To increase productivity, it is expected that a single user is able to operate multiple cybernetic avatars (CAs). However, the limited attention span of the user makes it difficult to send direct instructions to all CAs. Therefore, this chapter describes the essential technologies for CAs that solve these problems and behave autonomously according to the user's intentions. First, the realization of spatio-temporal recognition capabilities that enable CAs to move autonomously in an environments that change from moment to moment is described. Following that, methods to implement continuous learning and memory mechanisms to facilitate acquired information reuse in the future are described. In general, the observed data are time series, and future predictions are important to provide appropriate support to users. The time series analysis method is then explained, which is the most important technology. Advanced natural language processing technology is necessary to capture intentions through dialogue with the user and to process large amounts of textual data as prior knowledge and common sense. Examples of the application of these fundamental technologies in the medical field are also presented.