Constance Li | How AI could improve the lives of trillions of animals

[00:00:00] Constance: There's plenty that exists out in the world about AI and humanity, but very little about how AI will affect animals. And every time we've had some sort of technological advancement, it may not have gone well for animals. Like when we developed antibiotics, we were able to crowd more farmed animals into smaller indoor spaces instead of having them be outside.

If AIs are used in farms, could they be used to increase productivity so much more at the expense of the animals' welfare? And the converse is also true. Like, we could use AI to make animals so much happier, or we could use AI to create alternatives to the things that we currently use animals for, like cultivated meat.

[00:00:55] Beatrice: I'm here today with Constance Li from Sentient Futures, and the theme for today's conversation is AI for animals, which I think may seem like an unusual mix for most people. But why do you think — how do you connect those dots, AI and animals? Why does it matter?

[00:01:15] Constance: Yeah, so part of why I started this was because there was not much that was said on it. There's plenty that exists out in the world about AI and humanity, but very little about how AI will affect animals. And there are more animals alive than humans at any given point, by a pretty massive number. And every time we've had some sort of technological advancement, it may not have gone well for animals.

Like when we developed antibiotics, we were able to crowd more farmed animals into smaller indoor spaces instead of having them be outside. When the Industrial Revolution came about, we developed factory farming. And so there's not a great reason to think that as AI technology progresses, it's necessarily going to be good for animals.

So I think we should start thinking about this now and trying to plan carefully about how we could steer it to go better for animals than it otherwise would have.

[00:02:34] Beatrice: Yeah. How big would you say this problem is? Are there a lot of animals suffering today because of this?

[00:02:45] Constance: I don't know how much AI has been deployed to directly affect animals right now. I know it's being used in some research, and it actually has the promise of decreasing the number of animals that need to be experimented on. And it's also being used a bit in the dairy industry where animals are large and valuable. But by and large, a lot of farms don't have a lot of technology on them. They still rely on a lot of manual labor and documentation. And then most animals in the wild don't have any digital technology surrounding them at all. So right now there's not a lot of influence. But as AI progresses and controls more of what humans are capable of, I think that animals will see some consequences.

[00:03:44] Beatrice: Yeah, because we do know that a lot of animals are suffering due to factory farming.

[00:03:51] Constance: Yeah. So if you add up all the humans that have ever lived, that's about 117 billion. And if you take the amount of farmed animals, even just on land, that is roughly the amount that are killed for food every single year.

[00:04:15] Beatrice: Oh, wow.

[00:04:16] Constance: Yeah. And then if you go into fish, that's about one to two trillion every single year. And if you go into shrimp, then it's seven to 70 trillion. So the numbers are just unimaginable. Like, our brains aren't built to compute those numbers and imagine that many individuals

[00:04:39] Beatrice: Yeah.

[00:04:39] Constance: that are — yeah, like their whole purpose of being alive is just to be eaten by humans.

[00:04:49] Beatrice: Yeah. Those are shocking numbers. And we also know that the conditions that they exist in are not ideal. Normally, like, maybe most people when they think of animals for food think of cows roaming green grassland or something like that. But that's not necessarily the case for most.

[00:05:16] Constance: Yeah, that's what's usually put on the packaging. But 99% of animals that are raised for food are raised on what's called factory farms, which are intensive confinement systems. Many of them are in cages — like chickens that lay eggs, they're usually in cages with seven to nine other chickens and they have the average space of about a piece of paper. They can't even spread their wings throughout their entire lives. And if one of them dies — which there's a very high mortality rate in these systems — then oftentimes they'll just be in the same cage as them for long periods of time. Workers may not find the dead ones. So there's a lot of suffering on these farms.

[00:06:10] Beatrice: Yeah. I remember when we spoke before this episode, you said some shocking fact about shrimp and like how many die before they even get to the process where they're supposed to die — when we're going to eat them.

