Computing With Art – Computerphile
Articles,  Blog

Computing With Art – Computerphile


We’re going to talk about Can we compute with art? can paintings compute and it’s it’s a broader theme in terms of What are the links between physics and chemistry and biology and computing and indeed art? So I worked with somebody called Linda Jackson who’s a local nottingham artist She’d started to play with something called acrylic pouring where basically you take different types of paint You put them down onto a canvas you let them mix and you let them dry and you end up with what I think Incredibly striking an incredibly beautiful patterns like this sort of foam like what we Scientifically called cellular networks where you can see these cells on different lens scales and right across the board in terms of different colors Etc. Those types of patterns are absolutely ubiquitous in nature right across very very large scale and scales in fact The large-scale structure the universe in terms of our galaxies are distributed best thought of his in terms of a cellular network or this type of Foam like structure all the way down to the really really really small and we’re in nanoscience group So we’re very very keen on the really really small We took a droplet of this which are nanoparticles five nanometers across Tiny tiny particles of gold drop them onto a surface and let the solvent evaporate Left us with patterns like this. So these are not the individual particles themselves This is what the particles do in terms of how they collect together and how they dry and you end up with these incredible cellular type patterns which are something like if you compare this with this in terms of the length scale, this is on a length scale about 25,000 times smaller. This is a microscope emits an atomic force microscope image So you’re sitting there as a computer file if you were going Okay, that’s nice who’s talking about science and he’s talking about art. When’s he gonna start talking about the computing? There are strong links here in terms of the physics and the chemistry of these patterns But the question is can you compute with curtains? can you compute with art the question here is not can we you know generate something that does much much better than Silicon technology CMOS technology in terms of the speed or the processing power, etc so it’s about thinking about different main sets and it’s a broader and I would say more philosophical question what is information and How do we process information? It’s a lovely piece of art what sort of computational problem can a piece of art like that solve? I mean is this to do with colors? Not really to do with colors much more to do with patterns in that we could just take all strip all the colors from this And just make it black and white or even to a certain extent almost binary I said, but what’s remarkable? Is that the physical process which is underlying the art here? Can be used to do a computation Let’s say you’ve got a set of schools across a district or a county or indeed across the country and you Want to think about the catchment area for those skills? What’s the best way of? Dividing up the land in terms of catchment areas the fairest way and to do that we use an approach from computational geometry called the Verona tessellation and so to compute that which you can Compute pretty straightforwardly, in fact, a lot of languages MATLAB included basically have the command built-in So there’s a distribution of points. It’s fairly ordered What we want to do is for each one of these points on here. We want to find the region that is Closest to that particular point basically we want to divide this up as fairly as possible So mathematically this is actually an geometrically is actually straightforward problem. So what we do is we connect up One of the points to all its nearest neighbors and then what we do is we take the Perpendicular bisectors, let’s do that. So ones there Ones here. So we’re bisecting these lines and ones there Ones there if you’ll excuse the wobbly drawing So what we have is that this area or the points within here are the closest to that point? Okay? This is like numberphile stuff come on, it does feel like numberphile, but we’re going to do a computation I promise at the end and we’re going to do a physical computation I promise what you’d end up with then as you can see is just a set of hexagons So that was an example of a very ordered Let’s take a less ordered set and a somewhat more Natural set perhaps this is actually from a paper by Meredith P Richards and what she did was look at distribution of schools in a district or neighborhood in Washington, I believe if I remember correctly This is what it actually looks like the black dots are where the schools are and the question then is how would you divide up? this area to be the fairest in terms of the Neighborhoods or the areas of the district that is served by each school The algorithm is exactly as we’ve just done for that ordered set is you connected up You connect a given point to its nearest neighbors You take the perpendicular bisectors and you divide up the plane that way that’s the fairest way of Dividing up the land as it where I found being kind of picky here I’m gonna suggest that perhaps some of those are more densely populated than others Is that too complicated for this? So that’s a really good point Sean So yes We’re making a number of different assumptions here or assuming that the density of the population is the same right across the board Obviously there are complications in terms of even transport links etcetera. There’s a wide range of different contributions we’re going to do what all good physicists do even computational physicists and approximate cows a large sphere PI’s around three Density population density is even across the board. We can set up