Contributed talk

Neurorobot“ Vitaenoid” - an Interaction System between a Living- Neuronal-Network and Outer Environment

Ryu Umebayashi, Suguru N. Kudoh

In our Laboratory for Neuronal Intelligence Engineering (NIE LAB), the neurorobot VITAENOID, in which a cultured living neuronal circuit interacts with the outside environment via an artificial interface such as a robot body or a smart-speaker, has been developed as a model of the small brain. The ultimate goal of this research is an attempt to constitute the prototype of primitive artificial consciousness. We assume consciousness dwells in the physical layer of the complex system, so the feature of components of the system should be resemble to neuronal cells in the brain. There- fore, a network of neuronal cells as a physical component of the brain is considered to be an essential component. In our study, correspondence between the neuronal activity and suitable behavior is implemented by program. the assumption- based-design of connections between neuronal circuit and robot body (sensors and actuators) was employed. We call the system “ Vitaenoid.” The name Vitaenoid means “ a- living-like thing,” Vitae is“ Life” in Latin, and“ N” symbolizes“ Neuron.” A suffix“ oid ”means“ like ”. Various types of the Vitaenoid can be designed, such as a combination of a living neuronal circuit and a certain smart-speaker or a miniature moving robot. The former is speaking Vitanenoid (Vitaenoid-Babilon) and the latter is moving Vitaenoid (Vitaenoid-Animalis). As all other hybrid-type neurorobot, the major components of “ Vitaenoid ” is a living neuronal circuit (LNC), a robot body or smart speaker, and an extracellular potential multisites recording system as an interface between the LNC and electrical components. The most critical concept of the“ Vitaenoid ”is that its LNC is the central processing component without any other supervisor. Instead of explicit training of the LNC to perform ideal response, we designed the connection algorithm between the LNC and the robot body to perform the ideal behavior. We introduced two types of Vitaenoid, Vitaenoid-Babilon and Vitaenoid-Animalis . Vitaenoid-Babilon is a ”living ”dialogue system. The agent speaks only babling talk of 3 words selected from pool of the words in the cloud, such as the twitter, depending on the internal state of a living neuronal circuit. Our aim is not to realize logical talking, but to realize the agent tweeting only three words like babbling. A 3-words- response is generated from Tweets searched in Twitter containing cooccurrent words of the input sentences. Although it was still a preliminary result, we implemented an agent that selects responses depending on internal conditions of a neuronal network, as emotional polarity of the agent. Vitaenoid-Animalis is a combination of a living neuronal circuit and a miniature moving robot. In Vitaenoid-Animalis, we confirmed that the SOM-based “ Vitaenoid-Animalis ” performed a collision avoidance behavior, and find that the patterns of electrical activity of neurons in the circuit evoked by particular input changed to be more reproducible and more unique. The result suggest that Vitaenoid-Animalis mimicked the non-teacher learning of the animal behavior.