[00:06:24] Constance: Yeah. For farmed shrimp, shrimp are a very sensitive species, as are a lot of aquatic animals. If anybody's ever had an aquarium before, they know how difficult it is to regulate pH and ammonia. And when you have more of these animals in one place, that type of regulation is even harder. And they also get more easily stressed out. So 53% of farmed shrimp actually die because of these conditions before they ever reach slaughter. So that's 53% of the individuals that we don't even eat at the end of the whole process. And so that's literally trillions — yeah, trillions of individuals that just die for no reason other than we were not very good at managing their environment.

[00:07:24] Beatrice: Yeah. I think at this stage maybe people are shocked that this is the Existential Hope podcast. But if we're gonna get to the point where we can actually hopefully do something about this — I know your background is as a physician actually, but you are now working on this. How did that happen? How did you come to work on this?

[00:07:43] Constance: Yeah, so I went through medical school. I trained as a physician in physical medicine and rehab. This kind of discipline focuses on injuries and pain and how to restore people back to function. But I had always had a soft spot in my heart for animals. And I always knew, since the age of 12, about the scale of suffering that happened on factory farms. And then as I learned more about the suffering of animals in the wild and also the potential suffering of artificial minds that might exist in the future that can feel, I realized that this is a huge problem. And then I was looking around for who was doing something about it, and I really couldn't find anybody.

So I decided to stop being a physician and focus on building out this field of people who are asking these questions and developing solutions for them.

[00:08:57] Beatrice: Yeah. And I think we should touch a bit on what it is that you actually do. What does a sort of day-to-day look like for you? What is it that you're actually working on?

[00:09:12] Constance: Yeah, so every day looks very different. We do a number of different programs. So we hold conferences, we have an online fellowship, we have our Slack community, we also have our newsletter. For the last month or so I've just been in conference planning mode. We just had our largest conference ever — 276 in-person attendees in San Francisco. So it's been a lot of emailing speakers and making signs to make sure people can navigate through the building. But then, like last year was the year that we first became an organization, so it was a lot of paperwork on my end.

[00:10:02] Beatrice: Yeah.

[00:10:03] Constance: What I'm hoping to do more of in the future is strategy — so talking with leaders of other organizations that are working on animal welfare or AI safety or the welfare of artificial minds, and just building up this ecosystem of identifying where there are research gaps or action gaps and making sure that there are people there to fill them.

[00:10:36] Beatrice: Great. Yeah, thank you for doing that. If we take it back to the basics a bit then, and to the AI for animals question — what do we mean when we say AI for animals?

[00:10:50] Constance: Yeah. So "AI for Animals" was the original name of Sentient Futures. And back then I was looking at the fields of AI safety and seeing how rich and nuanced it was. They had disciplines for like governance — making good laws that make sure that AI is being used safely. AI alignment, which is trying to test the AIs in different situations to make sure that they're acting well and having values that humans would agree with. And also applied AI too — like who actually gets to use AI and making sure that it's not capable of doing dangerous things like creating bioweapons that could cause the next human pandemic. But none of these things existed for animals. And so I would say that you can also have governance for animals, insofar as how AI affects them. Like, if AIs are used in farms, could they be used to just increase productivity so much more at the expense of the animals' welfare? If we're developing AIs, are we measuring their attitudes towards animals and considering their interests, especially if there's a scenario where there's a trade-off against human interests? And then, finally, could certain humans use AIs to harm animals? And how could we prevent that?

And the converse is also true. Like, we could use AI to make animals so much happier. Or we could use AI to create alternatives to the things that we currently use animals for, like cultivated meat, or like predicting what kind of taste or texture certain molecules will have so that we can imitate the things that we would normally have to get from animals.

[00:13:21] Beatrice: You mentioned that you changed from "AI for Animals" to "Sentient Futures." Why did you change?

