Aza Raskin: Nature Speaks. Can AI Help Us Listen?

Aza Raskin — Keynote: Nature Speaks. Can AI Help Us Listen?

• Speaker: Aza Raskin, Co-Founder, Earth Species Project🔗
• Event: SXSW 2026, Mar 16, 2026, 1:00pm–2:00pm CT
• Venue: Hilton Austin Downtown🔗, Hilton Grand Ballroom (Salon HJK)

Official SXSW description

More than 8 million species share our planet. Until now, we’ve only understood the language of one. In this keynote, Aza Raskin, Co-founder of Earth Species Project, explores how AI is being used to decode the languages of other species—revealing complex, learned communication across the animal kingdom. For the first time in history, we can truly listen. The question isn’t if we’ll break the interspecies barrier. It’s whether we’ll do it in time to change everything.

Transcript: Aza Raskin

Hello, humans. You all doing good?

So I’m here today to talk about our interspecies project. I want to start with this sort of thought: What is the connective thread that unites all the largest problems that humanity faces — from climate change, to growing inequality, social media, traffic?

It’s that we continue to do narrow optimization at the expense of the whole.

We focus on just the GDP thing and not like, the technosphere eating the biosphere thing. And that’s just another way of saying disconnection. That we are disconnected from ourselves, from each other, and our systems, from the natural world. And really a lot of what Earth Species is about is, yes, learning how to understand the language and communication that of the other species and cultures of Earth. But more fundamentally, I think it’s this: it’s about this form of reconnection with ourselves to the rest of nature.

So, oh, and I should just say I’m standing here, but there’s a whole team. This is actually not our full team, it’s just the group that showed up on our last Zoom call. saying hello to everyone.

Guessing game time. Who here knows what animal makes this sound?

Any guesses?

Not a whale. Good guess. Other guesses? Not a penguin.

It is a seal. It’s a good one. This in fact is the— You will always get outstayed by animals.

This is actually though the mating call of the bearded seal, so if you got excited, now where to go. And the reason I like starting here is because when we think we know what animals sound like, we actually know just a tiny little fraction.

And it turns out a lot is already known. So these are a green parrotlet. It turns out— I’m not sure so many people know this— but it turns out that parrot parents Mamas spend the first couple of weeks leaned over their chick whispering a unique name to each of their chick’s ears, and they will use that name for the rest of their life.

Turns out elephants do the same thing, dolphins do the same thing, belugas do the same thing. The more we look— the first marmoset monkeys just this last year it was discovered that they also have names they call each other by. And names are of course one of the most profoundly important parts of my identity that people just don’t know.

We call this like the most humble graph that only includes all the time and all of space. But the reason why we put this graph up is because that little dot up there, that is the human experience. That’s what we can experience as humans. And we want you to stop and think about this for a second, feel it with me, because it means that most of the universe we cannot experience. Most of what is, we are blind to.

The world is vastly wider and vastly more vivid than we can experience. And one of the few ways for us to get outside of the box of our human mind, of our human brain, is the ability to understand other species that are experiencing a little bit more of that world. And the best way to do that is through the window of communication. Our ability to understand is limited by our ability to perceive, and what AI can do in best forms is throw open the aperture of our ability to perceive, throw open the doors of perception.

I want to talk a little bit about some of our work. So these are crows, and we’re working with biologists at the University of Lyon, and they have little backpacks on these crows. It turns out these crows are very interesting. They have unique family structures, and normally crows, they raise their chicks in pairs, but here they have a unique culture where multiple families come together to do communal childbearing. It’s almost like a commune.

And they have a unique vocabulary and unique dialect that they use to raise their chicks. They take outside adults in, they will teach them that vocabulary, and they will start participating in that culture. So they use “rach.” And one of the things we have recently discovered is, as our AI started to map all of the vocabulary of the crows, there was There was a whole bunch that the scientists had never told us about. So we went back to the scientists and we’re like, “Hey, you never mentioned that there’s this huge area of communication. What is that?” They’re like, “We don’t know.” And it turns out that, between 60 to 70% of crow communication, they are quiet, close, intimate calls that they make while they’re flying.

So these are the equivalent of So it was a whisper. Think about that. Like, we’re in 2026. An animal that we see all the time, we hear all the time, it turns out that the majority, the supermajority of their communication is below the surface. This is where your ears should start to perk up.

Just to give another example of this kind of thing, I think it’s fascinating, we’re talking a lot about other organisms they send back to a school, almost all recorded, by the way. Like, the actual data collection comes through biologist partners. And she was describing what’s known about dolphin communication. They just mentioned dolphins have names they call each other by, even in the third person, and they’re called signature whistles. And when she was showing what is known about dolphin communication, there’s sort of like a pie chart.

