D (10:36)

I'll now invite him to say a few words.

E (10:40)

It's such a privilege to be able to be involved with lse and particularly backing Jonathan. They eat without pleasure, cry without pain, grow without knowing it. They desire nothing, fear nothing. That was an 18th century philosopher, Nicolas Malebranche, what he thought of animal sentience. He was influenced in this by the godfather of the dumb animals movement, Rene Descartes, a man whose position on the subject was summed up recently as not suggesting throwing a puppy over a cliff is a good or wise thing to do, but not suggesting that it's particularly wrong either. Sounds outdated, but in 2025, humanity remains an extremely species species. People care about their pets, their cats and dogs. Few would justify throwing them off a cliff. But dismissing cows and pigs and chickens and crustaceans and all the rest as just animals allows us to justify all kinds of abuse, especially in factory farms. So when Larry asked me why I wanted to support this department's work, there were really two reasons. The first is that I'm a completely and utterly failed philosophy student. I did my ma back 41 years ago in 191984 and remember absolutely nothing beyond the title of my thesis, which I can't believe I actually had this title, AI and the Limits of Science of Man.

F (12:36)

But.

E (12:38)

But there's more to it than that. I became a vegetarian when I was 12 because I didn't believe in the way some animals were brought up and I thought I was too young to decide which ones. And over the years, for a number of other reasons and for a long time, I felt that I was doing the right thing. I was doing something positive, but I was just a bystander, Passive, not active. And literally, on one day in 2013, my life changed. I remembered that bad things happen when good people do nothing. So I decided to move from being a bystander to being an upstander by becoming an animal rights activist. I know there are plenty of other upstanders here tonight. You know who you are. We have one big thing in common. When we see other animals, we don't just see lunch. We see fellow sentient beings. So since that epiphany, as Larry kindly mentioned, I've launched many projects that educate and engage investors about the risks of factory farming, that support the growth of alternative proteins, develop law and policy around animal welfare and the Collar Doolittle Prize. The reason why I wanted to do the Collar Doolittle Prize is using AI, you know, like the Rosetta Stone, unlocked hieroglyphics. I want to ask a cow how she feels when her calf is taken away. The acceptance of animal sentience underpins all of that, and that is what led me to Professor Birch. Because there's nobody quite like Jonathan. Sorry, rest of panel. He has a rare talent for answering questions most people haven't thought to ask, doing so in a way that is both academically rigorous and easily accessible to those of us who are neither philosophy professors nor, or have forgotten every single little thing we ever learned about the subject. The work he's doing around AI and what it means for our understanding and the acceptance of animal sentience is absolutely fascinating and so relevant to where we are here and now. And actually, you can counterpose that that with we are the creators of AI. Will AI know who we are as their creators, etc. I mean, we're just going to. It's just different radar. And Jonathan's making such a big difference. There are plenty of think tanks and lobbyists out there. But it was Jonathan and his team, as Larry said, who got the UK government to add crustaceans to the list of animals considered sentiment under UK welfare legislation. And it was Jonathan, you know, Californian legislation. It was him who was quoted about the banning of octopuses. He's one hell of an upstander. So there's nobody better qualified to lead this work, and there's nowhere better suited to hosting it than the number one one top rated university in the uk, the London School of Economics. This is where minds go to be opened rather than shaped, where old assumptions are challenged rather than taught. Speak to Larry and you can quickly see how the values that made the LSE what it is today are the same values that they'll make. It's a success in the years to come and I really appreciate it. It's such an honour for my name to be associated with this centre and the wider lse and I can't wait to see the impact it has. Thank you. Chickens have rights.

B (17:20)

Thank you very much, Larry and Jeremy. It is now my great pleasure to formally introduce my colleague, Professor Jonathan Birch. He is, as you have heard, the founding director of the center. But he's much more than that. He is the Spiritus rector of the Center. Without Jonathan's voice vision, we would all not be gathered here tonight and celebrate the opening of the new center. Now, this vision is based on groundbreaking research that Jonathan has made very important contribution in his field. Indeed, discussions concerning the evolution of social behavior, the evolution of norms and of course the study of animal sentience and the relation between sentience and welfare would not be the same without his pioneering research. However, Jonathan has never merely been a philosopher's philosopher, someone who works in the rarefied air of the ivory tower. His work has always been deeply rooted in practical questions, particularly those concerning animal welfare. Of course, the arena of animal welfare would not be the same without Jonathan's contributions, just as things would not be the same in research without his discoveries. I wanted to mention one example here that has been already mentioned by Larry and by Jeremy. But I think good things can never be mentioned enough. So let me remind you that the Animal Welfare Sankins bill was extended to cover crabs and lobsters in 2022 and these protections are now included in the Animal Welfare Sentience Act. Now, new codes of practice for shellfish industries have also been proposed and Jonathan's research directly impacted these legislative changes, which would probably not have come to pass without his work. But Jonathan is a worldly philosopher and a philosophical man of the world and one cannot think of a better director for the new center. So please join me in wishing Jonathan and the center all the best for their journey, which I'm sure will be long and fruitful.

A (19:53)

