"Active inference" is not far distant from "activist inference" in that to "actively infer" is to adopt the posture and mode of an experimentalist.
We don't sit around waiting for the environment to hand us faits accompli, from which we're to abstract a worldview. Rather, we know what our worldview entails, in terms of actionable testing routines, and therefore have ways to provoke model failure faster than if we'd waited for it to fail anyway, of its own accord.
In other words, to actively infer is to stress test your model at every opportunity, which needn't mean you yourself feel lots of stress. When you strongly and justifiably believe your model to be correct, and it appears to fail, you get to find your mistake and debug it, one of the most valuable exercises when it comes to keeping in shape and doing activism. Mostly, if it's a good model, it will appear to succeed.
The more one debugs oneself in training, in the staging area, the less one is likely to be forced into doing live re-coding on the fly, in production. Not that rewriting on the fly is forbidden (on the contrary), but it gets to be a hassle sometimes -- why not work out the most obvious kinks ahead of time? That's called being prepared, a first motto of scouting.
"Do I know how to set up a tent and hunt for food?" The couch potato read-all-about-it type might infer an affirmative answer, having followed the hero this far.
But what if we wish a life beyond that of a couch potato? In that case, practice setting up that tent yourself, in your backyard, and see if you're able to scrounge any provisions from your environment, without resorting to shoplifting or other types of theft.
In the Active Inference namespace, we have Free Energy in proportion to our need for it, in terms of reconciling discrepancies, between our simulated, imaginary worlds, and the real one.
When the simulation matches the reality seamlessly, there's zero friction, no chaos, all is smooth, with no energy unaccounted for.
Free Energy is like noise, error in the signal, or at minimum unused, still-available bandwidth.
The "activist inferer" is not going to sit around waiting for what's too big to fail, to fail. Better to stress test and prompt failure in controlled anticipatory circumstances, when it's not all about reacting in ways unplanned.
So is our activist in this sense an accelerationist?
Sometimes the billionaires especially get accused of wanting to put events on fast forward to make houses of cards topple, and ponzi schemes explode, not in years, but in hours. That's a stereotype, and has the connotation of "irresponsible change agent" given the implied more haphazard approach.
To that I'd say I'm more humble than that, thinking the wheels of evolution have a more inexorable pace, not subject to my own druthers.
The most I might do is catalyze, which is where experimentation comes in.
I'm using that term loosely by the way, but with emphasis on "replicable" as I'm thinking "by imitation" is the most serious form of propagation. Check for yourself whether what I say works, actually does work. YMMV.
