FROM EXTERNAL TO INTERNAL STATES OF ANIMALS

From Quantamagazine

FROM the IMPRINT to GRAMMATICAL RULES of ANIMAL BEHAVIOR.

Konrad Lorenz, Nikolaas Tinbergen and Karl von Frisch (ethologists), won the Nobel Prize in Medicine in 1973, for having conducted studies of animal behavior, with meticulous descriptions of fish, insects, wild geese, behaviors, standing out within them: the animal filial imprint. From then until today, more rigorous, more quantitative studies were needed, with more argumentation and relationship with the neuro-emotional internal states of animals, including human behavior. Currently, there is a tendency to relate animal body behavior with: a) simultaneous associations to the corresponding neuronal codifications b) associations between animal behavior to measurements and quantifications associated with recent technological advances, using A. I.  techniques and the capture of deep neural networks, analyzing animal behavior beyond predictability. In this route, scientists from various parts of the world have adapted and modified machines with resulting quantitative versions, difficult to being questioned. The game: DeepPoseKit, equipped with algorithms that automatically follow the movements and orientations of desert locusts are used to obtain data of collective behaviors. While Lorenz's seminal studies focused on external behavior, modern research looks within animal behavior, trying to understand the meaning of body movements. In 2013, Sandeep Robert Datta (Harvard Medical School), acquired a Xbox Kinect device (which feels the movements of the players), extrapolating  and monitoring it to study in detail the body movements of environmental scouting mice, plotting their movements with clouds of spatial points, analyzing the rhythmic movements, with the conviction that giving meaning to their corresponding neural codes, a better understanding of the behavior of the mice would be accessed. It is necessary to build algorithms that automatically follow animal movements, record small angular changes of the wings of an insect or the arch of a mouse's back creating patterns, that  machines analyze and classify the data automatically and serve as guides for the internal states of animals. In an article recently published in Nature Neuroscience, by: Adam Calhoun, Mala Murthy and Jonathan Pillow, they built  a learning model, to observe animal behaviors by identifying internal states: courtship of fruit flies, manipulating the brain activity of flies and controlling the network of supervisory neurons of these movements and internal states, identifying movements and analyzing behaviors with contributions from neuroscience, genetics, evolution and medicine. Today, algorithms are used to identify the edges and contours of images of certain animal behavior problems: traces of contours of flies on surfaces, differentiation within multiple organisms, identification of certain body parts, training of neural networks to track the joints and body parts of any animal (deep learning),colors, transpositions on the animal body are used to identify its nose, tail, ears, legs, learning machines to map human movement. Jonathan Whitlock (Norwegian University of Science and Technology), uses small pieces, attached to 3 points of the animal along its back, to reconstruct the movements of the spine. The same investigator activates certain neurons of a fly making it walk backwards in a rotating spherical wheel, in addition to studying movements and positions of mice encoded in neurons of the cerebral cortex, committed to coordinated movements that explain how the mouse holds its head. Finally, Ilan Golani (Tel Aviv, University), after analyzing animal behavior for more than 60 years, argues that there are fundamental units of behavior within a common set of rules (grammar), based on animal anatomy. Benjamin de Bivort (Harvard University), who supports Golani, considers it important to properly interpret these blocks of behavior, which he believes are finally a type of hidden language. In 2008, 4 building blocks of worm movements were discovered. With algorithms elaborated in the Bob Datta's laboratory, explained in the Motion Sequencing, short units or syllables have been identified in the behavioral dynamics of mice, confirming that animal behaviors consist of small sets of behaviors (syllables), framed within certain rules grammatical. The Datta group managed to identify the neural network associated with the syllables "run forward", "down and dart", "get out!". using Motion Sequencing algorithms (MoSeq), predicting that a mouse could use 40 or 50 of them, with some corresponding to certain human behaviors. On the other hand, Datta discovered in the cerebral striatum, sets of neurons, responsible for representing different syllables identified by the MoSeq, being the grammar directly regulated by the brain. Datta adds that the neural representation of a syllable is not always the same, it changes to reflect the sequence in which the syllable is involved, so it cannot always be said whether a certain syllable is part of a fixed or variable sequence. According to Gordon Berman (Emory University), these studies will one day serve to predict social interactions between animals.