Thanks so much, Roman. It's a great responsibility, great privilege as well to be directing the world's first center for Animal Sentience, generously supported by the Jeremy Coller Foundation. My heartfelt thanks to Jeremy Coller personally and to all the staff at the Collar foundation as well, for making this possible. There's so much shared interest here. I think this is just the beginning of what will be be a long partnership built around shared values and a shared mission. The LSE is a place where people aspire to remold the world. That's what the Fabian Window depicts. It's an institution founded on a powerful idea that our sense of what is possible or impossible is constrained by limited understanding. That with better understanding, new possibilities, new futures, new ways the world could be, new ways to live come into view. And for 130 years people from all over the map, politically, socially, intellectually have come here to do research animated by that idea. And we are part of that. We're part of that tradition. The Jeremy Collar center for Animal Sentience is a team of people who believe in that idea. Our mission is to remould the way people see, think about and relate to other species. There's perhaps no other area where the words that's impossible are heard as often or where they so badly need challenging. So we're continuing a proud LSE tradition. I think we're continuing a really important British tradition too as Britain has a long history of standing up for animal welfare. We had the world's first anti cruelty law in 1822. We've had the RSPCA since 1824, the Vegetarian Society since 1847. We think of ourselves as a nation of animal lovers. But more recently we seem to have lost our sense of ambition. Progress has stalled and is in some areas reversing industrialized, intensive farming systems that have nothing to do with our values, nothing to do with stewardship or husbandry or care, or treating animals humanely. No regard for other animals as sentient beings with lives of their own have become far too common around the world and in this country too. Farmed animals and farmers alike are in danger of becoming cogs in corporate machines. Machines that put profit before care, compassion and dignity. A different vision of how we can relate to other species is needed. Our centre will be a place where that vision of a different future can be developed. Our goal is to do social sciences and humanities research that brings about positive change for other animals. Our work will be directed towards ethical moonshots, ambitious long term impact targets that if we achieve them, will put Britain back at the leading edge of animal welfare globally and help restore the harmonious relationship with other species that I think is one of the deepest human needs. That mission will be at the heart of everything we do. What unites us will not be any single discipline or methodology, but a shared research agenda and a shared commitment to producing research that benefits all sentient beings. We'll be doing strongly interdisciplinary work that will connect with my home discipline of philosophy, but also with veterinary medicine, evolutionary biology, psychology, Neuroscience, artificial intelligence, economics and law will all be about reaching out. Everything we do will be oriented towards achieving real world change. Our three initial priorities will be one, let's build a world where AI is used to benefit all animals and not to facilitate cruelty. AI is already having huge impacts on other species, yet they're often forgotten completely in discussions of AI governance, ethics and safety. That has to change. Driverless cars need rules about how to minimize harm to pets and to wildlife. What should those rules be when people want to outsource the training of their dog to an AI collar? What should be the ethical limits when chatbots are asked for instructions about how to exploit or abuse animals, what should those chatbots be saying? And perhaps most importantly of all, tech giants are already funding many startups, hundreds of them, that are building automated farming systems in which the traditional role of the human farmer is now played by AI systems in which AI, and not humans, monitor animals for health and for welfare, as well as for production parameters. There's huge opportunities here, but there's huge risks as well, risks of this entrenching welfare problems that are already serious. So how can we make sure this technology works for, rather than against the interests of other animals? With governments around the world taking an increasing interest in AI governance, the moment is now. Now is the moment to accelerate our efforts to promote the ethical use of AI in relation to other animals. That must mean limiting and constraining bad exploitative uses, uses that harm other animals and humans. At the same time, it must mean supporting and incentivizing good and beneficial uses, such as decoding animal communication to better understand what animals want. More on that later in the next part of the event. Priority 2 let's build a world in which all sentient beings are respected, even the smallest ones. Animal welfare laws have traditionally overlooked fishes and invertebrates, leading to notoriously inhumane practices such as dropping crabs and lobster us into pans of boiling water. Alexander Pope in 1713 lamented what he called the outrageous practice of cooking these animals alive. Three centuries later, it still goes on. We are changing things. My group's work led to the recognition of these animals as sentient beings in the UK's animal welfare sentience Act. But in my view, this hasn't led quickly enough to meaningful on the ground change. More work remains to be done. I think we're also seeing the end, the imminent end of octopus farming, one of the most straightforwardly indefensible kinds of farming that there is. You're putting predatory animals, easily able to injure each other, into close Proximity in small tanks. My group's work has influenced bans on this practice in California. It's another important step. But octopus farming, to be fair, is just a tiny drop in the ocean of global aquaculture, which is dominated by the farming of fishes, especially salmon, carp, tilapia and the farming of shrimps. This we're not seeing the end of imminently. In fact, on the contrary, the un, the World bank, the World Economic Forum have all called for a blue revolution. More and more of our food coming from aquaculture. But I think without an accompanying animal welfare revolution that starts taking these animals seriously as welfare subjects, this will mean more and more of our food coming from kinds of farming that violate basic ethical principles of animal welfare. It could mean millions of salmon being eaten alive by lice, billions of shrimp salmon having their eyes cut off. This can't be the future that we want to create. That can't be the world we want to bring into existence. But of course, bringing serious animal welfare standards to sectors that do not currently have them is very difficult. It's a crucial need for social sciences and humanities research of the kind we do at the lse, to understand the barriers to change and the ways to overcome them. Priority three. Let's build a world in which people are empowered to act in accordance with their love of other animals, rather than being nudged towards indifference. People say all the time that they love animals, but when it comes to making choices, the choices don't always reflect those values. How can we change this? Even when we all agree an animal is sentient, as with a pig or a chicken, this doesn't mean we will treat it well. On the contrary, the dire welfare problems experienced by these animals in intensive conditions, despite their acknowledged sentience, are well known. Leads to a question about what explains that gap between our beliefs and our values, between our self image as a nation of animal lovers and the practices that we nonetheless permit. It's a problem of cognitive dissonance. What can be done to understand where that gap comes from? What interventions can help us to close it? It's a question about human psychology, and so we'll study it with psychology experiments, not just with philosophy. Suppose you offer people a choice between cultivated pet food, which is a product recently approved for sale in the uk, and a traditional animal derived pet food. What information about the animals that were used to make the latter will actually shift people's real choices? And why, at a time when we as a society need to rein in our meat consumption for multiple different reasons, including environmental reasons, alongside animal welfare, what is holding back the wider adoption of alternatives to meat. Our goal is to produce work that will shape the animal welfare, labeling and public information campaigns of the future through a deeper understanding of how scientific information about animals influences human attitudes as we grow over time. These initial priorities, I very much hope, will be supplemented by new ones. But that shared mission to design better policies, laws and ways of caring for animals will remain constant and at the heart of everything we do. We'll always be doing research in the social sciences and humanities with the aim of benefiting other animals. And this is a promise to you that making now it will never be ivory tower research done purely for its own sake. We want to change the way humans relate to the rest of the natural world. Some might say this is an impossible dream, but that boundary between the possible and the allegedly impossible is one that we aim to move. So thank you all for listening and thanks very much for being with us at the very beginning of the Jeremy Coller Centre for Animal Center. So this is the point in the event where I go from speaking mode to listening mode because I've emphasized the importance of understanding the ways in which AI is impacting animals and what we can do to take the opportunities and mitigate the risks. And it's a real pleasure to be joined by four really important external collaborators of the new centre who are going to be working with me in various ways on trying to make progress on these issues. They are Jeff Sebo from New York University, Leonie Bossert from the University of Vienna, Jane Lawton from the Earth Species Project, and Kristen Andrews from the same City University of New York Graduate Center. And we know there's this huge space, this huge complicated space of ways in which AI is impacting animals. I've asked you all to think about one example that to you really crystallizes the issues we're facing. I think I'd like to start with you Jane, here because you direct this project called the Earth Species Project. We heard earlier from Jeremy about the Collar Doolittle prize, that is the, this 10 million pound prize on offer for cracking the code and achieving fluent two way communication with a non human animal. So of course I'm wondering here, you know, how realistic is this as a near term goal? If it is realistic, how are you going to get there? If you think it's not realistic, how close can we get?

