by Jacek Korneluk

Some time ago I started thinking if AI can help to decode and translate certain animal languages.

Yes, I mean an A n i m a l, for example, a Dog or Delfin or Parrot.

For many of you, it may sound bonkers, but for me, and I believe for many others it may be the interesting and realistic topic. At least to some extent. All with the help of rapidly growing AI capabilities and cooperation between Humans-and-Animals-in-the-loop.

We need to set up a couple of clear goals, and decent objectives, and keep persistent with what we want to achieve. It is very easy to lose interest or even give up – especially with so ambitious projects like non-human understanding.

But bear in mind that nowadays multiple AI-powered platforms are on our side. They are not only available to the Big Boys and Big Tech players, but to individuals like us. So, we can learn and reach the cutting edge of mind-blowing technologies, which was previously reserved for a few chosen ones. That is a fact that makes a significant difference. It may work to our advantage. Similarly, it is much easier to organise, prepare and handle a data set, or even generate a synthetic one.

So, let’s allow the AI force to be with Us.

For a lucky beginning let’s start to formulate and decipher the Problem**:**

“Can we better communicate with animals?”

Decoding animal language completely is extremely difficult, if not almost impossible ATM, with the current technology and knowledge we have.

Here are my recent thoughts regarding the challenges:

“Animals communicate in very different ways than humans – they use body language, scents, sounds, facial expressions, and more. We don’t fully understand all the complexity and meanings behind their communications”.  But this is why we need AI to help us deal with the problem.

“Each species has its unique language and there are thousands of species. Decoding one animal’s language does not mean we can decode them all”.  Let’s focus on one or two selected species. We certainly not aiming to decode all of them. One is enough for the lucky beginning.

“Animal “vocabularies” and sentence structures are likely far more basic than human language, so decoding meaning could be limited”.  I think it is an advantage rather than an obstacle. We can always generate some synthetic modifications to make simple language sounds more human-like if needed.

“Some animal communication may be driven more by instinct and emotion rather than conscious thought. This makes decoding and translation very difficult”.  Once again, AI is designed to solve a difficult problem, so we are on a good path, I guess.

“Studies attempting to teach animals human language have shown limited progress, suggesting animal vocal cords and brains are simply not wired for complex symbolic language.”   We need more studies and I see that quantum computing is coming! So, Quantum-powered AI will help to wire it better IMO.

Also, I am optimistic and enthusiastic about most decent innovative projects. Believing we will do it, is my second nature. Without this set of skills and proper a “can do” attitude, it may be difficult to manage a challenging project.  So, go, do, pivot, and repeat until you see the light in the pipeline.

According to multiple resources, we have made some progress in decoding animal communication signals mostly related to warnings, food, and social hierarchy. Some researchers are using AI/ML to analyse animal sounds and behaviours for patterns. As we know, AI is extremely good with recognising patterns or lack of patterns as well, so that is handy. But complete decoding of complex animal language still seems out of reach for now, based on current technology. More advanced (BCI) brain-computer interfaces 1(Wikipedia2023) or general AI (AGI) advances may help us in the near future.

Once again, quantum is on the way!

Animal communication has long fascinated humans, so I am not the only one with an extended interest in it. From bird songs to whale calls, dogs’ responses included scientists trying to find a deeper sense in “animal talks”.

But the questions return.

“Is decoding animal language with AI a Futuristic Vision or Near Reality?”.

“Could artificial intelligence help decode the subtle differences of animal   language and bring us closer to animals’ better understanding?”

This intriguing possibility makes me torn between a futuristic vision and near reality. I am a realistic visioner and innovator, and I am usually right!

Domain’s tailored teaching AI systems to parse the noises animals make could unlock revelations about how non-human species share information, express emotions, coordinate, and survive. AI has already decoded some animal communication signals related to danger, food, social status, navigation, and reproduction. With enough training data of animal sounds, movements, facial expressions, and environments, deep learning algorithms may begin to pick out consistent patterns across species and fish out the contexts. All step by step, diligently and persistently.

“Rome was not built in a day”, as the Romans used to say. So, did not the non-human to human AI-powered translator?

We certainly have time, and we can wait. Nevertheless, we should use our time actively building, and experimenting to achieve the step-by-step progress we aim for.