[00:13:30] Constance: Yeah, so when I first started the organization, it was just a side project and I was just thinking, AI is a big thing, I care a lot about animals, let's call it "AI for Animals." So it wasn't really that well thought through. As we progressed though, I realized that there are other technologies that could affect animals, such as gene editing or like engineering better equipment to handle animals. And so it's not necessarily just AI that can have a big impact on animals, although I think it probably does have the lion's share of potential impact.

And then similarly, it's not just animals that are being neglected in the AI conversation. Oftentimes we don't really talk about artificial minds.

So there are two categories of artificial minds. The one that people would think of most commonly is the AIs themselves, where all of their functions happen on servers. And if they can speak like humans, if they can sense like humans, if they can reason like humans, at what point could they also feel things subjectively like humans? I think it's too early to rule out that possibility. And once they could feel, then they would become moral patients.

[00:15:14] Beatrice: Like, we would need to care about their wellbeing.

[00:15:16] Beatrice: Yeah, we don't need to — we should.

[00:15:18] Constance: Yeah, we should. I think so. And yeah, unlike biological beings that would take a generation to replicate themselves, AIs could just be copied. So, is GPT one entity? Is it a million entities? Like, how should we count it? We have no idea how to answer these questions right now. But it seems worth investigating.

And then there's also a very little-known industry of biocomputing where they will actually put human neurons onto chips and then use this to power things similarly to AI on servers. And there have actually been some studies where things that these systems have not been intended to do — like process sound — have been observed in these tiny little brains. And they're just little tiny brains. They could evolve a bunch. They could grow different functions that would be morally relevant, but not have the systems of communication that humans have — like being able to voice whether they are suffering or not. And I think that's something that we should actually be asking questions about before letting these companies scale and make gigantic server racks of tiny little human brains.

[00:17:06] Beatrice: Yeah, seems worth looking into basically. Yeah. So thank you for doing that. When people just hear you say, "Oh yeah, we work on AI for animals," is there something you feel like they tend to misunderstand? Like, are there any assumptions that people get wrong?

[00:17:28] Constance: Yeah. I think the first thing that people's brains go towards is just using AI to do all the things that we had previously been doing for animals — so using it to create better communications about what's going on in factory farms, public awareness, or using it to improve the quality of certain campaigns, like getting chickens out of cages. But the further you go into this, the more you realize that the structures of government and society that we are used to are going to change.

Like, I don't think humanity has ever faced a technology that is recursively improving like AI. And just like compound interest, it doesn't go linearly. And our brains are normally wired to expect progress to happen linearly, but progress is happening at a much faster pace all the time now because of AI. And so I think the actors that are holding all the power right now may not be the actors that are holding all the power in, say, a year or two. And the things that we think are possible will change quite a bit, and we need to think a couple moves ahead about how we can shift the development of these systems to better impact animals. And I think that's going to be a much higher leverage course of action.

[00:19:24] Beatrice: Yeah. So I actually took your AI for Animals Fellowship recently, and it was really interesting. I definitely recommend it to people. And I think there were a few sort of themes that stood out to me as we were going through the sessions. The first one was something called precision livestock farming. Can you just explain to people what that is?

[00:19:52] Constance: Yeah. So precision livestock farming is using AI and other technologies within animal farming. So this could look many different ways. There are sensors involved that could be cameras looking at the animals, audio recording devices recording all the sounds that they make, even temperature sensors or water quality monitors. So you could gather a lot of data continuously from these farms, which previously hasn't been done before, and you can use AI to analyze it to get insights for the farmers. Like, it can detect disease a lot earlier so the farmer can go in and intervene and give medications before it gets too bad and they need to kill the entire population of animals, which has happened quite a few times now because of avian influenza and swine flu. Or you can actually just automate the whole thing, and based on how the animals are growing, automatically create a specialized feed and feed them at certain times that are optimal. So yeah, it's a developing field right now, but it basically is giving so much more control over how animals are raised on farms, either for better or worse.

