Richard Feynman

Friston traces variational free energy to Feynman, who used it to solve difficult integration problems in path integral formulations of quantum physics¹. Friston points out that the same bound is used widely in approximate Bayesian inference, including variational Bayes². On this reading, internal states parametrize a variational density that encodes Bayesian beliefs about external states. The dynamics of internal and active states can then be described as a gradient descent on free energy. Friston gives the expression two terms: a negative log evidence term and a relative entropy between the variational and posterior densities¹.

Bongard and Levin open with Feynman's lecture "There's Plenty of Room at the Bottom". They summarise it as an argument that large technological progress could come from manipulating matter and energy at ever smaller scales³. They add that natural selection could presumably exploit the same potential⁴. Their own title turns the lecture around. Biology cannot go down, because something is already there at every level. They argue that it instead overloads mechanisms with several functions, which they call polycomputing³.

In the TAME framework, Levin cites Feynman (1942) with several other authors. The claim is that a minimal degree of internal self-determination and goal-directedness is apparently present even in particles⁵. Levin concludes that there may be no true zero on the scale of consciousness. He also says TAME does not hold that mind is baked in regardless of structure, and that causal structure and cybernetic properties matter⁵. The passage does not say what the 1942 work argued.