Learning and Adaptation in Biological Cellular Processes

Adaptation and, consequently, learning in biology are emergent properties of the non-equilibrium nature of the systems. To be adaptive, these systems must form autocatalytic (cyclical) dissipative structures. Rather than dissipating only heat, as tornadoes and hurricanes do, these structures dissipate material through chemical reactions in which high-potential chemicals are transformed into lower-potential materials. Some of the dissipated material forms new dissipative structures (cells). Learning occurs because as new cells are formed, mutations during replication allow the cells to (1) adapt and move from a less probable metabolic state to a more probable metabolic state (the second law on the time scale of natural selection) and (2) adapt to take the most likely path (least or stationary action). Adaptation results in the learning and implementation of control. By autocatalytically and cyclically dissipating energy in the form of material, the chemical reaction networks learn to maximize cellular replication and, hence, the dissipation of energy from the environment. In other words, maximal thermodynamic efficiency enables the production of more cells, which consequently can dissipate more energy from the environment in less time.