On this side of the pie chart are signature whistles, i.e., names. And it turns out if you talk a little bit more, there’s lots of variation. There’s a whole bunch in there that people don’t understand. Let’s just lump it and say half of their communication are names. The other half of the communication scientists call non-signature whistles.

This is as if, for humans, somebody’s trying to study us and they’re like, OK, so they use their names and then everything else they say are just non-names. That’s the level of understanding we have in 2026. This is why understanding animal communication, we think, is the next frontier. OK, so we’re talking about animal communication, but how about plants?

We just saw an amazing plant wave concert. Do we think plants can communicate? Do we think that they can hear? Well, obviously, by the nature of the way I’m asking this question, the answer is going to be yes. But there’s a University of Tel Aviv study, I believe this was 2019, where they played different sounds to flowers.

So in this case, a primrose flower. They played high-pitched noises, low-pitched noises, and only when they played the sound of an approaching pollinator, a bee, did the flowers respond by producing more and sweeter nectar. And not slowly, they did it within around 3 seconds. So quite literally, the flower hears the bee coming and gets excited.

There are examples too of plants communicating back the same way. Lab stressed out tomato and tobacco plants, and it turns out the more stressed they were, the more sound they emitted in the ultrasound. And not quietly, at the sound level of human speech. And so that means if we could open the doors of our perception, the aperture of our ears, and we walk up into nature, we would actually be able to hear fields communicating. I just think that’s so beautiful.

Like, the experience of a cat out in the field is probably one of intensity, if it’s in the dry season. But one more thing about plants, because I can’t help myself, is, this was only discovered in 2013. Scientists were working on a particular vine, actually it’s a fairly common vine in South America, and they noticed that it seemed to mimic the leaves of the plant that it was growing on. So if it was growing on a red plant, it would grow red leaves. If it was growing on a green plant, it would grow green leaves.

If it was growing on a big leaf, it would grow big leaves. And biologists had no idea how they were doing it. It’s like, well, maybe it’s like sensing chemicals and it’s pre-programmed and has just a couple different plants that it can mimic. And so somebody in 2020 had the brilliant idea of like, let’s grow plant, this vine, on artificial plants. What do you think happened?

Same. It was able to mimic the artificial plants. So the only thing that biologists have not been able to rule out are what biologists call a “sally,” and that’s a face-saving term for “bias.” They think that the plant can see and can mimic without a central nervous system.

So the takeaway is that the world is awash in communication, in signal, in intelligence, and meaning, and it is just our limited view that is keeping us from recognizing it. OK, but let’s get into OK, maybe there’s meaning and signal, but is there really communication happening? How deep could that be? And there was a 1994 study from the University of Hawaii. That is, I think, one of our favorites.

And in this study, they taught dolphins two gestures. The first gesture was, “Do something you’ve never done before.” And the dolphins will innovate. And think about how crazy it is to be able to communicate that, which is they’re going to have to remember everything they’ve done before in that session, understand the concept of negation, “not one of those things,” and then come up with something whole cloth. Something random. But the dolphins can do it.

And then they’ll teach the dolphins a second gesture: do something together. And they’ll say to different dolphin pairs, “Do something you’ve never done before together in a bay together.” And the dolphins go down, exchange sonic information, come out, and they do the very same thing at the exact same time. So you’re like, “OK, that doesn’t prove language, but I now sort of feel that the burden of proof is on the other foot.” OK, so hopefully now you’re like, “OK, the world is awash in meaning and signals. There seems to be a very real ‘there there’ in terms of like the ability to communicate something rich and symbolic, but how can you possibly go about translating a language if you don’t have a Rosetta Stone or any examples really of translation. And that was impossible up until the year 2017, when a new technique came out of deep learning, AI, that let you translate languages without the need for Rosetta Stones.

And I want to walk everyone through how that works, because I also think it’ll boot up an intuition for why AI is going so extremely quickly now, and we’re seeing advances in every field at once. So this is how it works. This is English. This is how an AI sees English. The technical term is an embedding space if you want to look it up later, but I think of it as a galaxy or a shape.

In here, every star is a word, and words that mean similar things are placed in similar positions, and words that share a relationship share a distance and direction. That is, relationships get turned into geometry. So in this shape, king is to man as woman is to queen. So king is the same distance direction to man as woman is to queen, as girl is to princess, and it’s actually just princess minus girl, and then that’s a boy, a little prince. It’s sort of a— it’s a very fun thing to go play with.

But if you think about it, then, if you take the point which is dog is one spot in space, and it has a set of relationships to friend, to howl, to wolf, to man, to fur.

It sort of fixes the point which is dog in space. And if you solve the massive multidimensional Sudoku puzzle of how every concept relates to every other concept, you end up generating— a rigid structure that represents language. That’s the first thing you saw. And the AI doesn’t know what anything means, but it knows how everything relates. And then this was the insight.