F (32:27)

That's a huge question. Thank you. So just as a of an introduction, Earth Species Project is essentially using AI. Using AI, we're developing frontier AI to try to decode animal communication and we're doing it not just because we think we can. We definitely think we can, to your point, and not just because we think it's cool. I mean, a lot of people think this is very cool. In fact, when we tell them about it, the first thing they say is, when will I be able to speak to my cat or my dog? I think we are doing this because we fundamentally believe that communication is a really important window into the intelligences and the cultures, the minds, the lives of other creatures on the planet. And so we think that it's a really important way to reconnect human beings to other animals and, and to the rest of nature. Because fundamentally it's that disconnection that allows us to destroy, to exploit, to say we are different, we are separate from that is the other, and it's okay. So in terms of what we are actually doing, our unique niche as an organization is to develop essentially large language models to try to analyze the signal that are being sent by animals. So These are the GPTs of animal communication. We've just launched recently a model called Nature LM Audio, which is designed to analyze acoustic signals from other species. And the really interesting thing about this is we are looking at generalizable models. We're looking at models that, that can analyze communication across all species from a crow to a whale. We are also looking at models that can essentially help support biologists and ethologists in looking at tasks right across the spectrum, so everything from detection to classification. So we're looking to accelerate the science and through that advanced conservation. And at one point in the future, very soon, I hope to really create a shift in people's mindsets.

A (34:45)

I've got to ask you, say, generalizing from a crow to a whale, what does that involve?

F (34:51)

Well, I mean, we essentially this model, Nature LM Audio. Yeah, you can upload acoustic signals to the, to the model. It has very, very good levels of accuracy on species.

A (35:03)

It's been trained on recognizing the species.

F (35:06)

Recognizing the species, recognizing the type of vocalization that it's hearing. And the, the, this is the thing about AI. I mean these models are learning the emergent properties. We've seen is that it can actually, it can identify and then classify vocalizations from species it's never been trained on.

A (35:27)

Is this free? Can people use this?

F (35:29)

It is open source. It's just the initial user interface has just been open sourced on Hugging Face. So people can start to play around with it.

A (35:38)

People can use it. Right now it seems like there's a big. Well, not right now. Listen to the rest of the panel, there's obviously an ethical mission behind what you're doing. So you see positive benefits for other animals in doing this kind of work. What do you see as the moonshot?

F (35:55)

Completely? I mean, the whole purpose of our organization, organization is to try to exactly what you were talking about, transform the way human beings relate to the rest of nature. So our hypothesis is that if we can understand other animals better, we will be able to relate to them better, but we'll also be able to do much more positive things in the realms of conservation and animal welfare.

A (36:20)

I want to talk about the risks as well as the benefits. I suppose perhaps, Christian, you'd like to come in on this because you know, it's not unequivocally guaranteed to be a good thing. You think?

G (36:32)

Well, I think that if we're able to talk to the animals, then there's a responsibility that the humans have, the humans who've developed the technology, the humans who are having the conversations to listen to what the animals are saying back to them. And when you're listening, listening to a conversational partner, you are treating that other as an agent with the same moral standing as you have. So you don't get tiers of moral standing. You don't have humans being above the other animals. As soon as you have a conversation, you are implicitly acknowledging that moral status. But there's a question about whether human society, society is ready for that. And I think one thing I'd really like to see the Earth Species Project do and others who are working in this area is to take on, alongside the technological development work to help prepare society for this new technology.

A (37:32)

And Jane, do you think, is there any risk of being intrusive? I suppose, as it were, if you're communicating with an animal in the wild, inevitably that implies some degree of disruption to the animal's life.

F (37:44)

Yeah, I think there are all kinds of risks associated with this. And what's, what's interesting is that people around the world are already acknowledging these risks. We actually just did a survey, the first ever global survey of people around the world, more than a thousand people, 67 countries, to ask them what they thought about AI and animals communication. And they were both very excited, unusually open to the idea of potentially at some point in the future including other animals in even democratic processes, but also really worried about the potential risks of this. So I think one of the things we think is very important at Earth Species Project is to be careful not to jump immediately to the idea of two way communication. We believe really Strongly that the greatest value in the work we're doing is in understanding other animals. We not necessarily want to push it right now straight into that. We have to be in direct dialogue with other animals because that's where I think a lot of the risk actually comes.

A (38:51)

Recognizing the wildness is important, the rights of these animals to live their lives undisturbed.

F (38:57)

Absolutely, absolutely. And I think we're also really interested in one kick starting that conversation you were talking about. The regulatory frameworks have to start to be in place. This tech is moving very, very fast. It's going to be out there very quickly. How do we ensure it's used for positive benefit? We need the right regulatory frameworks. We all need to work together to try and kick start those frameworks. Yeah. So it's is definitely huge on our minds right now, but you've kind of got away, I think, the positive and the negative side to this.

A (39:29)

So we're thinking here about impacts of AI on wild animals. And Kristin, you know, you're very much like me in that you're a philosopher who works closely with scientists on empirical projects, including projects about using AI on to study animals in the wild.

