Have you ever listened to the quarrel of a group of sparrows and wondered what they were arguing about? Have you ever looked at your dog and wished you could have a back-and-forth conversation with them? With artificial intelligence (AI) growing in popularity, a common question starts to arise: Will AI become the window to animal communication?
Up to this point, communicating with animals has consisted of scientists observing the body language and vocalizations of different species and, in turn, categorizing these actions into logical assumptions about what the animal intended to convey. These speculations take years of observation and vary among species. Furthermore, this communication is one-sided: We observe and can guess what the animal is trying to “say,” but we cannot successfully reply. This is true for most animals, especially ones not in captivity or kept as pets. For animals that we are more familiar with, or that scientists have worked closely with—meaning singular individuals of a species—there have been instances in which back-and-forth communication has become a reality.
Take Bunny, the infamous sheepadoodle, for example. Bunny uses over 90 buttons to communicate with her owner in English. She uses words any dog would know, like “walk” or “food,” but also more complicated ones like “stranger” or “please,” even using a “hmm?” button to ask questions. Her fascinating usage of these buttons took the internet by storm. But this would not be the first time an animal and their companion could communicate with each other. Koko, the gorilla, used American Sign Language to “talk” with her caretaker.
Although it is fascinating to contemplate the idea that these animals are conversing freely with us, it is not authentic communication. In both scenarios, the animals learned to respond to external stimuli and therefore connect certain words or actions with that stimulus—which could be food, the cat, the park, etc. After all, they seemingly learned English and only used the words and actions they were taught—true communication between two species would not look like this. Instead, it would be more beneficial for us to completely comprehend the “language” that these individuals use and respond in a way that could directly translate our words into their language. This, of course, has not been done.
However, there has been an effort to use AI to achieve this goal. AI comes with a plethora of its own issues and discourse, but it has significantly impacted certain sciences and made some aspects of analysis easier for researchers. AI has its shortcomings when it comes to creativity and conceptual topics, but finding patterns and organizing data is not one of them. Because of this, using artificial intelligence to sort through collected data has made it more efficient for researchers to access complex and time-consuming information. So, although this technology is not quite ready to be chirping back to birds, it is helpful. For example, it can help in finding the differences between varying bird vocalizations, which may provide researchers with insight on what those signals and patterns mean to the birds.
Sensory ecologist and Professor of Biology and Director of Neuroscience and Behavior Megan Gall, shares that there are limitations, though. She states that: “even though AI can show us patterns and information in different signals, many things are still challenging.” To begin with, you have to utilize psychophysics to truly understand how these organisms are responding to these signals. Psychophysics is the study of how an organism’s sensory system responds to and translates external stimuli; there is no other way to analyze these signals unless you get inside the brain of the animal—quite literally. So even though you might completely analyze the data from a bird call and play it back to the bird, there is no guarantee that the meaning you attribute to the call matches what the bird itself understands.
Another limitation of AI that Dr. Gall mentions is that it cannot generate the signals for animals, and even if it could play back some form of communication, it is highly likely that the animal will not interpret the signal in the way the researchers intend. This is due to umwelt, or the way the organism perceives the world around it based on the kinds of senses and physical attributes that it has. A well-known example of this is how a dog is colorblind but can smell up to 100,000 times better than humans. Clearly, they see—or rather smell—the world vastly differently from us. As a society, we often forget this difference in perception. We assume that most, if not all, animals experience the same world as us, but this is not true. Unfortunately, this means that nearly all technology and architecture we make is for human convenience and would not be an effective way of sharing signals with animals, even if we did somehow “crack the code.”
It seems that AI will, unfortunately, not be the liaison for human-animal conversation anytime soon. Only through more data collection and analysis will we really be able to understand the umwelt of the species around us. Understanding the way the world works for non-human organisms will ultimately help better urban planning and coexistence, conservation efforts and mitigating wildlife conflicts. This is a vital effort worth working towards. And, I mean, honestly: Do you really want to know your cat thinks your new haircut is lame? I did not think so.