Discussion of hybrid images

Detailed commentary on slide from my recent presentation

Slide 29 - Albert Einstein or Marilyn Monroe?

(picture credit: That picture is the work of researcher Gregory T Huang, from New Scientist's 31 March 2007 issue at newscientist.com, subscription required. The original covers a larger portion of the torso and the effect is much more pronounced.

===== Extended commentary on how the way humans process images can cause interpersonal conflict ==

You can see smaller partial versions of it here. The MIT Hybrid Image page is here with a three faces that are both smiling or/and frowning, depending on your distance. This illustrates the problem with the terms "or" and "and" when considering phenomena that cover a broad range of scales simultaneously. It can cause amazing conflict by two viewing groups who can't understand why the other group, looking at the "same thing" on a different scale or from a different distance, can possibly be so stupid as to see the "wrong" thing.

Another "angry illusion" on www.hemmy.net shows an angry and a calm face that change to the other expresion as you move further away.

It may be that a number of international conflicts are due to this exact phenomenon, or its analog, where those close to the front lines perceive very clearly one thing, and those at a comfortable distance away perceive something entirely different, leading to total internal chaos because of the mistaken stereotype that a perception "must be one or the other". It could explain some "you had to be there" excuses.
The sad truth is, like thinking there must be a "best", this is another comforting and simplifying error on the part of humans. Some things are both "chickens and eggs", and almost everything takes on that property as soon as you close the feedback loop of causality so A causes B which in turn causes A, and let it stabilize into a strange resonance state that we simply never see when feedback is not involved.

Regardless, this is one way that the situation at "the front lines" of an organization might be astoundingly different than the way it is so clearly and unambiguously viewed at the top. Repeatedly, the question asked by CEO's or top officials is "Why bother going down there? I can see from here!" The answer is that even you would see why if you went. That's why Toyota has a policy of "Genchi Genbutsu" -

You have to go down there and take the time necessary to settle down and actually see things from that viewpoint before you rush to judgment or speak or come up with "a plan" that won't have such a high risk of being nonsense or worse.

All our images, including mental ones that use the same hardware in our brains, have the problem of "filling in" gaps for us (to be "helpful", like Microsoft Word's paperclip) whether we asked them to or not. There is no way for us to know by looking at our mental picture that our head has papered over details it couldn't make sense of and replaced them with something that made more sense to it. (almost the definition of a magnetic "stereotype" that grabs hold if we get anywhere near it.)

Life is "fractally complex" and sometimes the fine-grained details change the entire equation. There are shapes, like the famous snowflake curve, that can be filled with paint but not painted. Our intuition misguides us. The "genchi genbutsu" rule of thumb is probably the safest bet.

Along with the other rule of thumb - "the time to furl your mainsail is the first time it occurs to you that maybe you should furl your mainsail." If it occurs to you that maybe you should go and look for yourself, don't put it off.

(from Rules of Thumb.)

Put in other terms, it's possible that there is way more energy in the high-frequency "details" than the low-frequency "overview", and that the details do NOT "go away" and can NOT be "put off till later."

That is pretty much the case with the Escher "Waterfall" picture. The tiny details that were wrong, that our eyes "helpfully" insist on discarding entirely for us from each local area of the picture, actually are coherently and systematically wrong and do not "cancel out." They are not "negligible" precisely for that reason.

In mathematical physics terms, we are used to the high-self-energy, low-interaction-energy world, where the inner product is
dominated by |a| and |b|. We tend to forget about <,> and can get away with it when thinking about rocks and simple machines. But when we get to social interactions, or plasma physics, or galactic centers and black holes, the interaction term <.> dominates, and the self-energy terms are negligible. 1 + 1 becomes dominated by the nature of "+" and doesn't care much about "1" any more. We have about zero intuition regarding that world, although we can compute the equations for it and simulate it.

Our problem is that we are trying to use cold-earth mathematics to design policies for high-energy social interactions. We try to leave out all the feedback, and all the cross-product interaction terms, and then are baffled that our results don't match the data. Hmmm.