This was actually when— there were two papers that came out literally back to back within a day of each other that when I read it was like, now is the time to start Earth Species Project. Now is the time to start working on translating animal language. And the thought was this: can you take the shape for English and the shape for Spanish and match them up? And if you talk to linguists or anthropologists, they’d be like, “No, that won’t work. There’s such different ways of viewing the world, different conjugations, different verb structures.” But machine learning is like, “Whatever.” GPUs go grrr, and they line them up.

And indeed, the shapes are a little different, but the point which is dog ends up in the same spot The overall shapes are the same. And that wasn’t just true for English and Spanish. He’s like, OK, English and Spanish, those are super-related languages. It also works for Japanese. And it also works for Aramaic and Urdu and Finnish, which is a weird language.

It turns out every language fits in a kind of universal human meaning shape.

And you can translate to and from all languages in this way. And to me, this is so profound and so beautiful because it means that right now I would argue technology is doing more for like tearing us apart than putting us back together. And yet here technology is showing us there is a hidden underlying structure that unites us all.

Obviously the next place of thought that goes from here is like, OK, can you build this shape for animal communication? But if we’re going to build the shape for animal communication and line it up with the human shape, that seems like a promising idea, but I don’t know, like whales, like sperm whales, spend 80% of their life a kilometer deep in the ocean in complete darkness, seeing with sound. Like maybe there’s not enough that’s the same about our experience to be able to actually do the translation by matching shapes. But it actually goes much deeper than this, and I think this will give a really profound sense of like what’s going on in AI right now, is that this doesn’t just work for languages. This also works for essentially every modality.

So I’m sure everyone in here now has played with like image generators. You type in text, you get an image with AI. Probably. How does that work? Well, that can give you an intuition for how it works.

You can ask it to build the shape for languages, and then I can also ask AI to build the shape for images. So now every dot is an image, and relationships between images become geometric. So there’s a version of your face, version of your face was just smiling, that’s the distance and direction. I add that to any face, you get the smiling version of that face. You then align these two shapes by picking image-action pairs.

It’s no longer just a rotation, but it’s roughly that. And now you can correspond, you can translate from image to language and back again. So, “Portrait of a Chilean person” is this spot over in language space. I translate it into image space, and I say what image goes with it. That’s how that works.

But it goes deeper than that. It’s not just images. Everything. So here’s an example from, I think, 2023, where they put human beings inside of an fMRI machine and had them look at images that build a shape, shape, but to map them to each other and then be able to recreate what the human is seeing only by looking at the fMRI. So now the AI can only see the fMRI, it doesn’t see the image the person is looking at.

And it can reconstruct this.

How does it do this? This is awesome and terrifying. OK, thank you. Means that because your visual cortex runs in reverse when you dream, that means at some point we will be able to image dreams, which is cool and also terrifying from a security and state perspective. We’ll set that aside for this moment.

How does this work? Exactly the same way as everything else. You build the shape which represents FMIs, you build the shape for images, you correspond them, and you do this. And so why everything’s moving so incredibly quickly in AI is because everything can be represented in these kinds of shapes, and they can all be mapped to and from each other. From fMRI data to communication to DNA to robot Motion is just a different language.

DNA is just a sequence of letters, a different kind of language. Music is a different kind of language. Everything can be viewed as a kind of language that has been translated to and from. This, by the way, was my face when I first figured this all out. And so there’s actually something very, very profound ground here that lets us believe that we’ll be able to do this kind of translation.

Which, if you think about it, AI is learning the same sets of shapes for everything. Actually, one of the deepest results recently is that there’s a paper called The Platonic Representation Hypothesis that no matter which AI model you’re using, those shapes that it’s learning to represent text or images, they’re all converging. They’re all learning the same thing. And the more modalities you add, the more images, the more videos, the more DNA, the shapes start to converge. And so here’s a version of reality that doesn’t change dependent on which sense I’m using to look at it, which is to sort of say it’s the Platonic form, which maybe means it’s a little bit closer to reality as it is.

We’re getting a tiny little slice, but these things don’t change depending on which way you look at them.

We’re starting to get a glimpse at what a world looks like beyond our ability to sense it, a world beyond senses. And I, again, I just think this is so profound. There’s nothing written that says that deep learning had to work for absolutely everything, but that’s the way it’s occurred.

We see and experience almost nothing of the universe.

For us practically, as Earth species, what this means is that we now know everything that can be translated will be translated. Like, whatever is in animal communication, we will be able to understand. In the coming years, which is incredibly exciting. So back to guessing games. Which animal makes this sound?

European starling. You say starling or starlock? Let’s see who it was.

Not a bat, good guess. Not a dolphin. Frog. I’d love to eat that frog.

It’s a beluga whale. These are a couple belugas communicating while they hunt. And what’s crazy to me is like, you hear all that high frequency like digital sounding noise? None of that was added after the fact. This is literally just a microphone in the water.