G (39:47)

Yeah, we're interested in looking at how we can live better with other wild animals and minimizing conflict. So this field of human wildlife conflict is something we want to rename and kind of problematize. Because of course, if you're living in society, you have social living problems and if you're living in society with elephants around you, those social living problems are going to include elephants that want to raid your fruit gardens or your grain storages or who might be specifically India, the context. Well, we're working right now actually with wild street dogs in India is the case that we're going to be looking at first because the goal is to find whether there are social norms that shape the interactions between the humans and the non human animals. Using drone based imaging technology out of the University of Konstanz. And then we want to take that to look at the case of the tigers that are attacking villages in India, attacking some villages but not others. And so we're curious to find out whether there are social norms that some villages are following and others aren't.

A (41:03)

So this is project about filming the animals from the air over long period, generating huge amounts of video data from which you will attempt to use machine learning to extract the behavioral regularities, the rules that the animals are following that govern their interactions with people.

G (41:20)

That's right, that's right. And this technology is really robust, much More robust than the GPS data that we used to have when we're tracking animals, because you can see their postures, you can look and see who they're looking at, how they're communicating.

A (41:35)

It's showing us it's very much not just about the cause. The cause is the obvious thing you think about that could be analyzed with AI, but there's also the movement as.

G (41:45)

Viewed from above, all the body postures, all the morphology and all the interactions, because we can also then develop social networks and understand how these societies are structured.

A (41:58)

Yeah, I want to go to Jeff, perhaps about applications to farming. So we've been thinking about wild animals, translating communication and observing their behavior. One might think these are relatively benign applications of AI. Hopefully, in farming it gets a lot more controversial.

D (42:18)

Yes. And so this is not a project that I am personally doing, unlike the ones that they described. But the example that I thought I could discuss involves precision livestock farming. And this involves the use of AI along with other technologies in order to automate parts of industrial animal agriculture and aquaculture that previously required direct, manual human input. And so I think we all appreciate the context, and Jonathan briefly mentioned it in his introductory remarks. But to emphasize we are still in the midst of a very significant increase in the global population of farmed animals. We now have in order to feed an increasing demand for animal products. Conservatively, more than 100 billion farmed vertebrates per year. That includes cows and pigs and chickens and fishes, and Conservatively, more than 1 trillion farmed invertebrates per year. That includes cephalopod mollusks, decapod crustaceans, and insects, including black soldier fly larva. It could be as many as 50 trillion per year by the end of this decade, which is a staggering, unimaginable number. Now, animal farming at that scale, of course, is not compatible with the kind of idyllic individual farmer interacting with individual farmed animals picture that people still often associate with animal farming. We are no longer in a situation where an individual human can get to know two, three, or four farmed animals individually, understand their personalities, understand how much food and water to give them, monitor them for illness, and intervene immediately in order to. To treat them. So precision livestock farming is an attempt to recreate, to replicate that kind of individualized care for industrial animal agriculture and aquaculture at scale. It uses cameras, microphones, sensors, other technologies in order to ideally provide continuous monitoring of large farmed animal populations so that it can alert the farmers to when food needs to be adjusted, when water needs to be adjusted, when temperature, ventilation needs to be adjusted or when veterinary intervention is necessary. So as one example, there is a technology that will use microphones and other equipment to monitor the sounds at pig farms and detect patterns of coughing that suggest an illness is spreading. And some peer reviewed research show that in comparison to the traditional method, which is having some humans walk through every few days and listen for sounds of coughing along with other inputs, this does tend to lead to earlier detection and earlier intervention and reduced illness. So the claim, the hope, hope is that it will improve not only productivity, but health and welfare by allowing for more precision, more individualized care. Despite the increasing scale and industrialization of the industry. Obviously we can discuss whether that is really going to be the outcome.

A (45:41)

And some areas of farming are really far along this path as well. I mean, sometimes I speak to farmers about AI and they think I mean artificial and science insemination gets really confusing when I then start talking about super intelligent AI. But you know, other sectors, they're really, really enthusiastic early adopters of AI technologies. Salmon farming is very notable in this respect, where you have very high tech stuff, monitoring the salmon 24 hours a day under the water, and in one case even using lasers to try and shear off the lice that often infest these salmon. And these examples, they very quickly raise ethical issues, don't they? Because you see the opportunities and you see the risks as well. And it's very difficult to know how to strike an appropriate balance. What are your thoughts on this?

D (46:35)

Yeah, I think there are a lot of genuine ethical issues. I bet as a panel we could put together a long list off the top of my head. There are of course course traditional economic issues. For example, this could involve further centralization of power, further monopolization in this industry that can have all kinds of problems. It could involve a labor transition where a larger number of lower skilled jobs are replaced by a smaller number of higher skilled jobs. That can of course have benefits, but also costs for the workforce. But specifically with respect to animal health and welfare, there are a variety of problems here too. I will name two, and I bet others can name others. One is that unless this industry is regulated with effective oversight and enforcement, it is not necessarily going to voluntarily choose to optimize for animal health and welfare and rights and flourishing. It will optimize for productivity and efficiency and it will tell a story about how health and welfare and flourishing lives are happy byproducts of productivity and efficiency. Because when animals survive, they produce and so on and so forth. But that is not always the case. That is not often the case, especially in Highly intensive industrial animal agriculture and aquaculture settings where animals lived in cramped, toxic environments have very short lifespans, very unnatural lives, a lot of social disruption. So I am worried about the decoupling of what will be the goal and then what we would hope for in terms of health welfare rights. The second issue which is related is even if you did optimize for animal health welfare rights, flourishing lives, there is again a limit, I think, to how effectively you can accomplish that goal alongside public health and environmental and food security goals at scale in a global industrial animal agriculture and aquaculture system. And so the question for me is, while I acknowledge that this technology, if used thoughtfully, can mitigate some welfare risks in the short term, can reduce some welfare harms in the short term, is it going to, in the long run, build momentum towards true food system transformation or stall momentum towards true food system transformation by entrenching, reinforcing the perceived ethical legitimacy and environmental sustainability of industrial animal agriculture and aquaculture?

