A more honest title would have been "Check out our animal shape generator", though I suppose it wouldn't sound as interesting.
They touch on the cool idea of using an image identification neural network to single out desired shapes from an arbitrarily large corpus of semi-random generated shapes. So the way you program your random generator could determine the style, while the identifier network would determine what will be represented.
I think your linkbait detector might be evolving to work like the pareidolia phenomenon mentioned in the article (our animal and people detectors' tendency to produce false positives).
This might not be the best title possible, but it's not clickbait-ey, it's just a little vague.
For some inexplicable reason, I find this earth-shatteringly fascinating. If such a bewildering array of patterns can be found in random noise, maybe all patterns -- our attempts to ascribe order to the universe -- are illusive.
But be careful about drawing metaphysical conclusions from this. This is only an example of selection bias. Take random noise, filter it and select examples that probably look animal-like to humans. If the classification algorithm is worth its salt, the selected images are bound to look interesting.
A comparison could be made to the Rorschach test; if the patient finds clear images in the ink blot, that is supposed to say something about the patient, not the image, because the image is by construction random.
> If such a bewildering array of patterns can be found in random noise, maybe all patterns -- our attempts to ascribe order to the universe -- are illusive.
It's certainly a justification for using statistics to back up what "feels" like a "true pattern" when really it's just noise.
the shower of an apartment i rented long ago had old-school spatter tiles with small random shapes remarkably similar to the ones in the article - very close to what's in the links below, although on the ones i recall, the shapes had smoother edges and more small white dots inside the shapes
with the hot water enveloping me in a multi-sensory white noise, i would stand there for time=n finding faces, animals and aliens in the splotches - very zen
Awesome ! I also see a lot of living creatures in flames (try to take pictures with a DSLR during 1/2000 second ;) ), including phoenixes, seehorses and riders.
I love that every post I see linked from Wolfram is so over the top that it borders on satire. Especially the posts from the author of the software. It's so incredibly powerful in such odd ways.
The examples of him being super egotistical that I'm aware of are from a while ago. I'm thinking of this kind of thing, where he thinks he knows everything about every field: https://en.wikipedia.org/wiki/A_New_Kind_of_Science
The question of pattern recognition in random noise is interesting and a legitimate question. Animals (given our propensity to see then) and an interesting subject matter.
Generating random noise and then filtering it to "find" patterns seems dubious, at best. Why not just generate animal shapes in the first place?
Starting from a random image it's possible to manipulate and filter it to find literally anything. It doesn't mean there are "animal pictures" in the random data; it means the image has been manipulated to look like it. It's a tautology.
In a sufficiently large random image, you'll (probably) find as detailed of a picture of a rabbit as you like.
The question is about how often image recognition (either human or algorithmic) picks up a signal in random noise -- that is, how often we see things that aren't there (or if you prefer, are there, but by chance). Of course you can find anything, the question is with what frequency.
If you've ever spent much time looking at random patterns, you'd know that we're prone to seeing things that aren't there in noise. Modeling that phenomenon is interesting (at least to me).
I don't know much about wolfram things, but just looking through it, it appears that the manipulation and filtering is grouping the randomly generated data (I would argue it's still random), taking only groups between a certain size(this is the least random part, it makes sure there are enough "features" yet not too many), and then smoothing out the groups for display (still completely dependant on the data, which was random).
As the "animals" are smoothed groups of random data, (and differ every time) , I'd argue it's still (basically) random.
The premise of extracting meaning from noise is flawed. It doesn't matter what process you use to arrive at the meaningful conclusions if the source material is completely meaningless. It's analogous to a fallacy of false premise: you can literally arrive at any conclusion given a false assumption.
"random" encompasses an infinite array of variable parameters -- linear blur, Gaussian blur; white noise, brown noise, pink noise -- some of these can tend to make animals appear in dot patterns, or melodic phrases arise from static, or whisky taste of butterscotch and leather
They touch on the cool idea of using an image identification neural network to single out desired shapes from an arbitrarily large corpus of semi-random generated shapes. So the way you program your random generator could determine the style, while the identifier network would determine what will be represented.