That is how they communicate, and it appears that they communicate in these modem-like packets where they’re communicating a whole bunch of information, including like their name, including clan identity. There’s a lot that’s encoded, and the crazy thing to me is that one of our partners down at Raincoast Dr. Valeria Vergara. She has these little backpacks on the whales, or what we like to call whaleables.

And even though she has the whaleables recording audio and motion and video, she can’t tell who’s speaking. And they’re all speaking at the same time. You heard it here. She can’t disentangle them. So that means for her, roughly 3% of her data is useful.

And again, we’re back to this place where the most vocal underwater species, with the greatest vocabulary, their communication is only 3% explored. The ocean is 5% explored, which means that movement communication is a greater frontier in our ocean. Just another sort of puzzle in here is we hear up to 20 kHz. Belugas in here are communicating up to 120 kHz. Why?

High frequency does not travel very far in water. So what’s going on? Like, why are they communicating at such a high frequency? And here’s the answer, because this is what belugas are like in the wild. There are so many of them, and you can see they literally rub each other, so perhaps they’re directly vibrating, so like direct communication.

There are so many, it is— you understand how hard this problem is.

OK, so hopefully now I’ve sort of like demonstrated the world is awash in signal and meaning, there’s probably some kind of ‘there there,’ and it’s symbolic communication. We’ll likely be able to do some kind of translation, because AI is giving us representations, formats of understanding, that can transcend any modality, and transcend human understanding. But how deep are the things that they might communicate about? And here are just a couple examples where I think it actually might be very deep. So, how many people here know about the mirror test?

Test? I’m just curious. I see a couple answers just a scattering maybe 5%. So the mirror test is a way of seeing, is an animal self-aware? You do it by taking an animal, you paint a dot on them in a place they can’t see, you put them in front of a mirror, and if they start investigating the dot, well, that means they are understanding that that thing in the mirror that image, that is me.

And so this is an example of dolphins looking in mirrors.

Haha. Haha. Having perhaps the most universal mammalian experience, which is you look in the mirror and you check out your ass.

Haha. Dolphins pass this. There’s one kind of fish that passes it. And elephants actually pass the mirror test, but scientists thought that they didn’t for the longest time. And the reason why is because the mirror was too small.

I know, that is how very human.

And so then, if they have self-awareness, which is It’s pretty clear that they do. Chimps, by the way, are hilarious because they’ll also check out their butts.

if animals are communicating, they may well be communicating about one of the deepest parts of experience, which is consciousness, self-awareness, sense of self. Here’s another example. These are lemurs taking hits off of centipedes. So they find centipedes, they take bites off of them, and they enter this sort of like trance-like state. They’re just getting high.

And then they get super cuddly. It’s almost like a lemur Burning Man.

And it turns out a lot of animals will seek some kind of altered state or transcendent state. Dolphins— there’s long observational data and video of dolphins intentionally inflating pufferfish, getting high off their venom, and then passing them around. It’s sort of the original, the original clubbing device. Orangutans and chimpanzees will spin on vines and get dizzy. There’s recently a documented case of a chimpanzee male chimpanzees going inside of hollowed-out baobab trees in Senegal and doing these pant-hoot dances, banging on the inside.

You can’t hear it very much from the outside, but inside it’s deafening. And thinking that it’s some kind of some kind of trance state they’re entering. So, if animals communicate, they may well be communicating about transcendent states. Again, one of the most profound things that we can experience. And this was actually just like this last week, essentially, this hit the presses.

It turns out that chimps are just as obsessed with crystals as humans are. And they will like, if you give them a crystal, they’ll take it and they’ll hold it for the rest of the day and go right around with it. So, plus one for the crystal people, I guess.

And just one more example of shared experience. This is an orangutan baby watching a magic trick.

See what happens.

Which I think should It shows that another universal experience is a feature of magicians.

Right? But surprise, having your expectations frustrated giving you joy, that goes beyond the humans. And if they’re communicating, they may well be communicating about that. And that is the goal that started out with the first species project. We have many more tools than this now, but to build this shape that represents all of that rich context and understanding and use that to begin to understand another species.

I should say it’s really important because about this time people start saying, “Aha, so what you’re trying to do is break the interspecies communication barrier, trying to speak to animals.” And the answer is no, that’s not what we’re trying to do. What we’re trying to do is break the interspecies understanding barrier. We are trying to understand other species to open the aperture of our own empathy. Because, for human beings if you’re thinking about a group of human beings and they’re living somewhere else far away, it doesn’t really affect you that much. But if you learn their culture, learn to speak their language, they go from out there somehow to in here, to something we’re holding in our hearts.

