The Current - What are animals saying? AI is helping to decode

Episode Date: May 27, 2025

Do you speak dolphin? What about marmoset or nightingale? Did you know cuttlefish use a form of sign language to communicate? If your answer was no to all those questions, you should know that scientists are working to turn that into a yes — and AI is playing a key role. How to decode animal communication, and whether that's even a good idea. Three animal communication scientists join us to talk it all through - human to human.

Other People's Problems was the first podcast to take you inside real-life therapy sessions. I'm Dr. Hilary McBride, and again, we're doing something new. The ketamine really broke down a lot of my barriers. This work has this sort of immediate transformational effect. Therapy Using Psychedelics is the new frontier in mental health. Come along for the trip. Other People's Problems Season 5, available now. This is a CBC Podcast. Hello, I'm Matt Galloway and this is The Current Podcast.

You probably don't speak dolphin. Or Marmoset. And you probably don't understand the call of the Nightingale either. Well the new prize aims to help humans better understand animal communications. The inaugural Collard-Doolittle Prize for Two-Way Interspecies Communication was recently awarded to scientists working with dolphins. The other finalists were teams looking at nightingales, marmosets and cuttlefish. The top prize went to a team of scientists from the Woods Hole Oceanographic Institution and the Sarasota

Dolphin Research Program. And while we might not yet understand dolphin or marmoset or nightingale, artificial intelligence is actually helping us get closer to unlocking animal language. In a moment, we will hear from the lead judge of this $100,000 prize as well as a finalist. But first, I'm joined by a member of the prize winning project, Franz Havmann Jensen is a senior researcher in the Department of Eco-science at Aarhus University. Franz, good morning.

Good morning, Matt. Thank you so much for having me on the show. Congratulations on your win here. Thank you. Yes, it was quite unexpected. I think we're all very surprised. Why did you want to look at understanding what it is that dolphins are saying? Well, we've been working to understand dolphin communication for a long time, actually.

So over the past 40 years, Leila Saik at Wurzowszographic Institution, together with other colleagues, including me, Peter Taak and Vincent Janek from the University of St. Andrews have built a unique database of whistles from more than 300 button-nosed dolphins in Sarasota Bay. And our objective there is really to understand their rules of communication, so understand what the structure, function, and meaning of dolphin communication is. So, not just identifying the sounds they make, but uncovering what those sounds mean to them. How did you go about—you said you have this archive of dolphin whistles and calls. How did you go about recording them?

Most of the data comes from suction cup hydrophones. So these are underwater microphones that are placed directly, in this case, directly on the melon of animals that are studied in Sarasota. And this allows us to really collect data that we know are from specific individuals. Before your research, how much did we know about the meaning of the sounds that dolphins make? Well, a lot of research over the years has focused on signature whistles. So,

signature whistles are individually distinctive signals that animals or that botanist dolphins have. It's the dolphin equivalent of a human name. Dolphins use these signature whistles to maintain social bonds, recognize each other. I think what is really new here is that some 10 years ago, while Layla and colleagues were testing how dolphins could recognize each other, they were playing back non-signature whistles, whistles they didn't really know what the function of them were. And they were trying to see if dolphins would respond differently to familiar voices or unfamiliar voices. They didn't respond differently. So, but unexpectedly several dolphins responded with a very distinctive whistle.

So your team has essentially determined, as I understand, that the meaning of a couple of different dolphin sounds. And we have recordings of these sounds. I want to play them for you and get you to talk about what we're hearing. Have a listen to this. That's a slowed down version of a dolphin call. What have you learned about what dolphins are trying to communicate with that noise? So this signal here, which consists of a sharp up-down-up sweep, seems to almost always elicit a very strong avoidance response. So our hypothesis for this non-signature whistle is that it likely functions as some sort of alarm or alert signal.

As a signal to other dolphins to stay away or to express kind of concern about something that might be nearby. Exactly. That is so interesting. I wanna come back to that in a moment. I mentioned that there were a couple of sounds. This is another dolphin sound that your team tried to decode. Listen to

this. That went by very quickly. So let me play it one more time. As you understand it, what does that mean? So we're hypothesizing might function as a way of to signal curiosity or query. And many of the animals that were exposed to these whistles demonstrated or replied with this non-signature whistle, almost as if trying to figure out who was there. And remember, these animals can't recognize each other from voice cues. They really depend on this secondarily evolved signature whistle to identify who's there. So we think this might be some sort of an a query or curiosity signal that kind of elicits or tries to get other animals to