A (49:14)

Yeah. And to Leonie, you wanted to introduce the ways in which AI is being used in veterinary science.

H (49:23)

That's true. Like Jeff, this is also not like a project I'm conducting myself, but I wanted to highlight this example here because I think it also comes with a potentially huge impact on animals. So veterinary AI, AI used in the field of veterinary medicine. And of course this is used for both, both categories, farmed animals and companion animals. And I think if you are searching for AI that has the potential to benefit animals, at least in the current state, in the context of companion animal animals, there is the bigger chance that it will indeed benefit them because they are very often perceived as beloved family members. So obviously they have a computer completely different standing in our societies than farmed animals. So I think here truly there will be a big potential to benefit them. And AI in veterinary medicine is used in a variety of different ways, to name just a few. Like with farming context, it's used to for disease prediction and prevention, for example, with variables that have biosensors. And it's also used for diagnostic, for diagnosing. For example, studies have been shown that have shown that AI systems can successfully detect cancer cells in very early stages. That's also true for the human case. And often they can detect these cancer cells at earlier stages and more precisely than human doctors. So that obviously sounds highly promising. And cancer cells detection is just one diagnosing field where AI systems are successful. There are many more, and that sounds very promising. But as always, the using AI in the veterinary field also comes with a lot of ethical Challenges. And I just briefly want to mention one that is really linked to using AI and that is the issue of data. So data security, data availability and data quality. Because, for example, to train a model to detecting cancer cells you need a huge amount of data. And these is often not available. And especially it's not available in the sufficient quality that would be needed. And that could lead, if it nevertheless deployed, then it will lead to the product which is not as good as it should be. And we probably don't want to be used on our beloved family members. So there must be some regulation on not using these poor quality data for these products. And maybe to just mention a last challenge right now, what I think is also highly important in the this context you both mentioned it already is the question of care. Veterinary medicine obviously is related to care. And we have to ask, we as society, we have to ask ourselves what tasks of care we want to outsource to AI systems, especially in an existential context like duration of living and health.

A (52:42)

Yeah, I mean, there's certainly themes that cut right across all of these examples, including what are the kinds of caring relationships we want to stand in to animals. Is it possible for AI to mediate those relationships or not? I think there's also this fascinating connection between translation and veterinary care because we are already seeing apps that purport to translate your dog's emotions. I've seen one that purports to be able to tell you whether your cat is in pain or not by analyzing the cat's face. And yeah, the excitement of these things is real. It could be genuine benefits. I'm also really fearful because we know AI systems often have a tendency to tell the user what they want to hear. They're often biased in various ways. There's potential for this really difficult collision when these advancing technologies for decoding animal communication get adopted in veterinary contexts. And the veterinary profession has to make these big decisions about what it is going to admit as genuine sources of diagnostic evidence and what it is going to reject.

H (53:52)

Absolutely. And I think we have here kind of ambivalent situation that on the one hand, veterinary doctors say then like we all need AI literacy, they have to learn how to use these technologies. But also they still have to have enough of like animal literacy. They have to know enough about the animals that they are able to say how they potentially feel without using these technologies. And I think there has to change a lot also. And this is super interesting I think for us at universities as well, because we have to embed all of these in our Curricula.

A (54:33)

It's a big contrast with medical AI where the technology is advancing fast. But the ethical debate has also been very high profile and people are constantly talking about the ways in which algorithmic bias might have medical consequences for people. You know, can you trust AI or do you need human expertise in this? In the veterinary context, I'm just not seeing those debates at all yet at the same scale, but they urgently need to happen.

F (54:59)

Well, and it feels a lot more challenging because of course in human medical care you can ask the human whether it worked or not and how they're feeling, whereas you do not have the same ability with animals.

A (55:12)

So it's access the ground truth.

F (55:14)

You cannot access the ground truth. And so that's why I think one of the themes as well is how important it is that humans remain in the loop. One, because we don't want to create a situation where actually we create greater disconnection from other species. But also, you know, I mean, everything we do in terms of trying to understand other animals, this is not replacing human scientists. Right? I mean, there is absolutely no way we can ground truth without that. So I think that's one of the things that's really important to remember is that we can't take humans out of the equation. And in fact it would be wrong to do so. And I guess the other thing that I wanted to pick up on, that feels really important you talked about, Jeff, is just transformation.

D (56:00)

Right?

F (56:01)

We're talking a lot about practical applications of AI to make things easier on a day to day basis. But unless we can also seed really interesting radical new ideas of new futures, and that might involve how can we start to think about ways in which AI can discern the preferences, needs and desires of another animal to the point where their voice might be heard in a court of law as an example. Like those are the kinds of things that might take us into a new space. And it feels like we have to do those alongside some of the practical things.

A (56:37)

So we've emphasized we shouldn't be replacing humans. I want to put one final example out there, which is the possibility of replacing animals in animal research with AI. Again, you see huge excitement around this idea that we've seen technologies like AlphaFold that generate predictions of protein structures. A lot of people are saying, well, why not predictions of toxicity, predictions of carcinogenicity? We're already seeing technologies that are arguably equivalent to the kind of predictive value we get from testing on rats, because rats are not humans either. So maybe replacing the human in science. No, but replacing the animal. We can dream of this in animal research.

G (57:26)

Yes, there are a lot of alternatives in animal research and medical research. But I think if we want to understand animals and have relationships with them, we're not going to be able to substitute AI for the animals. We have to form relationships with the animals themselves. But I think we also have to recognize that we're doing it on their terms as well. Because one of the things, I think that that came up in all of our examples is that the animals, in fact, might not want us around. They might want to be left alone in some of these cases. They might have some privacy interests that we're violating by flying drones over them or recording what they're doing.

A (58:07)

I can't believe they're in favor of it.

F (58:10)

Well, we may get to a point with our technology where we understand the animals we study well enough to understand that they would prefer that we stopped collecting their data or that they might want to own their own data for their own benefit.