[00:21:38] Beatrice: Yeah. Do you know, are there examples of it being used already? Or do you know what kind of data it collects?

[00:21:47] Constance: Yeah. I'm not sure how it's being used commercially right now. It is being piloted on small scales in some places. So I'll share one example that I know of, which is ironically called "AI 4 Animals" — but it's a four with the number four. And this was developed by Deloitte Netherlands. It's a computer vision system that taps into video footage from slaughterhouses, and they work primarily with pigs.

And so one of the big problems with pig slaughterhouses is improper stunning. So normally, pigs will need to be stunned — with an electrical stunner to the head — in order to become unconscious so that they're not experiencing having their throat slit during the death process. But this is really easy to get wrong, and pigs might just be paralyzed but still fully conscious. And there are certain things you can look for to indicate this, such as their eyes moving around after the stunning or them twitching. And so these computer vision systems can identify when that is happening, they can identify if that's happening on certain shifts with certain techniques. And if you change the technique or the staff member around, if that gets better or worse. And that can be used as a training tool for the slaughterhouses to improve their practices and ultimately reduce unnecessary suffering for the animals.

[00:23:44] Beatrice: I mean, that sounds like a somewhat positive example — and they are trying to use it to improve welfare. I think, imagining the worst cases, it's relatively easy that you could just use it to increase productivity in general and not at all care for how it affects the animals' wellbeing.

What do you think would be a best-case scenario of precision livestock farming?

[00:24:16] Constance: Yeah. So the way I like to think about this is that you can optimize for either productivity — like how much milk a cow produces or how much meat a chicken ends up growing — or you could optimize for welfare. Some signs of that could be like how much the animals are displaying natural behaviors, or playing with each other, or how healthy their skin or feathers look.

And then there's the intersection of the two, which is the sweet spot of the win-win. So if chickens are less stressed, they will attack each other less. And cannibalism is actually a big problem in chicken farms, and if you decrease the rate of that, then the experience of the chickens is better and also the number of animals that make it to market is higher.

So I think realistically speaking, in order to scale these systems, you have to work with the existing industry and show that it is in their interest to do something that would also be in the interest of animals.

Oh, but to answer your question about the best-case scenario — right now there's not a lot of individual care being given to each animal. So I think there could be like two workers taking care of a hundred thousand chickens, and their main job is just to pick the dead chickens up off the floor. If we had AI systems that could track every individual animal and give them specialized care and attention, then you could see when this animal is starting to get sick, you could give it a special medication just for that animal. Or if a certain animal is being picked on, then you could isolate it and give it some better enrichment.

Yeah, there's a lot of room for improvement in the attention and care given to animals.

[00:26:44] Beatrice: Yeah. Best case, if we were able to lower the cost and increase the efficiency of being able to help them individually, that seems like it would be great.

And another of the sort of buckets that I remember we discussed a lot in the fellowship was interspecies communication — so like humans being able to talk to animals, ideally. Or talk — maybe that's a stretch. And this is something that I think I may have even mentioned on this podcast before, that I'm very excited about. Perhaps naively, because I think when we talked about it, it became a bit of a more nuanced picture for me. Because I think I had always thought of it as, oh, but if we're able to talk to animals, everyone is going to understand that animals are suffering, and we're going to change how we do things.

What do you think about this category of interventions? If we were to learn to communicate better with animals, would we be able to improve their lives?

[00:27:57] Constance: Yeah. I think it can certainly be a tool that we can use to better accommodate the interests of animals and know what their state of mind is. But I do not think that it's the silver bullet to help animals. We know that animals are suffering on farms. We know that a wild animal is suffering when it has a parasitic infection. And still there's not much that's being done about that — partly because resources are scarce and partly because people aren't listening, even if animals are communicating visually or audibly. And there are also many humans on this earth that we can understand perfectly well that are suffering, and that isn't helping them all that much.