respond with their signature whistle. How do you know that that is what, I mean you said that these are, this is a hypothesis, how do you know that that is what these dolphins are trying to communicate with those sounds? Well that is hard, right? So we're trying to investigate this through playback experiments where we go out, we follow animals, we measure animal responses using drones from the air or using these sound and movement recording D tags. And so we're trying to investigate these different hypotheses by playing back both of these two stereotype signature whistles as well as a couple of controls, both familiar and non unfamiliar signature whistles as well as a couple of controls both familiar and non unfamiliar signature whistles to animals. How was artificial intelligence

helping with your work? You mentioned the database and one of the things that AI can do is crunch enormous amounts of information in a very short period of time. How are you using AI to help with this? Well that is certainly one of the main applications. So we build this huge database of hours and thousands of whistles over the past 40 years. And so one of the biggest challenges that we're facing is the sheer volume and complexity of acoustic data, like thousands of hours of audio data that needs to be processed. So AI is now helping us both automatically detect whistles and we're using AI also to

uncover new or discover new shared whistle types that we don't yet know and looking into how using it for identifying patterns of use across individuals and contexts so that we can begin to infer meaning from how dolphins use these signals. Tell me more about that just finally, the infer meaning part of it. I mean, it's neat to be able to hear the different sounds and perhaps hypothesize as to what they mean, but why is that important for us to be able to better understand how they are communicating and what they are communicating. I think to some degree it's important for us too because we're curious about our natural environment. We're looking at a lot of different species to understand communication and by understanding how communication has evolved across many different species,

we can better understand the evolutionary roots of communication and of language. Part of it is also that it may actually help us improve dolphin conservation. Already now we're working on better ways where we can identify and recognize these individually distinctive signature whistles that really can be used for transforming population monitoring of dolphins. And some of these stereotype non-signature whistles may also help us become better stewards of our natural environment. This is so interesting.

Franz, thank you very much for this. Thank you for having me. Franz Hovhund Jensen is a senior researcher at the Department of Eco-science at Arhaz University. Not all of the research projects up for the caller do little price studied animals that communicate with sound. One of the shortlisted teams looked at cuttlefish and found that cuttlefish use a form of sign language. Sophie Cohen-Bodenes is a researcher

at Washington University in St. Louis, Missouri. Sophie, good morning to you. Good morning, Matt. Thank you for having us. For people who don't know, what is a cuttlefish? What do they look like? So a cuttlefish is actually not a fish. It's an invertebrate. So it's a fascinating animal that has the unique ability in the animal kingdom to be able to do dynamic camouflage. And so it will produce

with high fidelity the colors and textures of what it sees in the environment. And so this ability is also fascinating because it's versatile and that cuttlefish also use this dynamic skin patterning for communication. And so what we found in our studies that cuttlefish also use sign language to communicate. explain that how How does the cuttlefish use sign language? So we were acquiring some data. And so at some point I observed one of our cuttlefish expressing like very strange arm movement. It seemed like the cuttlefish really was expressing something.

And so after some extensive observation, we identified four different types of arm movements that we documented and we conducted two experiments to attempt to prove that there are communication signals. What are the four different arm waves that the cuttlefish does? So we named them according to their features. So there is an up sign where the coddlefish is extending the arm above the head on the vertical direction. We have the side sign. We call it side because it has some form of laterality, which means that the coddlefish can enroll the arm on one side or the other of the body. We have the roll sign, which is not expressed

very often, but it's very striking in its features because the cuttlefish enrolls the arm beneath the body and it changes actually the shape of the head and the eye. And at some point we were thinking that it would even resemble an octopus. And finally, the crown sign so refers to the arm crown of the animal. The animal is sort of spitting the crown, usually in repeated sequences. And what is really fascinating is that we saw that those signs can be combined and that they are also expressed along with skin patterns for communication. Do you know what these four types of arm waves from the cuttlefish might be saying? Okay, so we are still in the process of deciphering their meaning. So our first experimental

evidence shows that it's a sign that they would use to interact with one another first and that they could also use it to interact with prey. For example, when we insert the live prey in the tank, they would often display the crown sign, so it could be an aversive signal to prey or predators. But I think that they encrypt a variety of meanings, like for example, the up sign is usually expressed with zebra patterns and zebra patterns are patterns used during courtship. We still need to do more experiment

in various behavioral contexts to be able to decipher their meanings. What is this? You showed cuttlefish, a video of other cuttlefish doing different signs, is that right? Yes, absolutely. So it was an experiment that we conducted to test the hypothesis that there are communication signals that they use to interact with one another. So the experiment was quite simple because we recorded several instances of cuttlefish spontaneously signing, all the different types of signs expressed by all different types of individuals, male, females.