The lessons we should learn from "system dynamics" or books like "Feedback control of dynamic systems" is that not only is the feedback usually not negligible, it actually dominates every other factor. If you want a good "first approximation", leave out all the other stuff and put back in the feedback structure and see what that gives you!

Are important vertical or horizontal loops clearly broken? Are unexpected loops clearly present, given observed behavior? Start there. If you get a "hit" don't bother with any other details until you get that fixed, because they'll all become irrelevant as soon as you reconnect, or disconnect the flow power loops. These are the kinds of "facts" that are way more important than legacy "data" that so confidently (and incorrectly) extrapolate to tell you your quest is "impossible" and "nothing can be done about that."

This whole subject touches a point that Frank Drake and Carl Sagan used to make repeatedly in class, back in my astrophysics days at Cornell University in the late 1960's. (Sagan was briefly, my advisor before he left for JPL to work on launching Voyager) I was a grad student at the Center for Radiophysics and Space Research, acronym "CRSR", but we all called it "Charlie's Radio Service and Repair. " The group did many things, including running the world's largest radio-telescope in Arecibo, Puerto Rico - made famous in the movie Contact (Jodie Foster) and also some James Bond movie where they fought in huge antenna.

Speaking of which, I see the Arecibo radio telescope will close if they can't raise $4 million this year. It's a sad loss - I was yelled at once by the designer of the antenna (who was my roommate) because I wasn't"symmetric about my z-axis". That was the night, after two months of struggling, he finally solved the 4-page equation for his PhD thesis, which he couldn't resist phrasing in the thesis as "It is obvious that ... "

But the passing of Arecibo, a very large but single dish, is part of the larger trend that synthesized virtual arrays made of of may smaller component telescopes are far cheaper to make and maintain and extend. As with "supercomputers", they stopped making "one" huge computer finally and now are grappling with the problem of how to get 800,000 smaller computers to "work together".

Amazing how that question keeps coming up.

Anyway - in what I term "Drake's Other law", Frank taught us that

Every time humans move to looking at a different level of the electromagnetic spectrum, we don't see just different sides of the same old things -- we see ENTIRELY NEW PREVIOUSLY UNSUSPECTED things going on.

So, back in visual days, no one knew the center of the galaxy was off in the direction of Sagittarius, but blocked by dust. We had to look in the radio frequency spectrum to discover that we were, like the Andromeda galaxy, in the midst of a huge "dish" of stars, about 2/3 of the way out from the center in one of the spiral arms.

My point is that, the same thing seems to be true of our observations of this "LIFE" thing we are part of. Every time we rotate the microscope lenses and change to a larger view, the whole nature of LIFE changes, as dramatically as Marilyn Monroe's image changed into Albert Einstein when you changed the viewing scale.

I warned about this in another post as well, discussing the properties of "mid-field" components of a radiation pattern of an antenna. The field is "obvious" and we can measure it reliably with technical equipment, and it clearly falls of as the inverse of the radius.

I said then

The cleanest and least ambiguous example, physically, is the middle-range field of a radiating dipole antenna. As discussed in a prior post, very near the antenna the power falls off as the inverse of distance. At long distances, the stable pattern can be measured to be falling off as the inverse cube of distance. And, in between, in the really annoying and complicated mid-range field, the pattern is unstable and appears, if measured, to fall off as the inverse square of distance. Worse, in the mid-range, some fields build up that behave as if they are about to be radiated into space, but then sort of change their mind and get basically sucked back into the antenna.

Very near the antenna, less than 1/10 of a wavelength away, life is good and the equations are easy. Very far from the antenna, over 100 wavelengths away, life is good and the equations are different, but easy.

In between, things get extraordinarily messy. Power seems to get created out of thin air, then go away again, if you "neglect" all the terms that are "negligible" at each end, but not in the middle.

There's a lesson there. In the creation of LIFE ON EARTH, we're in the "middle" part. You can't assume any term is "negligible" -- you have to check it out and be sure it is. And even then, you could be wrong tomorrow about what it turns out was true today.

You end up learning what it means to say "The future isn't what it used to be."
If the past changes in the future, and the future has changed from the past, you really can't be sure any extrapolation of the "present" is reliable, whether you can "prove it" or not.