[00:29:03] Beatrice: Yeah. Do you have a sense of what's possible with this right now and what work is being done on interspecies communication?

[00:29:13] Constance: Yeah, so there are a couple of groups that have been working on this for a while. There's a project called CETI that works to decode whale communications. And I think one of their goals is to try to get legal representation for whales. And then there's also Earth Species Project, which covers a broad array of different animals. And they've both developed AI systems that can analyze these vocalizations and start to make some sense of them. It's very rudimentary right now. You could give the AI models a sound clip and ask, what species are in this sound clip, and how many of them, and what are they doing right now? Are they mating? Are they in distress?

So I think that's where we are right now. And there's a lot of thought being put into the ethics of this kind of technology, because yeah, it's interesting for us to know where these animals are, how many of them there are, what their general state of being is. But this can also be used by people that want to harm animals rather than just study them. So poachers could potentially use it to lure animals into a trap so that they can hunt them.

And also, if we don't know exactly what animals are saying, and then we can actually synthesize vocalizations of animals now using these models — if we play back these synthetic vocalizations to them and we're saying something distressing, that could cause them harm. And there's one example where they synthetically generated the voice of a matriarch elephant that had died, back to this elephant herd.

[00:31:50] Constance: And it caused the herd a lot of distress because they were very confused —

[00:31:51] Beatrice: "I know this voice. This voice hasn't been around for a while. Where is it coming from?" Yeah. There's so much to it when you start to think about these tools and things that you have to consider that just isn't maybe top of mind.

[00:32:06] Constance: Yeah.

[00:32:08] Beatrice: What do you think could be a positive case? Or if we are able to understand animals better, what could we improve with that?

[00:32:16] Constance: Yeah. So I'd say that if humans were really motivated to improve the subjective experiences of animals — as we currently are for our pets — then it could give us a lot of insight into how to do that. For example, prey animals tend to hide their distress quite a bit. So this includes rabbits, maybe hamsters. When they're scared, they tend to just freeze, but their heart is beating like crazy on the inside, and we may not necessarily recognize that as humans. And so we could start to actually translate that better to our own minds and know, oh, this animal actually needs some rest and for us not to handle them anymore, because they're scared — even though it's not acting like we would act when we were scared.

[00:33:29] Beatrice: Yeah. I know this from having spent time with horses. They also tend to hide when they're in pain, for example, because they don't want to get pushed out from the herd or something like that. And

[00:33:42] Constance: Yeah.

[00:33:44] Beatrice: it's really only very recently that people have understood that. We've worked with horses for hundreds, probably thousands of years. But it's really only recently that I think people have started to understand that. And so yeah, I'm sure there's just a lot of things. And I'm also guessing the further away from humans they are to some extent, the less we understand them.

[00:34:09] Constance: Yeah. Mammals — we have a lot of similar communication signals with each other across a lot of mammals. If you're in pain, you'll scrunch up your face a little bit, maybe arch your back. We don't really have that kind of signal for, say, chickens or lobsters. They don't have any facial expressions that we can rely on. So AI can really do a lot for translating whatever body motions they have when they're in states of fear or distress to ones that we can understand.

[00:34:54] Beatrice: Yeah, I think that would be so interesting, especially if we're thinking about those trillions of fish or shrimp — it would be great to be able to understand their pain signals better. Yeah.

I guess another thing that touches on this and that was something we talked about is this sort of realm of genetic welfare. Could you maybe explain what we mean when we talk about genetic welfare in this context?

[00:35:24] Constance: Yeah. So it's the propensity for having better or worse experiences based on your genetics. So maybe people can understand this best through things that are normally passed down through generations, like depression — depression can be more prominent in some families than others. Also chronic pain — some people are much more predisposed to having pain for a lot longer, even if the injury is the same as another person's. A lot of this is rooted in our genetics. And this might have been useful in the wild, but now we don't have as many selection pressures.