a computer program to solve this. It’s relatively straightforward Let’s take an example Here’s Nottingham. That’s probably – higher density of points for schools. But I don’t know. Let’s say it’s coffee shops Well, what wanted to do is to try and work out was your closest coffee shop, which is the closest coffee shop to you so the way to solve that is Actually to do a Verona tessellation and so here’s our points Just taking away the map and we can calculate the Verona tessellation and it looks like that so again It’s not an ordered distribution of points. Therefore. We see a range of different polygons ranging from adult See if we can see a triangle in there. I don’t think so I think the smallest is four-sided maybe up to seven or eight sided cells Okay, so where is the physical computation well we can do it on a computer or what we can do is we can Take this as our computer This is now a computer. This is a physical computer. And what we’re going to do is going to take a droplet of these particles Put it on a surface let the solvent evaporate And what’s going to happen? Is that those? Particles are going to be carried by the tide of the solvent and get pushed together and they’re going to create our own Verona tessellation, let me show you exactly what I mean with a simulation So the yellow dots here are the nanoparticles one question you might ask is. Well, I can’t if I hold it up oh, they’re definitely nanoparticles and there are most definitely Nanoparticles in it. In fact, we did a sixty simplest video on this some time ago. You might ask Why does it look red isn’t gold gold? Why does this look right? There’s a lot of very interesting physics as to why this is right. The yellow dots are the nanoparticles The white is actually the solvent. So that’s the liquid in which the particles are dissolved. What’s going to happen? Is that the dark bits meet you’re about to see you can already see some? Black patches that’s where the solvent is evaporating. And so what’s going to happen is when we run this These holes are going to open up as the solvent evaporates the nanoparticles love to be wet by the solvent They looked at to be dissolved in the solvent so they will track back as a solvent dee wets as a solvent Evaporates the remaining solvent left on the surface will spread back like this carrying the nanoparticles with it What we’ll end up with in the end is a Voronoi tessellation. So let me run this So you see these opening up? Carrying the nanoparticles with them and you can already see the density and they collide And so what’s happening here is that you can see as these holes spread out they force the nanoparticles together and ultimately you end up with something that looks like that those tessellations are absolutely everywhere just Earlier this year. There was a scientific American article On this is not a beautiful image This is a dragonfly’s wings and you see this same type of tessellation everywhere near to the large-scale structure the universe giant’s causway announced from this is the cross-section through a cork from a wine bottle and The point I want to try and make for computerphile audience is in each case. These are effectively physical computers you might argue about the the universe what the large-scale structure universe, but in terms of the Physical and chemical processes that have been weakened exploited to do a computation Which gets me back to this? This was a distribution of coffee shops. Here’s our points for our coffee shops. Here’s a Verona tessellation lots of calculated by the computer However, this is what happens when we take a droplet of our particles put them on a surface and we look at the final stage And if we overlay this on this the physics has done the computation force. It’s not cared about How do I join up the nearest neighbors and how to get the perpendicular bisectors? It’s fallen out of the physics and you might think hang on there seems to be a bit of a trick there maybe there is but it’s all about thinking about computation and Information processing in different ways. This is almost This is computation not by algorithm but computation by analogy By trying to look at what’s happening in a physical system and say well actually can we exploit that? To do a computation for us. There must have been some kind of control in this there You must have decided where those have operational right? So that’s in this case I’ve cheated a little bit shown reason the points were distributed like this is that I work backwards from the image So cheated a little bit, but could we actually program this system? How would we program the system? Well, what we’d have to do is control where the evaporation happens Can we do that? Yes We can take our surface and what we can do is we can oxidize it using the tip of something called a scanning probe Microscope in this case. Those lines were about a hundred Nanometers wide something like are very very small lines Obviously you can some some will recognize the logo at least And what we can do is we can control how the the particles this is the real experiment This is the simulation we can control how the solvent evaporates from the surface by patterning the surface and so that’s a couple examples and we can put a lot of different patterns down there the next step in this is can we actually Program the Verona tessellation. Can we put little points two four the evaporation of those points and calculate the tessellation that’s something were Actually is going to be stimulated by this video can we part on a surface and drive the formation of the Verona tessellation and can we do a computation that way and Then the values of those array elements can be whatever you want, but they might be for example How much you spent on beer and pizza and coffee and so you could write a very simple program that would let simply say