A (58:25)

It's a radically different kind of animal research that you're describing, isn't it? That you're saying what we've traditionally used animals for is as models of ourselves, as ways of trying to predict the effectiveness of drugs and so on. We can be phasing out those uses at the same time as continuing animal research of a fundamentally different kind.

G (58:48)

The kind of research that the primatologists were engaged in when they went into the field, lived with chimpanzees, and discovered that chimpanzees, in fact, are creating tools, do have social structures, do follow rules. So, yeah, it's a different kind of research, but it's a research that humans have been doing for a long time time in collaboration with animals who permitted them into their space.

H (59:13)

I think this also points to an important issue that has been mentioned already that, I mean, the decision that there is a need to replace animals in animal experimentation and that AI could be a very useful tool for that. That was a decision that is, well, hopefully built on, or can be easily built on moral arguments that we should not use them in these experiments. And probably we need to take these kind of decisions, like human decisions based on moral reasoning in the other fields as well, that we see enough reason to finally use this technology in a way that it will benefit the animals.

D (59:52)

Yeah, And I think one important point that all of these comments are bringing out is that there is no way of developing and deploying AI that in and of itself is going to have good effects on humans and animals and public health and the environment. There is no silver bullet solution to the problems that we are facing, the harms we are imposing on other species. And this is why this topic, I think, connects to some of the other topics that Jonathan, your center will be studying, including public attitudes and veterinary medicine. Because ultimately responsible development and deployment of AI is going to have to happen alongside other types of social and political and economic and technological interventions. Because even the most responsibly developed and deployed technology with good thoughtful regulations is not going to have good impacts, if anything, like current human societies are the ones to make use of it. And so we need that education and advocacy and these other types of change alongside better development of AI in order to get the good kinds of outcomes that people here are talking about.

A (61:03)

It's why social science focused institutions like the LSE is so important. I don't want to bring everything back to the LSE and know how brilliant it is. It's so important that the technology itself is advancing scarily fast, deployed in a chaotic, unregulated way. The impacts are almost certainly going to be bad. On balance, there's such an important role for social science and trying to understand how we can steer those impacts in a positive direction. People make this point often when it's humans, but it's just as true for non human animals as well. So really like to get some questions from the audience in as well. We've seen a lot of different ways in which AI is impacting animals and a lot of ways that social science and humanities research and ethical reflection can help us take the benefits and manage the risks. Questions about any of that, very welcome. I'd really like to take perhaps three questions at once and then come back to the panel. That way we get more questions in, in total. And let's start over this side of the room, five rows back there and then I'll just take two more.

F (62:12)

Hi, I'm interrupting this event to tell you about another awesome LSE podcast that we think you'd enjoy. Lseiq asks social scientists and other experts to answer one intelligent question like why do people believe in conspiracy theories? Or can we afford the super rich? Come check us out. Just search for lseiq wherever you get your podcasts. Now back to the event.

H (62:41)

Hi everyone, I'm Samarth.

A (62:44)

So my question is about.

H (62:46)

So I think, I think the risk.

A (62:47)

That I see with. With like the way, with.

H (62:51)

With trying to understand what an animal is saying through like monitoring it.

A (62:57)

Like the way you said, like Ms.

D (62:59)

Jane, like you spoke about what your work.

H (63:03)

At what point do you do we run the risk of anthropomorphizing what an animal is saying or doing. And at what point does that also then result in us institutionally being okay with animal exploitation? Because then you're saying that if you're in a position to say, if we.

A (63:23)

Are in a position to as humans, in an Anthropocene centric way, speculate that.

H (63:27)

Okay, this animal is saying yes, and then we do something, we don't really.

D (63:33)

We can never really be the animal.

A (63:35)

So we can never really, we can.

H (63:38)

Never really take a decision for it. So I think the solution, what we need to be, we shouldn't be looking at it in a welfare way, but.

A (63:45)

In an abolitionist way, which is to not use animals at all.

H (63:49)

And so in that case, I, I think AI like that, like that's all I want.

A (63:53)

Like, I wanted to know your thoughts.

D (63:54)

On this and the whole panel as well.

A (63:56)

Thanks. Yeah, great, great question. I am going to take three at once. So let's, let's go two, two rows up from there and then along. Okay.

B (64:08)

Yep.

A (64:09)

Hi, my name is Lionel. So kind of comment and question when we're talking about AI it, if I may.

C (64:18)

The elephant in the room is that.

A (64:19)

99% of the development and deployment is coming from commercial labs.

C (64:22)

It's dominated by three companies right now. Perhaps this center would benefit from partnership.

A (64:31)

With a commercial lab, some sort of engagement.

C (64:33)

You have Google, DeepMind, Upper Road in King's Cross, for example.

A (64:36)

And perhaps one of the most impactful subjects would be on AI alignment. I wanted to.

C (64:42)

Do you have any thoughts on AI alignment? Alignment means how models consider harm and ethical questions built into their thinking.

A (64:53)

And perhaps animal welfare isn't very highly weighted at the moment and can that be increased? Great. And let's take a third question. I want to get the other side of the room involved. Let's go to the, you know, we don't want to give all the ushers a chance to carry the microphone. Thank you.

F (65:12)

Hi. Thank you. I'm Ella. I was just wanting to ask, kind.

A (65:15)

Of in line with the first question.

F (65:17)

That was asked, what would be your fear of the co option of like communication with non human animals and then how that would be used in a possibly kind of exploitative or capitalistic system that we're currently in. So kind of touching on what you'd also already talked about and then continuing to. From that and opening that up kind.

H (65:35)

Of what do you see as being.

F (65:37)

The major issue caused by that and how would we possibly mitigate that? Thank you.

A (65:41)

Wonderful. So questions about anthropomorphism about our relation to big Tech and about our relation to capitalism. Great questions. I mean, who wants to go first? Geoff, you're just extremely eager.

