farmed animals and in wild birds, and that is avian influenza. Again, a viral infection which has huge mortality within both the poultry industry, but also, as I say, for wild birds. We're seeing that, I think, in the news just recently. The avian influenza has now reached Antarctica into the penguin population, which is very sad. One of the problems with avian influenza is each new strain does seem to be coming out of the poultry industry.

Of course, even though the poultry industry generally has quite high biosecurity, you have an awful lot of birds who are also all relatively closely related to one another, and that is a very good breeding ground for disease. One of the possibilities with gene editing is to breed our laying hens and potentially our meat chickens to not be susceptible to avian influenza, and that to be relatively broad spectrum, which would then reduce the likelihood of those animals from suffering from the disease. But it also reduces the likelihood of that disease mutating into a new strain.

So, there are some great benefits. But of course, that means that perhaps we focus less then on other potential treatments or other potential preventions of avian influenza from improving the conditions for chickens so that they experience better welfare when they're farmed, or from thinking about vaccinations, for example. In doing so, we actually leave our wild birds potentially more at risk in the meantime because we're not actually then focusing on other ways of combating this disease.

Erin McCann 27:49
Absolutely. And I think one key takeaway is for anyone listening who may think, "Well, I'm not particularly focused on chickens, but I care deeply about wildlife," whether we're thinking about penguins, other birds, locally, or in other geographies, we are seeing that when it comes to these strains, no matter what new strain it is, as you mentioned, and that continue to evolve and change over time, that they are resulting, at least in the vast majority of cases, right? We know that they're resulting from the poultry industry and that what may seem like a solution from one lens will likely not create an all-encompassing solution for all of the birds that are affected by this already. Am I hearing that right?

Dr. Fritha Langford 28:45
Absolutely. And there are other impacts as well. So, for example, there are many poultry breeds, but the huge majority of the poultry industry are using only one or two of the breeds, be it for meat chickens or for laying hens. Not all of the myriad of breeds that have arisen throughout time in all of the countries where humans are in the world. What we do then is we leave those other breeds that are perhaps really suited to the environment in which they have been bred into, to potentially suffer big consequences, high mortality consequences, from avian influenza.

We may lose the genetics that are held within those backyard flocks of interesting breeds from around the world. That can be problematic, too, for our future. So, when we consider, for example, say, in 20 years' time, when we look at climate models and the projections of where our climate is going to be then, we are likely to need birds -- if we are still farming chickens at that point -- we are likely to need birds that have a good resilience to heat and humidity, which our modern high-yielding breeds don't really have.

But some of our breeds, for example, that you see in the backyards of all sorts of tropical countries -- I was recently in the Philippines -- and you see different breeds in the Philippines of fantastically resilient birds. They're very resilient to, funnily enough, tropical conditions. Those genetics could be at risk if we just blanket gene edit our laying hens and meat chickens in the poultry industry and leave the others to their fate.

Erin McCann 30:39
So, there are many complicating factors when we think about the benefits.

Dr. Fritha Langford 30:44
Oh my goodness, it's very, very complex, and everything is interlinked. When you say absolutely, some people who are listening to this may have no interest in chickens. I mean, honestly, they should have an interest in chickens. Chickens are wonderful. But people might not have an interest in chickens, but they may well care for wildlife. They need to realize that our farmed livestock are everywhere. So, anything that impacts them can have a devastating effect on our wildlife.

Erin McCann 31:13
So, when we talk about the choices that we make as consumers -- and we'll talk about this later on in more detail -- the choices we make every day for our own meals, for the meals of our family, for what we choose to eat, matter. We'll talk about how those matter for the species in direct question, right? Whether we're talking about chickens, pigs, cows, and more, cattle and more, but they also matter for wildlife.