And if we can do the same thing for the other cultures in other species of Earth. They go from nature as a thing that’s out there to something that’s in our hearts, and that starts to change us. we don’t really change when we speak, we change when we listen. And so what we’re now doing, sort of like more technical front, is we are working on creating sort of analog to natural language processing. There’s things like ChatGPT, GPT, Gemini, all those things, and now applying it to animal language processing, building this kind of frontier AI for the rest of nature.

And one of the interesting things about that is that then one of the questions will often get asked is like, OK, which species are you working on? And one of the lessons from AI is that making the problem bigger actually makes easier. So if I teach an AI about Russian and English and Finnish, it gets much better at doing German. That’s because there are patterns that it’s learned that are the same across the whole. So what we’re doing now is building generalized models that work across the entire tree of life.

So it turns out they work really well for orangutans, they’ll also work well for jumping spiders. And the goal now is to start expanding so we can ask questions at a scale we could never ask before. For instance, here we’ve now seen that a couple species have names. How many species have names? When did that appear in evolution?

How about abstract thought? When did abstract thought appear in evolution? How many species can do abstract thought? We can start asking questions of the entirety of evolution. And that, I think, also is incredibly exciting.

And just, I mentioned jumping spiders, and it was— I really wanted to do that mostly so I could show you this video. This is jumping spider communication. Oh, whoa, cool. Yeah, most of jumping spider communication, a little bit is visual, and a lot of it is vibratory. Every jumping spider essentially has a different vibratory communication system, and really this is beyond the ability for human beings, because you have to— they’re that fast.

It is beyond the ability of humans to like sit there and categorize all these different kinds of vibrations across the— whatever it is— hundreds of thousands of species. This is the thing that really only AI can do, and then use that to map it to and all the other communication systems. One of the pieces of work we’re really excited about is we’ve been working with zebra finches, which are these very cute birds. And we’ve been trying to understand what is their conversation like? Because in order to understand, we’re going to be able to do some form of playback where we like interrogate the system.

And so we’ve done this sort of experiment of This, by the way, is what zebra finches sound like. They’re very cute.

So what we’ve done is we’ve sort of done this Indiana Jones battle swap, where we’ve taken two birds that are talking and we swapped one out for our AI. And have them get into some kind of back-and-forth communication. And the goal here is to be able to understand how they converse, in what ways they change over time, and as they converse, how much of the past conversation predicts the future conversation, which means how much memory do they have, can they refer to things that aren’t happening right now. Actually, just an interesting— well, I’ll come back to that story.

And just to give you a sense of like where the tools have been and where they’re going, we’ve now had an AI, or our little zebra finch AI, in conversation, if you will, going back and forth communication for 48 hours with a zebra finch. So the longest sort of like playback experiments that happened sort of before are around 10 seconds. So it’s as if we’re trying to understand animal communication, but the only tool biologists have is the ability to hit a button and shout once. Imagine how hard it would be to understand animal communication if that was your only tool. So we’ve actually been able to do what’s like 17,000 times longer back and forth, and what we’re discovering is absolutely new biology.

It was thought that zebra finch females that their songs are innate, that once they learn them, or once they like singing them, they don’t change. But what we’ve discovered is that they will change the way they chirp and communicate to sort of match the way our AI is speaking. It’s actually very similar in some sense to ChatGPT, where you can sort of tell which people have been using ChatGPT too much and hear how they speak. And animals do the same thing. They start to learn the way that the AI speaks and start to mimic it.

And that’ll get into a little bit of the ethical implications, which we’ll get to in just a second. One other way of just thinking about, in the end, translation is we sort of talked about how you can translate from text to image. One of the other things you can do is go directly from audio to behavior. So we have a pilot project with this wonderful group called the Raincoast project off the coast of British Columbia. Turns out in the last 2 decades, a lot of progress has been made on everything about orca behavior, but almost no progress—no progress—has been made on orca communication.

It’s just way too complex. And but there are 2 decades worth of very high-quality recordings that have no annotations and no way of knowing what they doing. So the thought in the pilot that we’re running right now, first in British Columbia and then in Iceland and then probably propagated around the world, is can we go from just audio to behavior? So that is listening, and can we then regenerate the behavior that was happening? So generate the behavior.

And if we can do that, record them all daily.

If we can do that, then that means one, you start to understand what sound one whale made before to cause another whale to dive. But it also lets us understand all of this past data, which means we can unlock the data you need to actually do the translation.

I want to talk though a little bit about the ethical implications of what crossing the interspecies communication and understanding barrier really does, and why it is that we focus so much on understanding over communication.

One of our sort of inspir— like, heroes or inspirations is a woman named Katy Payne, and she was the woman who discovered that elephants communicate infrasonically, that is below the level that human beings can hear. She did it literally by sitting there and watching and seeing their foreheads like move a little bit and then watching their behavior change. She’s really quite incredible. And early on she tried an experiment where she played the mother’s calls for the matriarch of the herd back to her, but the mother had died a couple years ago. And all the elephants came running, very excited, and then that excitement turned to frustration, and then that frustration turned into grief.

