Ben Goertzel @bengoertzel has a new paper that formalizes the notion of 'Occam's Razor'. I certainly agree with this notion that we have to think of Occam's Razor in more general terms. arxiv.org/abs/2004.05269
My hypothesis is that it is not only general intelligences but also all of life evolves through a principle that is a general form of Occam's Razor. I am glad Goertzel has taken a stab at formalizing this.
The three most common invocations of Occam's Razor that he writes is: (1) Grounded in Physics (i.e. least energy) (2) Grounded via Cognitive Theory (i.e. simple models) and (3) Grounded via Computing (i.e. short descriptions).
This is of course does not cover the entire space of possible notions of Occam's Razor. The biological notion is one that isn't covered by the 3 mentioned above. Although it is related by analogy to the first.
Biology does not have simple models of this world nor do they have short description. In stark contrast, biology tends to be bureaucratic and gratuitously complex. So how does one square with the idea of Occam's razor?
Underlying biology is a set of principles that reveal a higher implicate order. These same principles are I hypothesize are the same principle that creates minds.
This discussion of Occam's Razor cannot is entangled also with the notion of a complexity measure. The problem with complexity measures is that there are so many to chose from!
Complexity measures are emergent phenomena in the same category as entropy and temperature. They are what are observed as a macroscale as a consequence of emergent phenomena in microscale.
As a consequence, there are fundamental limits to their resolving power. We cannot understand a microprocessor by simply measuring what parts are hotter than other parts. It is informative, but not informative enough.
Therefore we must think differently here and identify the first principles. An excellent example of this thought process can be found in Constructor Theory where a version of the 2nd Law is derived from first principles.
A kind of Occam's Razor is one that is based on first principles. Similar to the conservation of energy. That is, one where all interactions must abide to.
In fact @ylecun argues for this kind even though he may not be aware of it. That is, when devising a new neural network one uses as few priors as possible. In effect, using as few principles as possible.
The complexity of this world and its biosphere are a consequence of a few principles that give consistent rise to new emergences. The ultimate cognitive emergence is known as empathy.
I agree with @bengoertzel that 'one doesn't want one's theory of cognition to depend on particular aspects of physics and computing infrastructure'. This can cause a fit for many philosophically, but I think he's in solid ground here.
The explanation for this relates to an understanding of emergent phenomena. It's a non-issue, however, philosophers will continue to debate this.
Anyway, there are lots of nuggets of wisdom in this paper that I think I can definitely use in my book. gum.co/empathy
@threadreaderapp unroll
