Toward Experimental Techniques
If I recognize a pattern, a machine should be able to as well, given enough training data. However, a machine might also recognize something other than what I am seeing. Therefore it is better to indicate to the machine what exactly I have in mind. This is ‘supervised learning.’
But what is actually happening when I ‘recognize’ something? Yes there is generalization and yes, there is repetition. But it would seem that at least a few fundamental processes are at work. These include:
Transformation (as in matrix transformation) What we see as we walk along a window and watch the panes transform from squares to skewed parallelograms.
Sequential zoom — Zooming in and zooming out, both simultaneously and continuously. By which I mean: there is a simultaneous processing of information at an atomic, a molecular, and an object/class level. We recognize via scanning, transformation, and memory.
But which goes where? How do we figure out whether we’re zooming in or zooming out, or moving one direction or another from a shift in perspective, or whether the idea becomes weaker with the inverse of the square of the distance (to what?… to what it was just before we began this process of répérer*)? It’s the ‘moving along’ with which we communicate. The moving along, the changing, relative to what just was, and anticipation (prediction) of what will be. (This is reminiscent of the example given in Jeff Hawkins’s excellent On Intelligence, in which he describes the expectation, as one climbs down a ladder, of where that last rung will be. If it isn’t there, we are surprised, and must react.
If I say, “Well, that’s all in a day’s _________” and you fill in the blank with “cheetah” instead of “work,” I’m confused. Are you making a joke? Is there something else you’re about to tell me, like, maybe that’s the name of a new movie? Have you lost your mind? Or what? In this way it’s a directional curve, and potentially more than one. It may in fact always be a series of intersecting curves, or even of ‘vaguely parallel’ or correlational curve-shapes. It does seem that any any point you can stop and calculate the derivative of the idea shape, aka the derivative of thought, unless of course the thought itself is the derivative. It may be that thinking about the thought-shape is a different activity precisely because it is the derivative, or the integral, of what came before, or in parallel.
However this is done, it must be done through a multiplicity of data points, and a multiplicity of techniques. One technique must be repeated as it applies to what works for one type of abstracted shape, and another for another. So, like music, it becomes an interplay of variations on a theme, surprise, delight, the sudden shifting of gears intermingled with expected harmonious interplay like the ebb and flow of waves on the shore; it also becomes variations on the variations, ie another methodology entirely.
It becomes, in a way, the Calculus of Everyday Life.
We can start by defining, with an intelligent hand-wave, some agreed-upon data points. The experience of traveling through a story, in a kind of eager trance, is familiar to all of us who watch tv, see movies, or read novels.
Not just the physics of meaning, in an absolute sense, without an observer, but perspective, which includes the observer. Can we incorporate things like perspective (seen at a distance, the 30,00 foot view, zooming in and zooming out)?
“Where are you going with this?” When I decide to go take the van to the body shop, I have a destination. When I decide to go water the plants, I have a destination. The destination is the faucet outside, because I’ve set up a complicated sprinkler system. By decoupling the circular sprinkler from the impact rain gun sprinklers, the oscillating sprinklers from the soak hoses, and by having each juncture individually controllable in terms of flow, I’m able to have a single destination: the faucet. I turn it on, and the flowerbeds get correctly watered.
Now, this is true of just about everything: every goal has preparation attached to it, and consists of some pre-work in order to get to the destination. Getting to the destination itself (the activity) is also composed of multiple pre-steps. If I’m going to the body shop, I have to get the car key, walk to the car, create a flash on the route of how to get there, push down on the accelerator, etc. This is the anatomy of the activity — these are the things that are done by habit and arent really thought about much. What’s in my head is the goal: get to the body shop and ask what is going to cost to fix the rust patch.
At each step along the way I can pay attention to the step and not the goal. I have both available in my head. I will only attend to one of the steps along the way if something goes wrong (where are my keys? why won’t the car start? what? there’s a detour?) but the point is that I have the large idea in my head.
This is similar to following a story. I’ve lately started watching Designated Survivor on Netflix. The premise is that the U.S. capitol is attacked during the State of the Union address, and the entire government (with one exception) is wiped out. That happens within the first few minutes of episode one. Then, they do a ‘fifteen hours earlier’ flashback. That future event then hangs over all of what we see, and colors how we see it.
(“Colors how we see it ”— that is a color transform.)
The way it ‘hangs over us’ must be derived from the way a destination ‘hangs over us.’ It’s a goal, a purpose, and it’s larger than the individual steps required to get there. (derived from — does this suggest that it is a derivative?)
What if we take
Jared Kushner was always a little too blank, too lifeless — as one Twitter wag had it, “the Madame Tussauds version of Jared Kushner.”
We flash on a wax likeness of someone, and, having been pointed in the desired direction (primed, really) by the phrase “a little too blank, too lifeless,” we’re able to get the idea when we get to the Madame Tussaud’s version of Jared Kushner.
This is what idea-shaping does. It points us in a direction, shapes that direction. The destination kind of doesn’t exist, at least not until we get there. This is why the expression ‘where are you going with this?’ works. What we expect, though, is that there will be some destination: there will be a point, after all. Otherwise we are left with the long and meandering road to nowhere, which is the opposite of communication, and our natural response to this will be boredom, irritation, confusion, anger, and the eventual shunning of the speaker. Who wants to listen to stories that go nowhere? The general elements of a destination as a place toward which we go doesn’t necessarily mean that we actually arrive there, or even that it is a definite, definable place. We live in a world of instances but we perceive them as a generalized representation. Where I’m going, when I go to the body shop, is a generalized pre-imagined version of a shop. It’s just some place, a place where they do body work. If it’s the first time I’ve been there, it’s a vague idea of a garage, maybe with paint-sprayers if I think it through to that level of detail.
*répérer – a French word, meaning to orient, to navigate, or to locate with precision in space. The reflexive form se répérer means to get your bearings. I prefer it to the English because of the implications it contains