Think about what that experience must be like, to hear your mother call and then can’t find her.

So starting to communicate back before we understand what we say, we have to be very careful. And there are many other ways that just blindly communicating back can cause harm. Turns out humpbacks, their songs can go viral. So for whatever reason, Australian humpbacks are sort of like the K-pop singers, and for whatever reason their songs will get picked up and sung sometimes by the world population within like a season or two. We don’t know why.

So if you just create a synthetic whale that sings and puts it in the ocean, we might well create a viral meme that disrupts whale culture. And whale culture is exceptional because it looks like, they’ve been passing down culture, dialects, dialects that go into mutually unintelligible languages, and they’ve been doing that for 34 million years. Can you imagine that? Like, what wisdom can be contained in 34 million years of cultural evolution that we might then disrupt. And I would argue, by the way, when people say like, what is the thing you most want to like say to a whale?

It’s like, what is it that we can learn? Whatever it is that is the solution to our problems, I would argue, is not in human imagination. Because if it was in our imagination, we’d be doing it. So we are looking for things and concepts that are outside the sphere of human imagination. Where better to look than in cultures that extend back 34 million years, 85 times longer than our species has existed.

We’ve thought a lot about what it means to be good stewards of a technology that will eventually roll out to everyone and what I think that means is, whenever you invent a new technology, you uncover a new class, a new species of responsibility. We didn’t need the right to be forgotten until the internet could remember us forever. We didn’t need the right to privacy to be put into the US Constitution until Kodak produced the first mass camera, and we could be captured everywhere. We are going to need to have a kind of prime directive as we break the interspecies understanding barrier. And that means norms, it means laws, it means international treaties, so that before there is an economic incentive to exploit, we can already create the protections.

There is something really deeper and more exciting though about what happens on the other side. Of breaking the understanding barrier. I’m curious, how many people here have seen my octopus teacher? And for those who have seen it, how many people stopped eating octopus? I want to— this is really important because normally change is created or attempted to be created by, doing like shame and like, you’re bad.

But notice my octopus teacher never said that we were bad or you should feel shame. He just created connection. Like, instead of coming through the front door of the mind, it came through the side door of the mind with awe and wonder and love. And it shifted behavior. It changed who we were on the inside because it changed how we relate on the outside.

And so we think of Earth species as like a scaled-up, a 10,000 times scaled-up version of the My Octopus Teacher theory of change. That is the goal, to shift our relationship with the rest of nature. And if we can do that, you start to realize, OK, if animals have a deep interiority, are communicating rich, complex thoughts, have long cultures. Well, that starts to change the basis of law, like who gets standing in court, who can be a person, can ecosystems own property, can they be their own sovereign nation. It all changes when we begin to understand.

It changes the basis of religion, like who gets a soul, sort of like, it starts to change. It changes the basis of agriculture. A lot of the presumptions of how our society and civilization has been built start to shift. And it’s that paradigmatic change I think is so important for us as a species to continue to survive on a finite planet with exponential technology. So I want to leave with just one more thought, which is that all of the AI really has been trained on top of modeling and extending human intelligence.

But for all of our ingenuity and our brilliance as a species, there is a problem we have not yet been able to solve, and that is how to successfully play the infinite game on a finite planet. But we actually have examples of systems that have figured that out. That’s of course nature and ecosystems. Before we hit, AGI, which is coming very, very soon, like Indra Madhavan from Anthropic, Mustafa from OpenAI, Microsoft AI, they point out that at some point soon there’s going to be some kind of handoff to these super intelligent things that were created. Hopefully they’re controllable.

Before that happens, you probably shouldn’t have them just learn and recapitulate all the mistakes that human beings have made, right, by modeling and extending how we have solved problems. If we actually want to make this work, we need to learn from the only systems that have learned how to do infinite being, finite planet, with exponential technology, even when you have cutthroat competition, and that is nature. And that is understood through a window of communication. So, in 1995, I think it was December 18th, astronomers pointed the Hubble telescope at an empty patch of sky. And, what did they discover?

Well, they didn’t discover nothing. What they discovered was the most number of galaxies humanity had ever seen in one spot before. And this is the position we are now in. The rest of nature. We are building the new scientific tools that let us point at where humans thought nothing existed, or science thought was empty, and what we’re discovering is not nothing, but everything.

And we started this conversation by saying what was the challenge we had to face The challenge we all face is near optimization at the expense of the whole, which is to say disconnection. This is the way to do the reconnection. And the last time we invented the telescope, we pointed it out at the universe, and what did we discover? We discovered that Earth was not the center. This time point our new telescope at the universe.

And what are we going to discover? We’re going to discover that humanity is not the center. And that’s the shift I think we need in order to survive. Thank you. For people that want to stay involved, scan this QR code, it will send an email to our team, you can sign up for the newsletter.