And we did a visual playback experiment. So we showed them, we displayed them on the screen and naturally they were curious about it. They would come themselves in front of the screen, and they would frontally to the screen display the different combination of signs. It's like they were communicating to each other. I mean, we believe that it's possible

that they are interacting as if they were interacting with other cuttlefish. It could also be like they recognize their own sign and that they possibly understand that we expect them to sign back. So we are still trying to prove indeed that it's a communication signal, but we have established specific controls that shows that they recognize specific social cues in the signal and that indicate that it is possibly communication signals indeed. How has artificial intelligence helped with your work?

Since I'm telling that we are still in the process of deciphering the language, we really capitalize on the power of artificial intelligence because artificial intelligence helps us to decipher patterns, those hidden patterns. And so we would like to, we are in the process of collecting large, large, large data sets. And in many behavioral contexts, to give those data to artificial intelligence algorithm that could find in an objective ways, the different correlations between different arm signs and, for example, between different skin patterns. For example, sometimes you have a coddled fish that

is doing a chrome plus orange plus black spots. And sometimes in other contexts, coddled fish would do like up and zebra. And so an algorithm could tell us, okay, so zebra correspond to corry sheep, and it could tell us like, for example, crown is for adversity context. That's wild to think of the various combinations. But just finally, what do you think we learn when we are able to get closer to understanding what these animals are communicating to each other? What is it that we learn from that? That we learn from that, actually, it's very fascinating because carapace, other researchers that use this metaphor to say that they have a neural net that we can see on their skin. It means that their skin is

almost like a window to their internal states. And so when you see the skin patterns of the color fish, they tell you a little bit what's in their mind. Why it's important for our research is that it's a way to better assess their welfare, to better understand their needs and to, I mean, my personal goal is to improve their protection, possibly their protection in the wild, but also in captivity. So I think this is very impactful. that I think so, I think this is very impactful. It's fascinating. Your brain kind of melts as you think about what the cuttlefish may or not be saying. Sophie, thank you very much for this. Thank you so much, Matt. Sophie Cohen-Bodinez is a researcher at Washington University in St. Louis, Missouri.

For a brief moment in the year 2000, the city of Phoenix was on fire. You could see a glow in the sky. An arsonist was on the loose. How many more ways can we fuck this thing up? The culprit will shock you. I had several dreams about that house behind me. Your dreams about setting it on fire one night I did. From Novel and Sony Music Entertainment, listen to The Arsonist Next Door.

Subscribe on Apple Podcasts to binge all episodes now, or listen weekly wherever you get your podcasts. Jose Ovelle led the judging panel for the Caller Doolittle Prize in Two-Way Interspecies Communication. He's a professor of zoology at Tel Aviv University, and we've reached him in Montreal this morning. You'll say hello to you. Hi, good morning, Matt. The ultimate goal of this prize is two-way multi-context communication with animals using the animals' own signals. What does that mean?

That sounds a bit enigmatic, I agree. Yeah, the idea is first of all to decode animal communication. So just to understand what you know, what types of signals they're using, what kind of messages they use, how do they concatenate signals in order to create what we would call sentences if this was human language. So that's the first step. Once you get some understanding, once you think you understand this, the next step would be to prove that indeed what you found is important for the animal. So if we think of, I don't know, these cuttlefish that you've just discussed, for example, the next step would be to present signals that you've just discovered to the animal and observe

their response and to show that you can do this in multiple contexts. So we want to show that you really decoded the animal's communication system in multiple different meanings. The chair of the foundation that leads this prize has compared artificial intelligence You really decoded the animal's communication system in multiple different meanings. The chair of the foundation that leads this prize has compared artificial intelligence to the Rosetta Stone, saying, just as the Rosetta Stone unlocked the secrets of the hieroglyphics, I am convinced the power of AI can help us unlock interspecies communication. We've heard how the researchers are using AI.