For those listening who are really interested in this conversation from a sustainability perspective, the food choices we make very much matter for the sustainability of wildlife and other implications, which we'll talk about in more detail. But I really appreciate you highlighting those connection points, because I think we can often summarize these statements in very high-level ways, right? It may be a simple message that our food choices matter, and what that means in practical terms for particular species, for particular wildlife populations, I think is a very helpful part of this discussion that you're really illuminating in concrete ways.

Dr. Fritha Langford 32:22
Thank you so much. And I promise I won't always just answer every question about chickens.

Erin McCann 32:15
I care about chickens, too.

Dr. Fritha Langford 32:16
However, chickens are so important because they are the most common land-based farmed animal that we have. It's many, many billions every year. So, yeah, I think without occasionally considering the chicken, we're leaving out quite a big chunk of how animals and sustainability interact.

Erin McCann 32:50
Absolutely. For our listeners, I had the great privilege of learning from Dr. Fritha Langford as part of the International Animal Welfare, Ethics, and Law program at the University of Edinburgh. One of the questions that we unpacked was, "When we think about the lived experience of chickens, do they have lives worth living?" I can certainly include some links to some helpful resources on that topic. This is an ongoing question in the animal welfare community and an urgent one when we think about chickens as well as other species. I really appreciate you raising the importance of that, and I hope that our listeners have a greater appreciation for chickens coming out of this conversation as well.

So, when we think about technology and innovation, we've talked a little bit about alternative proteins, cellular agriculture, and gene editing. What do you see as opportunities to leverage AI, data analytics, gene editing, and other technology advancements to enhance the care, health, and welfare of livestock while addressing sustainability concerns? I would also love to extend this question to companion animal care as well. When we're thinking about these things, what are the opportunities? What other obstacles do you foresee in these realms, too?

Dr. Fritha Langford 34:10
Again, a fascinating question, and one that I think we'll probably be able to answer better in five years' time compared to what we can do just now. We're just at the tip, aren't we, of understanding the possibilities of AI. So, in this case, we're talking artificial intelligence. We've just been answering questions about avian influenza, also AI, and occasionally I get mixed up because of artificial insemination, also AI. But let's focus on artificial intelligence.

Erin McCann 34:41
Good reminder not to use abbreviations. Yes, thank you.

Dr. Fritha Langford 34:45
Yes, it's very confusing, particularly for us in animal science. But yes, I think there are huge potential benefits, and we will learn so much more in the coming years. When I'm thinking about artificial intelligence, I'm considering mostly machine learning at the moment, that aspect of artificial intelligence. One of the ways in which that can be really helpful for animal welfare is being able to sort through data, being able to understand differences in patterns far quicker than humans can, especially when we're looking at large numbers of animals.

I'll start with a farm animal example, and then we'll maybe go into some companion animal examples. For dairy cattle, for example, dairy cattle are commonly farmed across the world, and they have different welfare challenges depending on where they're being farmed. If we take UK and US examples just now to help our listeners consider this, one of the biggest health-related welfare challenges that dairy cattle face is to do with lameness. Lameness, where an animal is showing a difference in gait -- so they're walking unevenly, they're not putting the same amount of weight on each hoof -- is down to the feeling of pain. The reason that the animal is limping is because she's in pain.

It's a big problem. What we know in terms of evidence is that if you look across UK farms, you see that, on average, a farm will have roughly 20% to 25% of its cows with some degree of lameness at any one time. We see similar numbers, slightly higher numbers, in the US and across Europe as well. There are different challenges in different parts of the world, but let's focus on the UK and US. This is a big problem area, and it's quite an intractable problem, by which I mean those numbers have not really changed in the past 20 years.

If we go back to 20 years ago -- in fact, I was going out around farms in 2005 to -- so, 19 years ago -- to assess cows on different farms in the UK. What we found was that the farms we were looking at, we would see 20% to 25% of cows with some degree of lameness. It's exactly the same numbers. Conversely, in another health-related problem that dairy cows suffer from, which is mastitis, the painful inflammation of the udder, we have seen extremely good work from farmers in terms of treating that disease and getting on top of it very quickly.