We are always looking for partners or nonprofits that will come help us out, fundraising to partnerships. And then this The last thing, since it was meaningful, Jane Goodall died just a little bit ago. We were very lucky enough to get to spend some time with her before that happened, and, she said that she shared with us that— oh yes, please. that she had got to spend— that she— Doctor Doolittle, when she was growing up, was— that was her hero. Hero, and that she was so moved that that dream of hers is now finally within the realm of the possible.

So thank you, everyone. Thank you. I think there’s going to be Q&A now, right? Yeah.

OK.

Hello, fellow primates. Hi everyone, my name is Steve Van Roekel. I’m the new CEO of the Earth Species Project. I couldn’t be more excited to give you this presentation at this pivotal time. I can— you can only imagine after seeing talk the type of employees we have that are species project.

I’ve worked with a lot of organizations and this group blows my mind. It’s incredible. What an incredible team. We’re so lucky to be here. All right, so let’s start with questions.

I think you’ve heard at the beginning, if you use the official SXSW app, there’s a link to submit questions. We’ve already got some coming in. But Aza, I know when I, when I was considering this position, You and I spoke a lot, and you promised me that someday I’d be able to speak to my dog. But how close are we to things like that? Well, again, it’s really about understanding versus speaking.

And everyone who has a dog understands this already rich communication that’s going on there. And the question is really, how do we expand that sense that we all have in our daily life to the rest of the natural world? The world. This is new science, so you never know. But really, it is the next couple of years.

It’s like a 2030 kind of thing that we are having, and we’re going to have a long way to very rich, deep breakthroughs. Yeah. A question came in, great question came in, I think you’re going to really enjoy. Is there evidence that animals of a certain species share dialects, considering animals activity.

Yeah, actually there are a lot of examples of species with dialects. Orcas are the most famous example, and they definitely have like accents and they’ll have like some shared words.

Those dialects will drift far enough apart to become languages that they actually can talk. And I don’t know how many people know this, but there are significantly different cultures in.

There was a fad that happened a couple years ago where orcas were wearing salmon as hats, and then it started spreading, and it was there for a couple seasons, I think, and then it disappeared. And this is actually really important because when we think about animals, we often don’t realize that they actually have a difference in their cultures.

So, we often have an image of chimpanzees as warlike. Why is that?

It was a little bit of a fluke. It turned out there’s a whole distribution of behavior of chimps, from like warlike to bonobo-like, like peace-like. And Jane Goodall, when she studied them, happened to study a more war-like culture. And so that’s ended up sticking, and that changes our perception of who we think chimps are, and that has changed the perception and the stories we tell ourselves about who we think humanity is and what people say human nature is. Human nature is this like more like thing, because it comes closer to chimps, but that is just the fruit of believing that only one culture represents the other.

I know you talked about that experiment in your talk today. Can you go a little bit deeper? There’s a question submitted, is how do we ensure the ability to understand different perspectives and empathize with not manipulative other systems? Can you go a little bit deeper? Yeah, this is a deep and hard problem, but just to name it, the fundamental paradox of technology is that it better enables you to understand, it better enables you to protect, but also to exploit.

And so there’s two ways to think about where the harms are going to come from. One is from bad actors, and the other is from bad incentives. So bad actors, those are things like poachers can use the technology to attract animals to poach them. Now obviously you can already use like the simple version when you record a breeding call, like bats actually do, so it doesn’t increase so much the surface area. But what this means is that we think very hard about releasing our models.

So the problem with technology is that it always complexifies things. It continues because if you went back 5 years, I’d probably said open science is the way to go. We want open source everything because we want to accelerate science as much as possible. But now it’s like, ooh, that’s actually— is that quite right? We have to be more nuanced about how and why we release models that can communicate that.

So that’s on the bad actor side, but then there is bad incentives. So, one of the things that keeps me up at night is what happens if Factory Farms discover that actually cows going through grief produce more money. That would be political. So I think before there’s an economic incentive, we have to do our work to start putting in the laws and the regulations— well, essentially nobody still cares— that says these are the ways that you do a prime directive. Like, how do you reach across and do this kind of interspecies diplomacy?

I just want to say there’s one more really interesting thing this is like, here’s another question here about human, non-human species making meaningful influence in human decision-making. And that is, we ran a global study with our partners at Collective Intelligence Project. I can’t remember, it was like 30, 40 different countries, over 1,000 people. And what surprised us was that already, I don’t know the exact number now, but between 60 and 70% of people said that animals should participate in our democratic governance with this understanding. And to me, this is the direction to point at, because once we better understand, we can start to— the idea of consent, the idea of incorporating into our power structures nature itself is the long-term solution.

