Will AI enable us to talk to animals? 🐋

Good morning, good afternoon, good evening, and good night wherever you are in the World.

Last week’s newsletter is a bit late… For the observant reader, It is not even last week anymore… 🙂 

Does this mean you will get two newsletters this week? Maybe…

But for this edition, you have already read the heading: I am talking about how AI might allow us to talk with animals.

So, without further ado, let’s jump into it!

Research projects are leveraging AI to understand animals communicating

When I started this newsletter back in December 2022, I would not in my wildest dreams think I would cover a topic such as this.

But here we are…

And while I can see an argument for “we don’t need this, this is not good,” there are plenty of opportunities for good by understanding animals.

Maybe I am applying too much “greater good” thinking here, but at the very least, I find this research and the use of AI interesting.

With AI, scientists worldwide are making strides in research and discovery.

In a recent study published by Oxford Academic, researchers were able to translate Cuneiform with AI - one of the earliest writing systems in recorded human history (ca. 3,400 BCE–75 CE), as well as Egyptian hieroglyphs.

However, all projects related to translating ancient or lost languages have one thing in common: they are all human languages, and therefore, large datasets of “understanding of human communication” exist.

We do not have the same datasets when we move over to the animal kingdom.

Thus, simply finding patterns and translating them is a whole different ballgame.

With the capabilities of the LLMs, researchers are hopeful that we will be able to interpret non-human communication once we have larger datasets available.

However, there are no guarantees that animals “talk” or communicate in a way that makes sense to us, for example, non-verbal communication.

A good example of non-verbal communication in the animal kingdom is the honey bee dance communication system.

It is quite interesting to see what scientists manage to create, like the scientists at Cornell University who created a bee robot that successfully communicated with other bees and recruited them to forage.

Projects using AI to learn about non-human communication

Let’s do a quick run-through of two interesting projects going on right now where they are utilizing AI in their research.

ProjectCETI

Project CETI (Cetacean Translation Initiative) is a nonprofit dedicated to understanding the intricate communication patterns of sperm whales.

They use AI to decode these communications, focusing on the distinct rhythmic clicks, or 'codas,' sperm whales use.

I think it is awesome that they have found out that different sperm whale communities have unique coda repertoires. They serve as identity markers and indicate a whale's membership in a specific cultural group.

CETI's research is primarily conducted in the Eastern Caribbean, with its base in the Dominican Republic.

CETI uses both aerial and underwater drones to capture the daily routines and behaviors of sperm whales.

Have a look at how they record it all:

[If it is hard to see the details, then here is the link to the original image]

The leading researcher for CETI (Shane Gero) has spent nearly two decades studying Caribbean sperm whales, providing invaluable data for the initiative.

The Machine learning algorithms are analogous to those used for human language translation and are employed to identify patterns and structures within the myriad of clicks.

You can read more about ProjectCETI here, here, and here.

Orcasound and AI for Orcas

Orcasound has been at the forefront of efforts to understand and protect the enigmatic killer whales, specifically the Southern Resident Killer Whales (SRKWs).

As part of its multifaceted approach to conservation, Orcasound has spearheaded the "AI for Orcas" project.

This initiative aims to harness the power of artificial intelligence in studying orca vocalizations.

The "AI for Orcas" project represents a synthesis of advanced machine learning models and open-source collaboration. Its primary objective is to classify the distinct vocal signals – calls, clicks, and whistles – produced by killer whales in real time.

A cornerstone of this initiative is creating and utilizing open-access training and testing datasets. These datasets serve as the foundation upon which machine learning models are trained, enabling them to discern and interpret the myriad of sounds produced by the orcas.

As I mentioned in the beginning, building a solid dataset is essential for an AI to do its thing.

But the cool thing with Orcasound is that they are developing datasets and sharing bioacoustic software. They hope will not only aid in understanding orcas but potentially be able to work with other marine mammals and soniferous species in non-marine environments.

I have no clue what “soniferous species in non-marine environments” means, but it sounds fancy, so it is probably important…

Other ongoing projects

There are a lot of ongoing projects to understand the animal kingdom better.

Here are a few I found. So if you find this stuff interesting (I did not want to make this newsletter toooooo long), here are the links:

• HALLO (Humans and Algorithms Listening to Orcas) is a project that partners with Orcasound and uses its infrastructure for its research. The project is based out of Simon Fraser University.

• Earth Species Project (ESP) is cataloging the calls of Hawaiian crows, attempting to build new technology that could help humans talk to animals (this one is very interesting, with lots of great articles). There is also an hour-long video here (Youtube) if you care to watch it.

• DeepSqueak is a software that uses deep learning algorithms to identify rodents.

• Bird Sounds is project is an open-source program that visualises thousands of bird sounds using Google’s AIs together with Cornell Lab of Ornithology.

Other interesting articles, should you wish to read even more…

• How Scientists are Using AI to Talk to Animals (Scientific American)

• LLMs Docode Enigmatic World of Animal Language (AIM)

• Can AI really help us talk to animals (The Guardian)? The article talks more extensively about what I talk about in this newsletter.

Ethical considerations?

I decided to add a short paragraph about ethical considerations. If you made it this far, maybe you are thinking about these questions too…

If we are able to interpret animal communication, wouldn’t someone find ways to use that for nefarious purposes? Imagine the whaling fleet of Norway back in the 1940s to 1960s, when my countrymen almost eradicated the great whales in Antarctica, had access to modern technology, and could understand the communications of the whales. They would empty the sea in no time…

Or what about an ethical question like “Do we have the right to eavesdrop on animals without their consent?”

Or how do we actually interact with the animals? What is the purpose? What is the right way?

Personally I think it would be great if we could use technology to understand better and help animal life. But then again, the reason why animals need “helping” is probably our doings in the first place…

That’s all for today!

What do you think? Are you positive or negative about the prospects of communicating with the animal kingdom?

If you have any questions about AI or any feedback, just hit reply or tweet me @thomassorheim 

PS! 
What did one whale say to the other during their date?

.

.

Whale you be mine?

(Today’s dad joke courtesy of ChatGPT)