Similarly for animals, they don't really have many selection pressures right now. Like, I think if you put a chicken in the wild, it would not survive. But yet chickens are one of the most populous animals in the world right now.

So if we can select for traits that are more adapted to the environments that individuals will live in, then they could experience better lives than they might otherwise have. So maybe an animal example I could give you is: imagine having a small New York City studio apartment and you adopt a dog, and the dog is a pug. The pug will probably be decently happy there and sleep most of the time. But if you had adopted a Border Collie, that Border Collie might go a little bit insane in the studio apartment, because it's predisposed to needing to be able to run around

[00:37:32] Beatrice: a lot. Yeah.

[00:37:32] Constance: Yeah. Its genetics are not adapted to the environment that you put it in.

[00:37:36] Beatrice: Yeah. Are there — do you know if this is being used already to some extent to improve animal lives?

[00:37:47] Constance: Yeah. So genetics for farmed animals has historically been used to increase productivity. So the animals that we have today are nothing remotely similar. I know you said "to improve their lives," but I'm going to start with the example of how it has made their lives worse.

Like chickens — they're usually not that big. But the faster they grow and the earlier you can slaughter them, the less feed you have to give them, the less energy they're wasting on just breathing and being alive. So chickens are normally killed at about two months old, when they're literally just babies. And they grow to such unnatural sizes that if you think about a human infant, it would be the equivalent of the infant reaching about 600 pounds at the age of two months.

[00:38:59] Beatrice: That's insane to consider — very clearly extremely unnatural.

[00:39:06] Constance: Yeah. And it has consequences for the animals. They've optimized for the growth of breast meat, but the bone development hasn't caught up. And so a lot of these chickens end up with broken bones and not being able to walk. But it doesn't matter that much as long as they're growing the meat, because they're going to be slaughtered anyway.

And so I think that genetics, as much as it's been used for productivity, could also be used for health. There's no reason why we couldn't choose the chickens that have the strongest leg bones and then breed them over time.

[00:39:48] Beatrice: Yeah. I think the realm of genetic welfare is to some extent the most promising if I think about trying to think more ambitiously about how we could improve the future for animals. We had for example the philosopher David Pearce on the podcast before, who talks about the idea that we should genetically modify all animals basically — also wild animals, for example.

So in many ways, I think he calls it paradise engineering or something like that, which I recommend you check out. We can link it in the podcast description. But yeah, what if you think ambitiously about what we could do with genetic welfare — what do you think we could actually make better for animals with this?

[00:40:45] Constance: Yeah. I think so. David Pearce is extremely ambitious and probably represents the extreme end of genetic welfare. I think that is within the realm of possibilities — it's a matter of how socially acceptable doing that is. There are a lot of ways in which we currently manipulate the genetics of animals through selective breeding that people don't really blink an eye at. But once you get into the realm of genetic engineering, even for humans, there's a lot more stigma around it and a lot more caution.

So a lot of the things that David Pearce talks about is actually genetic engineering, where we go in and change the DNA through scientific techniques rather than just selecting for certain traits and then pairing those animals together over generations.

Yeah. I think that over time we could just change animals — and also humans — to be perfectly adapted to their environments. Like, maybe they have the pain signals that they need in order to not put their hand over a stove, or to remove their foot if they're on a sharp rock. But maybe not so much pain that at end of life, when normally there's a lot of suffering involved, they need to experience that suffering as deeply.

[00:42:32] Beatrice: Yeah. So if we take it back a bit and think about all of these different sorts of tools at hand, what do you think in the next five to ten years we could change if we were able to push for a best-case, realistic future on this topic?

[00:43:00] Constance: Yeah. Honestly, it's really hard for me to project out five to ten years. I only started this organization two years ago and it's already progressed beyond my expectations. But I think that as AI progresses, there's going to be a lot more abundance in society. People are going to be richer. They're going to have their basic needs met, if the value is distributed fairly. And I think it may not be distributed totally fairly, but I think the general average will go up.