Yeah, I’m sure you have, knowing your other work, strong opinions on the response possibility of coming up with those prime directors, and Frost, they have. Yeah, absolutely. It’s incredible. Has this research and work of Earth Species changed anything about your everyday life, general philosophy? What’s evolved for you?

Yeah, well, a big one— there are two. One is that I no longer eat factory-farmed animals. And a big part of that was both your species and also my partner, who doesn’t eat— she’s vegan. But it really has shifted the way that I feel about care. Because when I go out into nature now, it’s no longer, “Oh, here’s a nice bird.” It’s starting to see an ecosystem, an interconnected whole, and that feeling when you’re in the middle of the Redwood trees and you put your head back and you look up.

I get that much more now, and I just hear animals communicating. And so I walk around the world somehow, even though I didn’t fully accept this, with like a greater sense of awe, which is— I would encourage everyone in this room and everyone to take a breath, look up, look at those birds, listen, watch what’s going on around you. It’s really amazing. The U.S. population is projected to be about 89% city-based by 2050. The world is over 50% based in cities, and we’re losing that opportunity as humans to connect to nature, and it can be a really powerful thing.

And if you take it to aquatic places, 89% of people are living in cities, that means 89% of the world’s population is living where they can’t see stars. Exactly. We can’t see the sky, we can’t see the thing that causes the long-term floor. Like, that’s a profound change. Definitely.

For better or worse. Great question came in, and I think a lot of us, especially using chat apps and things through AI, ask this question: How do we know that AI is not making false claims? Maybe don’t understand the underlying language? Yeah, that’s a great question. And that’s just one of the ethics problems we need to face is, first, how do we know that we’re not saying something that’s false or incorrect?

And how do we know that we’re not interpreting it incorrectly and then making, say, government decisions based on that? And it’s not an easy problem, but that’s— this is why we actually have to get into these closed loops where we’re communicating, making guesses about what’s going to happen, seeing whether that was right, and like checking our hypotheses. And I just want to name like one of the big plot twists as we started to make Earth species is that it turns out we will be able to communicate fluently before we understand what we’re saying. And it’s sort of like ChatGPT, right? ChatGPT doesn’t actually really understand.

It’s as if you had the superpower to walk with anybody, sort of like listen to them for a little while, pop your head to the side, and then start mimicking their patterns, and you’re just speaking back a language you don’t understand, and they’re like, “Wow, that’s really insightful,” right? It’s like that. And so because the power to communicate outstrips the speed at which we’ll understand, this is why we have to take it so seriously, the ethics side of things, and do the harder work of first understanding, before we did the human, especially at the bottom. Any perspective on using AI for the climate crisis and maybe the overlay that animals might have in the climate crisis? Yeah, and it’s complex.

Obviously, the better we understand how the things that human beings are doing affect nature, the better we’re able to intervene in some way. But really, it’s not like we don’t already have much of the technology to be solving climate change. It’s really much more of a coordination problem than it is a technology problem. We sort of make industrial-era tech and we create super-industrial problems that industrial I can’t quite solve, but you can solve if you coordinate. So it’s not really a problem with the technology, it’s a problem with how we understand ourselves as well and how we come together to solve those problems, right?

If you can solve the coordination problem, you can solve a lot of other problems. Working on animal communication, I think, gives us access to a whole new range of potential ways that we can collaborate. My favorite example of this is some elephant research where it looks like elephants— actually, there’s some more research that elephants, when they communicate, they overlap 90% of the time. That is to say, when we do human things, like you speak and I speak and you speak, but elephants do the opposite, they’re overlapping. And so the question is like, why and how Is there a kind of communal grammar that’s happening?

Is there a way— it does appear that there’s like more information the more of them that speak, which means there’s a greater amount of like coherence happening perhaps the more of them that speak. That seems like a thing that humans need to learn how to do, but the more of us speak, the more coherence versus the less. This is, I think, my favorite question that I’ve seen pop up is, is there a way to understand why my cat yells at us all the time? She’s clearly communicating. She’s probably just mad at me.

That’s the great thing, we don’t know, but we’re gonna find out. Yeah, parts of science actually, this is really interesting, cats have said that cats have developed meowing as a human interaction, right? Yeah. Versus they don’t really meow to each other, which is great. Which is really to say they’re like, you dumb human, I clearly communicated If I indicated by turning my back this way, you still don’t understand I’m going to have to react.

How uncouth of you. Exactly, exactly. Last question. Thank you for this question. Wonderful job.

How can we help this project? Yes, Steve. Steve, as our new CEO, how do you want to answer that question? Well, I think you, one, stay informed. go to the link that Aza put up earlier, earthspecies.org.

To sign up for our newsletter. We’ve got some jobs posting up there occasionally. I think there’s one active right now around data science. And then we are a nonprofit, so helping us cultivate donors, relationships with researchers, all that stuff to really kind of send the world the message that AI can be good. And we’re one of those places where that’s happening.