This could extend our comprehension beyond obvious warning cries and mating calls to the more subtle social exchanges and survival strategies encoded in animal talks. We may even be able to gather insights into animal psychology – are they expressing happiness, anger, or jealousy? The potential to bridge the human-animal divide through AI is great.

If you are still reading it, then I will tell you more. My futuristic vision goes further – two-way communication.

If AI systems could translate human speech into an animal’s language, and vice versa, we could engage in simple dialogue to exchange needs, thoughts, and emotions. Children’s movies have imagined this world, where a device allows us to ‘talk to the animals.’ This remains effective in the field of science fiction, but AI at least gives us a hint that basic voice-to-voice translation could become possible.

With all my optimism, the challenges standing between the vision and reality should not be underestimated.

Animal brains, mouths, vocal cords, and social structures have all evolved very differently than humans. Most species rely far more on body language, scents, facial expressions, and environmental or behavioural signals layered on top of vocalisations.

“Usually, the context is key. So, even if AI could crack the code of these complex, multi-modal languages, what meaningful information could we extract?”

Of course, that is the Big Question, and probably we need to experiment more, before trying to answer it.

“If I knew it was not possible, I would never do it. Likely I did not know and finally, I did it!”         Jacek Korneluk 2023

Additionally, each species has a unique language specific to its culture. Progress in decoding a dog communication does not automatically translate to marsupials or songbirds. But we do not expect a universal animal translator – do we?  We are just looking for the first which may come anytime soon.

For example, useful applications could include alerting homeowners to signs of stress in domestic pets, monitoring livestock health, and preventing poaching of endangered species. That’s a few of the pioneering potential use cases.

Truly conversational interfaces seem unlikely in our lifetime without immense advances in both AI and our comprehension of animal minds. But artificial intelligence may help to unlock some of those secrets, bringing us a little bit closer to the alien brilliance of non-human intelligence. And I am not scared to talk about it.

My final thoughts.

Since animal communication methods are still not fully understood by scientists, it’s difficult to make a definitive comparison to any specific computer system or programming language. However, here is my educated guess on which languages may share some similarities in logic to the general communication patterns seen across some animal species.

Visual Programming Languages

Many animals rely heavily on visual signals like body language, facial expressions, and behavioural displays. Visual programming languages like Scratch 2(MIT), and Blackly 3(Google) that use graphic blocks to create code to follow a visual logic that could parallel how animals interpret visual cues.

Logic Programming Languages

Some animals like bees and whales rely on sophisticated navigation techniques that involve complex spatial reasoning and logic. Logic programming languages like Prolog 4(Wikipedia,2023) which use facts and rules to draw conclusions are closer to this deductive reasoning than procedural languages IMO.

Reactive Programming Languages

Animal communication is very much driven by reactions to environmental stimuli and each other. Reactive languages like Elm 5(Wikipedia, 2022) that focus on data streams and event responses may share that immediacy of reacting to signals that are common across species.

Neural Network Languages

Neural networks take inspiration from animal brains. Languages used to build them like Pytorch 6(Wikipedia, 2023) have some similarities to how animals may instinctively learn to interpret communications through sensing input patterns over time.

According to my current understanding, no single programming language maps perfectly to the countless methods animals use to transmit information. Elements across visual, logical, reactive, and neural network languages likely apprehend fragments of the complexities inherent in animal communication. With the help of AI and quantum computing, we can work together to work out it. But as we decode more of these species-specific languages, we may uncover stronger parallels with our modes of human and computer communication.

That’s all for now. Thank you for reading.

References:

1. Brain–computer interface (2023) Wikipedia. Available at: https://en.wikipedia.org/wiki/Brain–computer_interface

2. MIT Imagine, program, share, Scratch. Available at: https://scratch.mit.edu/

3. Blockly  |  google for developers (no date) Google. Available at: https://developers.google.com/blockly/

4. Prolog (2023) Wikipedia. Available at: https://en.wikipedia.org/wiki/Prolog

5. Elm (programming language) (2023) Wikipedia. Available at: https://en.wikipedia.org/wiki/Elm_(programming_language)

6. Pytorch (2023) Wikipedia. Available at: https://en.wikipedia.org/wiki/PyTorch

Originally published at https://spektrumlab.io on October 2, 2023.