I think people will have more time, energy, and resources to think about those around them.

[00:43:54] Beatrice: Yeah.

[00:43:54] Constance: Including animals. And I think that more attention will be focused on how we could improve their lives, how we might reduce our reliance on meat products that come from them or on research that currently uses them. And already there's a lot of progress being made on this front. The United States government just said that they want to phase out animal testing. And I think that in the next couple of years, probably, there will be a lot more bans on animal testing.

And this is made possible through technological advances that can simulate what these animal experiments would have done in the first place — either by creating organs on a chip or doing computational modeling based on data that we've gotten before.

So I think replacements for animals will become a lot more prevalent. And I also think attention to animals, because we have our needs met, will also continue to increase.

[00:45:13] Beatrice: That's great. Yeah. Do you know — is there good data historically on whether, as countries have gotten richer, they tend to treat animals better? That would be my hunch, but yes, yeah.

[00:45:30] Constance: I think we are just starting to see this. So a lot of animals in the EU have some of the best welfare protections in the world. I believe Norway just banned the use of these fast-growing broiler chicken breeds — the ones that I described before. And there are the first signs of things getting better for animals once countries get developed.

[00:46:09] Beatrice: Yeah. And what are sort of the biggest hurdles that you think we need to overcome in order to get there?

[00:46:18] Constance: One is just resources. If you think about all the farmed animals that are out there in this world — and like I said before, every year just for land animals alone, it's more than all the humans that have ever lived — and they're in conditions that people would never really imagine. The organizations that are working to make their lives better have funding that is less than the yearly budget of the New York City Metropolitan Museum of Art.

[00:46:58] Beatrice: That's crazy. Yeah. I think maybe people think it's more or something, but to me that's insane — that as a society we choose to spend this on an art museum. Like, that's great, I love art, but it just seems like not the right prioritization to have made. Yeah.

So if we — hopefully people are listening to this and feel like, okay, oh my God, I want to help here — where would you point them? Where can they get started?

[00:47:37] Constance: We offer a number of different programs. I'd say probably the best way to get started is just subscribing to our Substack. So we put out a newsletter every month or two that collects all the news and developments from this field of technology affecting non-humans. And also opportunities like jobs or funding, and events like conferences to go to. It's not just ours — there are also a lot of other conferences that are popping up around this subject. So that's a really good way to stay on top of things and see what interests you.

If people want to dive even deeper, we have a Slack community that you can join where there are channels where people are talking about genetic welfare, wild animals, artificial minds. And that's a way to get connected with the community. We also have monthly socials that are virtual.

And then for those that are interested in getting a really good knowledge foundation, there's the fellowship, which you were part of the first cohort of, and we are now open for the second round. This is an eight-week course where there's a reading for each week, and then you get together in a small group with a facilitator and have discussions. And that's a really great way to get a broad overview of all the different possible interventions. I highly recommend it.

And then for those that are looking to get even more involved and start to meet people and maybe start projects and get hired, I would highly recommend coming to one

[00:49:30] Beatrice: of our conferences.

That's great. Yeah. I think the last question that I want to ask you is just: what are things that actually give you hope that we could do something here?

[00:49:42] Constance: Yeah. So there are a lot of things that give me hope. The growth of the community, the adaptability of those working in animal welfare to change and be flexible and adapt their interventions to technological progress. One thing that gives me a lot of hope is actually the fact that a lot of people working to create AI care about non-human welfare.

So historically, those that have had a lot of power and influence have not cared about animals very much. But I talk to people that work at these labs and this is one of their highest priorities, which is just really inspiring. Because the way that AI gets developed and used will have a really big downstream impact on everything else. And if we can get the starting point right, I think that will probably have the biggest impact for the least amount of effort.

[00:51:00] Beatrice: That's great. Yeah, that's very promising. Thank you so much, Constance. That's it.

Constance: Yeah, thanks Beatrice. It was great to be here.