How AI Revolutionizes Established Tech — Part 2
With rapid advances in data discovery and management, older technologies have no choice but to keep adapting in order to find a way to profit from AI. But it is not just humans that reap the benefits of digital transformations that machine learning algorithms have to offer. The natural world with its diversity of species is finding itself in the midst of new technological transformations as we keep changing our relationship with it. Continuing with the Series: How AI Revolutionizes Established Tech, we take a brief look at how AI can help crack the code of whistles, chirps and growls of animal communication.
Part 2: AI Decodes Animal Language
It has been nearly a century since Frisch managed to translate the meaning of the “waggle dance” in honey bees. The sophisticated dance would relay information to other bees about food sources that may be at a distance, so that they too can find their way to this precious resource. The bees’ movements describe with amazing precision the direction of flight with respect to the position of the sun even if there are hindrances that they must detour around.
Frisch’s pioneering work around sensory perceptions of these tiny creatures would later earn him the Nobel Prize for Physiology or Medicine together with Konrad Lorenz and Nikolaas Tinbergen who also made groundbreaking contributions in comparative behavioral physiology. Lorenz studied instinctive behavior in graylag geese and jackdaws, whereas Tinbergen explained the role of internal and external stimuli in controlling the expression of animal behavior.
Animal communication has continued to fascinate humans to the point where we nowadays apply AI to further penetrate its mysteries. Data analysis and technology tools are now used by cognitive psychologists to better understand the animal kingdom. Together with sensory tools and digital interactive systems designed specifically for other creatures, a wealth of data is being collected, organized and analyzed. It is no surprise that natural language processing is applied to not only humans, but also to animal tongues. Deep learning algorithms developed for unsupervised machine translation of human languages are now applied to the interpretation of vocalizations of whales, primates and birds.
Complex sonic arrangements are fed to AI-powered animal translators similar to Rosetta Stone tools. These projects are based on the assumption that there is a universal grammar that characterizes both human and animal sign systems. By statistically extracting the patterns of language, a machine learning algorithm can distinguish semantic elements and thereby reveal the language structure. Once this structure is uncovered, it can be deconstructed into a series of phonemes which can then be used to build an AI-translator. Another approach involves using long short-term memory (LSTM) and convolutional neural networks (CNN) to classify spectrograms from acoustic data based on location and thereby learn meaningful representations of animal vocalizations.
One challenge with this sort of AI software is verifying and validating the machine learning algorithms. With human languages, these are usually benchmarked against a human expert, but who is to say that a translation of a bird’s chirp or a monkey’s growl is accurate or not? It seems that cognitive psychologists have found an answer to this, too. Expertly trained animals in mixed animal/human environments have grown up versed in both animal and human languages. These skilled bilingual interpreters can translate the slang of their species into the parlance of people, and vice versa. For instance, hoots, barks, grunts and cries can be confidently used to check the accuracy of animal communication, if the animal can be coaxed to gather around a computer screen.
It seems that all this effort is having a big impact on environmentalism, promoting conservation movements. Initiatives likes these are more and more backed by private foundations and corporate investments who can guide the testing of shared AI platforms. Whether interspecies communication is accurate or not at this stage of AI applications is of lesser importance. What stands out is that these kind of AI endeavors can be used to further synthesize environmental conservation plans and facilitate future digital communications. Technological advances aside, the motivation behind these emerging fields can be boiled down to having more compassion and empathy for other creatures, if we can have a chat with them.