Although mastitis is still a common disease within dairy farming, we see fewer cows suffering from mastitis over a longer period of time because of the way we treat that disease. We use antibiotics where they're needed, painkillers are often used, and so we know that cows are not suffering from mastitis pain for too long a period of time when they get the disease in the first place. There are some cows, of course, that end up having problems with that disease, but in the main, cows are being treated very well for mastitis.

The reason for this is there are very high penalties if disease-causing organisms go into the milk tank, the big tank where all the milk from the dairy cows go, and also there are very high penalties if any drug residue -- so antibiotic, antimicrobial residue -- gets into the milk. This means that farmers want to treat it as quickly as possible because there's a huge loss in milk and therefore loss in money if either disease-causing organisms or drug residues get into milk.

Erin McCann 39:22
Those penalties are imposed by laws and regulations?

Dr. Fritha Langford 39:26
Correct, but also from the milk sellers, the cooperatives or producer-related organizations that sell milk. So, yes, there are big penalties there. For lameness, there isn't the same incentive. Although a cow will end up having a reduced milk yield overall when she's been lame for a period of time, the treatment isn't going to impact the milk tank, the large quantity of milk that the farmer will gather from all his or her cows, and the disease-causing organisms that live on the hoof, if it's an infectious cause, don't get into the milk.

So, there isn't the same incentive to treat, even though most farmers -- we know, again, from evidence -- that most farmers agree that lameness is a painful condition.

Erin McCann 40:21
In summary, the farmers know that lameness causes pain for the animals, but because that pain and the possibility of infection are really only affecting the animals -- they're not affecting human consumption of milk -- we don't see anywhere near the level of penalties or addressing of these issues in the way that you were describing they've been addressed with mastitis because those have direct implications for the safety and the consumption of the milk by humans. Is that a fair summary?

Dr. Fritha Langford 40:48
Correct. There are other complications as well that make it more intractable. There are roughly 60 different causes of lameness that interact with one another. So, I'm not saying that the dairy farmers only think about that milk going down the drain because there are these other issues as well. But it is one of the big differences between mastitis and lameness, and you can see that they are treated differently on the farm by most farmers.

Okay, so we have this issue where many of our cows are lame. Not many of them are really severely lame. That might be a much smaller percentage on the farm, but those animals that are feeling their feet are not walking in a normal way. We would say they are mildly lame, but that still indicates that they are suffering some degree of pain and discomfort from their feet. But they are quite difficult to pick out.

People can do it. You can become trained as a lameness scorer and go and score cows when they're coming out of the parlor or they're walking in a straight line away from you. It's very time-consuming. It's quite tricky to do in terms of keeping reliable with yourself. It's also quite difficult to maintain what we call interobserver reliability. If you have two people, they might not see the exact same level of lameness in a large number of cattle. It takes a long time to train people as well -- roughly about 15 hours of training to become self-reliable in terms of your lameness scoring. You need to do that on a regular basis in order to keep being reliable over time.

You can see that there are possibilities in terms of human scoring, but that it's quite tricky to do, quite expensive to do. What if we could train a computer to do it? What if actually a computer might be better at recognizing very subtle behavioral traits that cows are showing when they have this lower level of lameness? What if actually a computer could pick it out far earlier than a human once they are trained on a big enough dataset?

That's one of the really exciting areas of using artificial intelligence, machine learning, really, to understand large numbers of animals and be able to help in quick treatment and stop these chronic problems being around for so long, but also being such a big problem on farms.

So, yes, that's one example from dairy cows, and I have other examples from other different farming sectors as well. But I think we've talked quite a lot about farming, so perhaps we should go on to companion animals. There is potential for machine learning in improving the detection of problems for companion animals as well. This area is perhaps less mature. We're seeing quite a lot of products, I think, coming to market now with regards to farm animal welfare and using what I've just described for lameness in cattle, whereas we're not seeing that quite so quickly for our companion animals.