How do you think that AI can help with this? Yeah, so I might be a little bit more skeptic than Jeremy Koller, who's indeed the head of the foundation. I think that AI is definitely improving our ability to decode animal communication. So it allows us to take these huge datasets and to seek for patterns. Whether this will be enough in order to generate a machine that an algorithm that can talk to animals like you and I are talking right now. I'm a bit more skeptical about that. A, I think as I said before, I think we need to find ways to measure the animal's response. And that's not always clear. You can

think of yourself, you know, I could say different sentences to you. You might be processing them completely differently inside your brain, but when measuring the output, when looking at you, you know, it will be hard to measure a different response. So that's one obstacle. And second major obstacle is that I'm not sure that animals have the type of complexity, their communication system is not as complex in our language in a way that, meaning that you and I are using words and we're concatenating them into sentences and this allows us to discuss any subject that we would like to and I am quite sure that animal communication is not as complex, so we might be able to communicate with animals, but talking

about talking with animals, I'm more skeptical about that. But that's, I mean, it's called the Doolittle Prize, right, in some ways, and it's like Dr. Doolittle, if I could talk to the animals and walk with the animals, what have you. Peter Gabriel, who was part of the team that helped come up with the concept for this prize, he was interested in this because he played music with bonobo apes. Why is it, what is it that you think is so interesting to us about the idea, not just of understanding what these creatures might be saying, but in figuring out some way that we might be able to communicate them?

What's that about? Yeah, yeah, well, I think, you know, I can answer as a human being or as a scientist. As a human being, I mean, of course, well, a scientist is a human being, of course, but, you know, as a human being, of course, but, you know, as a human being, you know, since I was a child, I always wanted to understand the animals around me, right? Whether it was my pets or the birds sitting on a branch or the frogs that I sometimes caught in the pond, right?

I mean, we have this fascination, you know, the literature is full about, there are a lot of myths about, you know, King Solomon, Dr. Doolittle, whoever, Aristotle, I think, also who could speak with animals. So, I mean, we've always been fascinated by this. As a scientist, I can say there are different reasons. One of the major ones that is driving me is simply this curiosity about the origin of our own human language. So we have, as I said before, we have this language which is extremely complex

and it seems to stand out in comparison to other animal communication systems. But you know, in order to understand whether it stands out and how big is this gap, we need to understand more what's going out there in nature. And the more we study, the more we appreciate animal complexity. That's one of the main goals, I think, of this prize. Do you worry, as a scientist, and maybe as a person, a human being who's not a scientist, about anthropomorphizing them? I mean, we're using this language, right?

We're talking about language, we're talking about names, we're talking about sign language. We don't know, to your point, whether these are just forms of communication or whether they are an evolved language, but we want to think that because that allows us to create something in our mind. Do you worry about that? Yeah, definitely.

Definitely. And when I talk about my own work, we study bats in the lab and I always emphasize that when we talk about meaning or context, we always do this based on our own human observations and that could be limited and it could even be wrong. So yeah, definitely, as scientists, we have to be objective and to try to find criteria that are not dependent on our own human bias. But during the past decades,

I would say until a decade or two ago, maybe there was a movement or a notion among scientists that we can't talk about, we cannot use human terms at all when we discuss animals. And I think scientists went too far with this notion. Once you have to prove everything, then you cannot do science at some point anymore. Just before I let you go, just because we can do something, there are people who would ask whether we should do something.

I might want to understand what my dog is saying when my dog is barking away at me. But for dolphins, for example, there can be an element of control that can come out of this. And we've already, in many ways, altered or destroyed parts of their habitat. Should we try to understand or communicate, if that's the long extrapolation of this, with animals? I think that's a very good point and I've been thinking about it a lot and more and more during the past year or two, especially when we have now also commercial organizations

who are interested in this. If I talk about my own motivation, it's all about the conservation and understanding. And I would never, I think I would never misuse this ability to harm an animal. You may not be everybody who uses that technology. Exactly. So exactly. So I can completely see how, you know, maybe even a birdwatchers, it doesn't have to be something really malicious. You know, you could think of birdwatchers who want to lure birds to some kind of a point in order to photograph them. So I can definitely see how this technology can be exploited.

And we should clearly think about how to limit this. I should say, you know, that this it's always been possible. So, you know, hunters poachers always use the animal signals in order to lure them. This is not new. Yeah, AI is definitely, it's an amplifier as always, but this has always been around and maybe this is a good time to think about it. This is really, really interesting.

And that last point is fascinating as well. Yousef, thank you very much for talking to us. Thank you very much. Have a good day. Yousef Ovelle led the judging panel for the Collar Doolittle Prize in Two-Way Interspecies Communication. He's a professor of zoology at Tel Aviv University. You've been listening to The Current Podcast. My name is Matt Galloway. Thanks for listening. I'll talk to you soon. For more CBC podcasts go to cbc.ca slash podcasts.