D (65:57)

Well, I love all of those questions questions and we'll try to be very brief because I would love to at least briefly touch on all three of them, but I know others will have plenty to say too. So first of all, I really like your point about will this reinforce anthropomorphism and then give us license to continue exploiting animals if they appear to be consenting or assenting? I want to add to that by noting there could also be a risk of reinforcing anthropo denial if this communication translation technology really helps us be aware, makes more salient the difference between our ways of thinking and talking and those of other animals in a way that then leads us to see them as lesser than because we are still applying a human standard to who matters and how much they matter. So I would be really curious about strategies for mitigating both of those risks with respect to AI alignment. We are, and by we I mean our center at NYU and Jonathan and many others talking to the major AI labs and that includes anthropic and Google and OpenAI around animal issues. Around AI issues. We have the lines of communication open. That is not to overstate how much influence academics can have over these huge companies who have major economic incentives. But we are working on that kind of strategy in addition to public health engagement. And I agree that including sentience and animal welfare and AI safety and alignment is pivotal. We are going to be doing some work on that. And with respect to the third question about capitalism, I think that that is a really, really important problem. I am worried about the infohazards created by these technologies, the way that they could be used by actors with bad intentions to manipulate animals not to receive information and update our behaviors, but to send information and drive them out of habitat so that we can make use of them and so on and so forth. And I would like to see good strategies for mitigating those risks too.

A (67:59)

Yeah, I mean, if I can. I can't resist commenting. I mean, Jeff and I, we both worked extensively with people from Anthropic and from Google. In both cases, one of the things we've asked is can you please get some constraints on what your chatbots say about animals? Can you get something added to their constitution that is choosing the response of the different options that maximizes animal welfare or something like that? That's proved really, really hard to get for what to us seemed like a very feasible ask, met with surprising resistance. So this is something we have to keep chipping away at.

D (68:35)

Absolutely.

A (68:36)

But, yeah, I do want to get the other panelists in.

H (68:38)

Leonie, just briefly, maybe to the last question that also relates a lot to research ethics, that if we want to conduct open field research with communicating with animals in the wild, we have to know, or I think we definitely have to know what we are doing and not doing it until we are very sure that we won't cause disturbance to the animals. So we should follow the principle, do not disturb them. And there are currently, I think, no ethical guidelines for this kind of field research. And we need them.

F (69:15)

Yeah, I think, I mean, that's. I think that's really important. I mean, there is, there is a real risk with generative AI that could be used in playback experiments in particular, which is a method used by scientists to test the hypotheses they have about the meaning of a communication. And so I think it's really, really important that the ethical standards and certainly other ethics committees are thinking about giving approvals to field research, that they are now updating their standards related to the power of AI at this moment. But I also just wanted to reflect on the anthropomorphism question, which is, you know, has been a question time immemorial, right. When we're. We're trying to not influence our research with our own views. And I think there is actually potentially a benefit here. When we think about the models we're developing and their analysis of acoustic signals, the model doesn't know whether it's coming from something that's cute and cuddly or not. And so it has actually less of a bias toward wanting to have a human perspective. And then I think the other thing that I'd add, like we could, you know, there are so many different risks potentially around the misuse of the tech by bad actors. And we definitely need regulatory frameworks to get ahead of that. But I also think one of the biggest risks possible risks of this is trivialization.

A (70:49)

Right.

F (70:49)

It's almost like if the first thing that happens from this is an app on your phone so that you can talk to your cat or your dog, that is not going to lead us anywhere near the transformation that we are looking for in our relationship to other species. So I think what you were saying, Jeff, about just needing to make sure that we are advancing those kinds of ideas at the same time is really, really important in order to mitigate and outweigh some of the risks that will Inevitably come.

A (71:18)

Should we take some more questions? We'll do another three. I like that way of getting more questions in. Let's start this side of the room this time. Yeah. About five rows up. Sometimes hard to count those. Please wait for the microphone to come to you so that we can hear the question. Thank you. Hi.

G (71:36)

Hi.

F (71:36)

My name is Mara and I don't have much or any experience in designing AI models. And I'm just curious whether there's a.

G (71:44)

Way in which, if you design a.

F (71:47)

Model to pick up on pig costs or other expressions of discomfort, that at some point the AI model would pick up on the distress cries and actually learn that it's not just the CO cuff that matters to the pigs, but the pigs are saying, it's way too crowded in here. We don't want to be in here. So that they, you know, I don't know. Is there any way that you can.

H (72:09)

Give that model a heads up? Please watch out for these sorts of.

F (72:14)

Things so that the AI can work on behalf of the other animals.

A (72:18)

I don't know. Yes. Let's take one from this side. Yep. Hi, I'm Mike. I was just curious because my thinking.

C (72:29)

About how translation generally works is you have a kind of a shared basis.

A (72:33)

Of meaning between people or points of reference. And there's always a kind of leap of faith you're making understanding that the words I'm using can be understood by you because there's these shared reference points. And I'm wondering if, because of this notion of translation I have, whether it's ever completely be possible to translate from humans to animals, or if it is, does that require, like accepting inherently, like an anthropocentric notion of what language and what meaning is as well? Yes. Thanks very much. And let's take. Let's go to the back of the room. We've not gone right up to the back yet. The second row from the bottom. Thank you. I'm very much interested in artificial intelligence, but unfortunately I feel that artificial intelligence.

E (73:20)

And animal communication is still in the novelty.

A (73:25)

I was walking today and a met dog owner passing through with a dog, and I already know what a dog doesn't want or want. You don't need a machine to tell you that. I mean, how long have we had dogs? We've had them for thousands of years. And we already know that we don't want anything from a dog. It will tell you. But what also understand is that 90% of all food comes from the ocean. And the whole point of communication is that when we talk with between species, we're passing information between each other. I mean, when we're talking about philosophy. Do you believe that animals have egos? I didn't quite catch the last bit. Do you believe that animals have egos? Self. Because if they have consciousness, have a.

D (74:16)

Self or an ego.

A (74:16)

Okay, so some fascinating, some skeptical challenges to the Earth Species Project in that set of questions, as well as a question about animals and egos. Who would like to.