But it may well be likely that we can use similar techniques to recognize certain behaviors, skin conditions, or other types of early warning signs of health problems in our cats and dogs.

One difficulty, particularly with dogs, is that they're so morphologically distinct from one another. As in, there are so many different breeds, and they look very different from one another, and that can be quite tricky for computers to understand. So, we have to get over that problem, particularly if we need the facial features of a dog, to be able to train the computers to make the right choices.

That's quite interesting, considering how well we know dogs and how well we think we understand their behavior. Being able to get an appropriate data set to train the computers to recognize different facial expressions, for example, has proven to be quite difficult. But I don't think it's insurmountable. It's just requiring enough of all the different types of dogs to be able to do it. But yes, there are plenty of possibilities for using artificial intelligence, machine learning in particular, for improving animal health.

Erin McCann 46:15
Absolutely. So, if I'm hearing you right, essentially, we can think about the opportunities for machine learning applying to our pets, and we need to recognize that -- if we just pick two breeds, right? A Great Dane and a Chihuahua are going to have tremendous differences. So, there is that additional work that needs to go into training the models, right? And inputting all of those different breeds and features. So, it is early stage. However, there is plenty of opportunity for machine learning to advance veterinary medicine, the health of pets, and help, presumably, to inform the way we care for our animals at home as well. However, that likely depends on engaging with our veterinary teams and that interaction with the health side of our animal care.

Dr. Fritha Langford 47:09
Absolutely. But yeah, there's all sorts of ways we could think that artificial intelligence could be beneficial for particularly picking up on early warning signs of disease, but also what we might consider as behavioral problems of our companion animals, or perhaps unwanted behaviors within the home – maybe I would prefer to call them.

Erin McCann 47:35
Understandable. Absolutely.

Dr. Fritha Langford 47:39
But there are some risks. So again, whenever we consider these technologies, we need to think, well, what are they replacing? And is that problematic? So, if we take one of these technologies that we might use within our farmed animals to help us sift through large quantities of animals to pick out problems, maybe that means that the humans need to be involved less with those animals.

Perhaps that once again reduces our human-animal bond with our farmed animals -- one that used to be quite important when we had only small numbers of animals on farm, but now is becoming less and less important as we have more and more animals per human in a farmed environment. And perhaps that just brings us one step further away from understanding our animals. I think the human recognition of things when they go wrong, or being able to put two and two together in ways that computers can't do. Creative thinking is not possible at the moment from machines, and so they will just do what the algorithm says, right, as opposed to making critical decisions at critical times. So, I think it's important to ensure that there is human watch on the process as well.

Erin McCann 49:11
Absolutely. And just to draw on the pet example from a lens that many of us can think about in our day-to-day experience. So, many of us start with the love of our pets as our entry point into animal welfare, and then think about the lived experience of other animals, and we're caring for those animals every day. If we think about the possibility to leverage more data, to leverage AI or machine learning, when we think about enhancing our own care of our animals, we can think about these same themes for ourselves, right?

We have the bond that we have with our pets because we're engaged with them, because we are observing and interacting with them on a day-to-day basis. If we start to focus more on the data and getting whatever interface we're reviewing, we're focusing on outsourcing that, to relying on the data and not our own observations. If we're thinking, "Well, I'll get an alert when there's a problem, so I don't need to be paying attention to these various aspects of my animal's care." That's another way to think about what I'm hearing you describing, which is that we can remove ourselves so far from the equation that we're losing that bond with the animal, we're losing other insights that really do require human awareness and observation.

So, we want to think about the benefits, and we also need to be mindful of over-reliance, of removing ourselves from engaging with animals and the costs that go with that. Whether we're thinking about our companion animals and our love of our pets, whether we're thinking about the industry implications, or both.

Dr. Fritha Langford 50:52
Absolutely right.