100 Comments

  • LapTop006

    I thought this was going to be on McMillen's SIGBOVIK 2019 paper "93% of Paint Splatters are Valid Perl Programs"

  • Ari Rahikkala

    Scott Aaronson's "NP-complete Problems and Physical Reality" is a great way to get into more depth on this subject. Actually, I'd love to see the soap bubble experiment on the channel, I'm not sure I've ever seen it actually done on video. I mean, not that that's the important part, but it would be cool to see anyway.

  • Flurki

    Yea there actually sounds to be a bit of a trick there: When your definition of a computer is "something that behaves according to a set of rules" then everything is a computer, including your nanoparticles. Computation by analogy – can we "exploit" the laws of physics to achieve our goals? Well yes, that's what life has been doing ever since it came into existence. I can't see what's supposed to be special about that, because it's happening everywhere all the time.

    Prof. Moriarty, being interested in philosophical questions like "What is information?", seems to be too wound up in the apparent 'magic' of those analogies to see that there's really nothing spectacular going on here.

  • GZ

    I remember reading a story saying that people found out the way the city blocks were formed and connected somewhere in Japan resembles the patterns formed by bacteria if you plant the nutrients in a similar manner as the natural resources are distributed geographically.

  • Gabriel Medina

    Hi there, could you explain the relation of these concept with Mandelbrot Fractal, I understand his different aproach, but… exist some relationship with?
    Thanks, very interest topic
    gab (@gxsoft)

  • tkykhs

    This is certainly not "physical computation." He is talking more about simulating some interesting physical patterns in nature with computers. The end result is an approximation of a physical pattern, both of which can never be the same (i.e., not "computation" in its strict definition).

  • Ben Rowe

    I don't care that this isn't the usual kind of content we get from Computerphile. I'll take any excuse to watch Professor Moriarty.

  • Stefan v.i

    0:54 I don't see how these structures are similar at all… This cellular network structure is very densely packed, whilst the galaxies in the universe are incredibly spread out.

  • Luiz Sarchis

    Sorry to be close-minded but…I don't get it…this voronoi tessellation seems to be already solved by our algorithms…why have the trouble of dealing with nanoparticles and solvents and whatnot to do a "physical computation" of something that is already computed? Maybe a little motivation with some other idea that would solve a "hard" computational problem physically could be nice

  • Todd Marshall

    He's got it kind of backwards doesn't he. Take a spline. it's a simple way of running a smooth curve through a set of points. The computation challenge comes from simulating this spline. You're already using it is a computer.

  • Marc Shanahan

    I really wish there were subtitles, it's hard to catch the complicated terms. Does anyone know the two terms that are being said at 5:46 and then immediately at 5:47? McKauzalard's Sphere? Pyzorone 3? Google ain't finding any of this… :/ Please help.

  • blomeup2day

    It would seem this would be more of making art out of computing than vice versa. You are dictating the spaces for the art based on the computation. You manipulated the surface based on computation instead of achieving any data from random art that would be useful in any possible way. Then the question becomes why even involve art when it’s not necessary for computation?

  • Alex Kluew

    damn homies — that's some dope pattern recognition and pattern matching applications! nice! thanks for the video!

  • Greg Sherman

    I saw a video where they compute minimum distance between points with soap bubbles, but sometimes they only find local minima, so they have to prod the bubbles past the local, does this have the same drawbacks?

  • Shido

    Isn't the reason for the Nano particle liquid appearing red the same reason that yellow food coloring appears red? The higher concentration and smaller size cause the only reflected wavelengths to be the higher red ones?

  • eyeballin

    Reminds me of how a scientist simulated the most efficient possible transit routes between key locations using slime mold growth.

  • Penny Lane

    But if it's so straightforward to compute those tessellations, how would setting up such a physical computation ever be advantageous?

  • OrangeC7

    So, is the point that we may end up solving a computational problem (like P vs NP) not through a clever algorithm, but through a clever physical process?

  • Lucid Moses

    I don't know man, This guy lost a lot of credibility when he argued that scientist don't follow the scientific method and even, shouldn't follow it.

  • Matthew Jenkinson

    Ah thought it would be about Voronoi when I saw that thumbnail. Interesting to know more about it.

  • Andrew Kelley

    Oh heat.
    Take each building stick it on the opposite side of a piece of glass. Make each building be a reflection of its population density to its surface area and have each building be made out of a conductive material. Think copper blocks. Then carve your rodes to depth to speed limit on the upper side and the building's side face down. Level your surface then preform.
    On larger scale factors air current would be a problem but could be solved with a more complicated set up with the same basic scenario.
    There is also a factor of imprecision that can also accumulate on more complex patterns.

  • Aadit Doshi

    A video on shaders would be pretty cool. It's about programming materials and effects. They use a lot of cool tricks.

  • Arthur K

    This is kind of like the force-directed graph layout. For that, you can simulate the springs computationally, but can you not simulate liquid to achieve the same for Voronoi tessellations?

