Friston's "Free Energy" principle and PCT

This initial message is a copy-paste of the CSGnet message [Martin Taylor 2019.10.23.09.43]

On 2019/10/23 9:19 AM, Fred Nickols (fwnickols@gmail.com via csgnet Mailing List) wrote:

A Brit named Karl Friston has a theory he calls “free energy.” Here’s a snippet that caught my attention in a post on another list:

“The second law of thermodynamics tells us that the universe tends toward entropy, toward dissolution; but living things fiercely resist it. We wake up every morning nearly the same person we were the day before, with clear separations between our cells and organs, and between us and the world without. How? Friston’s free energy principle says that all life, at every scale of organization—from single cells to the human brain, with its billions of neurons—is driven by the same universal imperative, which can be reduced to a mathematical function. To be alive, he says, is to act in ways that reduce the gulf between your expectations and your sensory inputs. Or, in Fristonian terms, it is to minimize free energy .”

Sounds a bit like PCT. Here’s a link to the much longer piece.

The Genius Neuroscientist Who Might Hold the Key to True AI | WIRED

Yes, it’s not only very like PCT, but is actually implied by PCT. In other words, if PCT is correct, then so is Friston’s argument, but the reverse is not necessarily true. It’s a bit like the false syllogism: All men are mortal/ Fido is mortal/ Therefore Fido is a man .

I first came across Friston because Warren and Rick were arguing with him some years ago on a blog whose URL I forget. So far as I can tell, the key difference between PCT and Friston’s work is the implication he and his many followers draw from the idea in the quoted paragraph. In that paragraph, I would disagree strongly with the second-last sentence, for which I would substitute something along the lines of " To be alive is to act in ways that reduce the gulf between what you want to perceive and what your sensory inputs tell you is 'out there '."

“Fristonians” think reduction of free energy implies the necessity for computing output in real time in order to make perception become what it was predicted to be, whereas PCT says that you act in a direction that prior experience, in the form of reorganization, says would push the perception in a direction towards its reference value. For PCT, no on-line real-time computation is necessary.

Of course, in conscious control when we are thinking “how can we do this” rather than just tracking a cursor, we are actually doing “Fristonian” real-time computation of output, imagining – predicting – what the results of doing this as opposed to that might be, and ignoring changes in the disturbance while we follow the plan we concocted. That’s much slower than simply moving the perception in the right direction without further computation, but it’s what we have to do when we are confronted with a new problem that requires effective control of sequence that is not (yet) reorganized into the hierarchy.

Friston and PCT use the same hierarchic structure of control, but with a slightly different set of connections between levels on the perceptual side of the hierarchy. I have argued on CSGnet in couple of different threads over the last two or three years that the Friston circuit, which performs mathematically identically to the Powers circuit but is more flexible in operation, is likely to be the better approximation to the biological facts.

Friston has many followers, and it is possible that one of the ways we might propagate PCT is to persuade some of them that they are right to follow Friston’s free-energy principle and determine what it necessarily implies, but that the PCT concept of bringing perceptions closer to their reference values with little or no on-line computation is more parsimonious both as theory and in practical demand on neural resources than analyzing failures of prediction into patterns of output in real time.

Martin