This paper examines the potential uses of the computer for amplifying and guiding human learning. The nature of teaching and learning is first considered and an operational definition of learning is arrived at—i.e., the regulation of the stimulus environment," specifically, "a beneficial control engendering optimal learning conditions." "Intelligent automata" are concluded to be the ideal instructors, according to the needs included in these definitions. Some of the functional characteristics of this ideal instructor are then specified and the idea of evolving a model of the ideal instructor is proposed. Some of the physical, hardware requirements of these automata are also briefly considered. A rough outline and characterization of research needs is presented. These needs are classified arbitrarily into six major categories: 1) subject-matter structure; 2) instructor-student information coupling; 3) student-instructor information coupling; 4) measuring the behavioral modification in process; 5) motivation; and 6) differences in the learners. Some of the possible social consequences of the proposals presented here are briefly considered. These include the ethical problem of too much human control over human behavior via the automata, the possible elimination of both poor instructors and poor students, the limitless duplicability of ideal instructors wherever they are needed, and the fact educational research results could "be directly injected into educational practice" and educational administration could thus be immensely improved.