Sunday, April 17, 2016

Perceiving the Flow of Reality

It seems natural to say that we see the world as it is presented to us, a continuous flow of new situations and new information. On the other hand, if you've ever seen an optical illusion, you know that perception is imperfect. How do we interact with the world?
This study says that our perceptions come to us in the form of "'time slices' lasting only milliseconds." Our mind perceives the world as a moment: all at once for about a quarter of a second, then it goes on to perceive the next moment.
The description reminds me of Ridley Scott's filming technique in the most intense battle scenes of Gladiator. (starting at about 4 minutes)

A couple of programmers read this and it reminded them of real-time programming:
I write code like this and it doesn't imply 20 frame per second is the limit to human have to think like an algorithmic programmer: one thread captures the raw data (the 400ms time slice) and then another thread processes and exploits it (the 50ms time slice). They run in parallel. At least they do in my code. Serial processing would induce noticeable lag (400ms is a LONG time). It also makes sense that the exploitation thread works faster than the capture thread. When needed, you can amp up the attention and still keep up.
Unless you've been drinking, in which case the processing thread slows below the capture thread. That's explains why all the girls look good the closer you get to last call.
Another says,
That sort of graceful degradation is why people don't go from stone cold sober straight to passing out, when their time to process the data regrettably exceeds the allocated time slice. The processing just gets less detailed.
Alas, that's why us old folks seem stupid when we're hurried, too.
And another,
It seems very likely biological information processes is extremely different from the kind of code we write. The roadblock is that the individual neuron "clock" speed is no higher than about 250 Hz. That needn't mean, as you say, that the brain can't comprehend more than, say, 20 complex events per second, but it has to do it in parallel, and it has to "assemble" the comprehension after the events -- that is, it can't be done in real time. Similarly, to command high speed complex chains of actions, e.g. a gymnastic flip, the brain has to assemble the train ahead of time and anticipate what will happen, a genuine tour de force of solution of some pretty complicated differential equations.
Yeah, you hit it. I design "High-Performance" systems. A few years back this involved what many would call super-computers: multiple behemoths weighing as much as a car.
The last HPC system I designed used 4,800 little machines working in parallel. Once proven, the customer doubled that. That would take your 250 Hz clock and multiply it linearly.
The brain, I think, works much the same way. We (HPC nerds) talk about it now and again. It's obvious our grey matter is exceptionally efficient in processing inputs in parallel, even though we consciously attend to things in a single thread. I don't want to dumb down the ideas too much, but it's like a software program processing data in the background while the user is looking at the user interface.
For all our technology, we are barely catching up with nature.
another other:
" ...the elements of a visual scene are first unconsciously analysed."
That's not exactly a new model, it has been around for well over a decade and is epitomized by the saying "you don't run from the bear because you are afraid of it, you are afraid of the bear because you run from it."
This seems related to Veritasium's observations about eye-tracking.

(May'16) What that study says about all perception is similar to the better known visual phenomenon called "persistance of vision" or "how movies work". Destin explains it all three different ways.

(Dec'16) Why time "slows down" in times of danger. It's not that you "see more clearly" or that your "senses are sharper". It's that you remember more frames: more time slices.

No comments:

Post a Comment