G (74:26)

I mean, I'm happy to jump in with some of those great questions. The last question is your comment about, well, we don't need a machine because we know how to interact with animals that we've co evolved with like dogs, I think is a great point. And it also goes to answer apart the second question, that we need a kind of common ground in order to get started with any sort of communicative system. So when we're interacting with animals over a long time, like humans have with dogs, we do have a certain starting point, but it ends to a certain extent. The dog can't tell us, oh, my tooth hurts and that's why I'm so lethargic, or I have a pain in my stomach. And there's some anecdotal reports that the technology of the dog buttons that Fluent Pet has designed actually lets dog guardians know what is wrong with their dog. And then they can get them treated a way that they wouldn't have been able to before. If there is a common ground that we can start from, and I think that there will be when we're able to form relationships with other species, then I think that this kind of radical translation project that we have to do when we're communicating across species is going to be possible, but it is going to be grounded in certain assumptions. We worry about anthropomorphic anthropomorphism, of course, but anthropomorphism, if it allows us to make predictions over time so that we're interacting with other animals over, you know, periods of weeks and months and years, then that raises our credence that we're getting it right. And I think that that's one way of avoiding that.

A (76:13)

Do you think sometimes there is no common ground? I'm thinking of the Collar Doolittle prize and the entries in the first year. Probably the most romantic entry was this cuttlefish study that was studying the arm movements of cuttlefish. They have these amazing swaying arm patterns through which they appear to communicate with each other. No one has any idea what these mean, what they're doing. They're absolutely fascinating. We, of course, want to decode them. What if there's just not enough common ground between us and the cuttlefish to make that possible?

G (76:46)

I think that that might be underestimating the kinds of relationships that humans can have with other species, if they spend enough time with them and are open to developing those relationships. I read things about people seeming to form relationships with species that are very distant from us. Well, we've seen my octopus teacher, for example. So you've got a cephalopod mollusk and a human apparently having a relationship, at least from one person's perspective.

A (77:15)

How much of that is just misunderstanding?

G (77:17)

Yeah, I don't know. In that case.

F (77:19)

Yeah, I mean, I guess my. Yes. And on that one would be that, you know, if you think about translating across human languages, even there is an interpreter in the middle of that.

H (77:31)

Right.

F (77:32)

I mean, this is not a direct. This means this. There is all always an element of interpretation, even across human languages and animals. Many animals occupy a completely different umfel. Right. Even their. Their sensory perception is entirely different from ours. So the idea that we will simply be able to have direct translation is actually a need for humans. What I think we're aiming for is a better understanding. So can we understand better what cuttlefish might be trying to signal to each other rather than saying this arm movement means this in human. I think we're aiming for the better understanding and potentially also needing to appreciate the mystery and the beauty in the mystery of the overlaps, the similarities between us and other species and the differences which will always be there. And so how can we cultivate that approach, appreciation and that understanding without needing the direct translation?

A (78:32)

So Jackie in the front row is in charge of the online audience. Jackie, do you want to give us like 3, 3 favourites from the online questions?

F (78:41)

Does this kind of research have an assumption about what animals already know and understand? For instance, if we assume animals want better conditions to. Does that not imply that they already know that there are other types of conditions?

A (79:00)

Yeah. Do you want to read two, two more as well and then we'll come back to them.

F (79:03)

Could AI help us to understand organisms of the deep sea, particularly given how other their existence is from us?

A (79:13)

Yeah.

F (79:14)

And is it necessary to believe the hype about AI?

A (79:19)

I mean, the answer to that last one is no, of course. We want to try and be, I suppose, critical friends of the industry, you know, willing to criticize as well as willing to believe in the opportunities. There's questions about, you know, could animals, I suppose, have what philosophers call adaptive preferences, Right. Where their Environment is not very good, but they don't know of any better environment, and so they end up preferring it anyway. That's a very real concern. And the question about are some organisms of the deep sea uniquely difficult to understand? Was so fascinating.

D (79:53)

Yeah, we worry a lot about adaptive preferences because a lot of animals, of course, do adapt to captive settings, to imperfect wild settings. A related issue is in this imperfect world where we are creating the entire world by and for, for humans, there are not good places at all for many animals. So not good alternatives at all for many animals. So not only do we have to worry about adaptive preferences in, say, captive conditions, but we also have to ask how should we assess these captive conditions, given that there are not right now, other options that are clearly much better? And so what is the right standard to hold our treatment of these animals to? Very briefly, I also want to make sure we respond to the first question from the last round, which I thought was really interesting. And it was about whether we can train AI to detect signs of distress and pain and suffering in farmed animals alongside signs of productivity and efficiency. And I think that we can. I also think that is not what is going to happen by default. And that is why we need regulation and oversight and enforcement. But it would be very interesting if an engineer at one of these precision livestock farming companies took it upon themselves to secretly train an AI system to detect signs of distress and pain and suffering so that they could be the new generation of undercover investigators and whistleblowers and tell us about what is actually happening in these viruses.

A (81:26)

I mean, these products are out there for cats. And in a way, that's a tragedy, isn't it? It's just showing that bias, that cognitive dissonance that we were talking about. The kinds of products that are being developed for cats and dogs fundamentally different from the kind of products being developed for farmed animals. And that's the kind of injustice we need to be getting farmed animals. The kinds of benefits that we expect for cats and dogs.

H (81:51)

If I may just briefly add to the question about adaptive preferences, I mean, we have the same with us humans. We also can adapt to a lot of things and then have this kind of adaptive preferences. But still, some circumstances we would evaluate as being a rights violation. And I think this is important to stress because it also relates to the question, if a full translation is ever possible. And it was already mentioned that probably also not not the case for all kinds of very different human languages. And I think that's quite important to keep in mind or to highlight that many of the claims we are doing in animal ethics also hold true for the intra human ethics. Basically because animals are not categorily different to us humans.

A (82:39)

Wonderful. I think I don't want to keep you. And after this event we have a rest outside where we can celebrate the launch of the Jeremy Collar center for Animal Sentience with LSE Catering's finest plant based offerings. I'm really looking forward to seeing what they what they come up with, and you're all very welcome to stick around with us to celebrate that moment. Thanks very much for being here for this event. Thank you for listening. You can subscribe to the LSE Events podcast on your favourite podcast app and help other listeners discover us by leaving a review. Visit lse.ac.ukevents to find out what's on next. We hope you join us at another LSE event soon.