  • Unpronouncable

    Okay how do you control it? This technique is not going to help divide school map fairly if the pattern generated is random. Lets call the nano-particle suspension "public bitmap Tessellation_function(point[] a)" based on this video the function returns a bitmap with tessellation but has no inputs and it should have an array of points as input to be useful. also the function is unstable for few points. lets say I put an array of 1 on the surface there will be other cells generated around that point and the final image will be not what is expected. If my function returns 10 regions when I declared 3 points that kind of tessellation function is useless to me.

  • Chris W

    Since the artist relied on physics to make the pattern, if the art is used to compute, then physics is used to compute. That is not surprising since physics is generally described with math.

  • Charles Cox

    You are not defining what "fairest" is. In terms of area? In terms of population? In terms of power capita income? In terms of racial minority distribution? You can't approach a solution until you properly define the problem. Please, quit using incomplete examples.

  • Kyle Buchanan

    Seems to me if they made the surface as uniform as possible and then pulsed a laser on the specific points that the evaporation would then happen where they want.

  • Robert Boran

    I really don't like the term computation. It is misleading and very inaccurate. Computers are not computing. It is just a smart movement of electrons. The same here. The idea to use evaporation to create movement in order to achieve a pattern is computation and it is a brilliant idea. Our reality is generated in the future by the smart laws of quantum mechanics.

  • Kelly Stratton

    Question: Does anyone know of an example of this kind of 'non-traditional' computation that /isn't/ the result of the physical system falling into some local extrema? I've only ever seen examples where the 'result' is read after the system reaches an equilibrium. (soap films + steiner trees, or slime mold + network optimizations) – I'm having a hard time proving to myself that this always needs to be the case. More philosophically: why does using soap film or mold or gold nano particles feel fundamentally different than say, gathering up some bunny rabbits and waiting a while to calculate the nth fibonacci? (yeah, yeah it's not really that but you get what I mean).

  • David Fisher

    Nature creates cells that assign areas to nearest point. Voronoi invents an algorithm to reproduce that. Physicist is impressed by nature's ability to reproduce the algorithm that reproduces nature?

    Can it compute? No, it just grows/shrinks. What question can it answer other than the one it embodies?

  • totlyepic

    1. Surprised (but happy!) to see an South Carolina mug sitting on his desk.

    2. Would've been a better video if it were an ML person talking about the topic (since that's really where the geometric stuff shows up in computer science) or by a theory/complexity person if you wanted to talk about arbitrary NP-Complete systems or something.

    In the end, this had nothing at all to do with computation.

  • Ryan Roberson

    this comes out shortly after news that analogue computing is about to make a comeback using photonic reflection patterns to solve calculus

  • Alwin Mao

    A lot of comments disparage the examples shown in the video but the point isnt the specific examples but rather the way of thinking. Can we use beautiful behaviors in nature, capture them, and by analogy solve problems? Spaghetti sort. Light bulbs as the earliest nbody gravity solvers. Slimes growing to resemble highways. beauty as function in the animal world helps us optimize shapes for flight and flow. Analog computers or mechanisms that can do more. Can we use bacteria to compute, using their DNA? Can we use the physics of quantum mechanics to do problem X? The point is that we can use other disciplines to help us understand computable problems or in rare cases to do the computation more cheaply. This is most useful for problems we dont know how to do well, like factoring or protein folding, and esp. useful for thinking about problems that people havent even started to imagine using a computer for, because it would be too hard. Some leeway needs to be given to account for what the video leads to but is very difficult to imagine.

  • kebakent

    If we start piggybacking our calculations on the universe physics engine, the designer is just gonna shut it off. (joke)

  • jellyfishbones0

    In structural engineering they can also use physical analogies for computation. Hanging a chain between two points will give you a curve that is the ideal shape for an arch (say for a bridge or a wide doorway).

  • Daniel Abrahms

    This is such a tease. Keeps telling us hes gonna make these patterns do a real calculation, then at the end, explain that he means calculation as an "analogy". I tried that on a math test once. They didn't buy it, and I'm not buying this.

  • Frank Shrank

    what dialect is he speaking? "divide this obb". Is he forming his lips to voronoi-like diagrams while speaking? the mublephile edition.

  • Tyler Jackson

    What is the time complexity of performing the tessellation? If we consider this physical process as a single step of computing (albeit a slow one) can we solve any NP problems in polynomial time?

  • Rickey Bowers

    Making smaller substrate features in the production of microprocessors might benefit from a similar technique – lithography creates patterns at one scales, while physical computational processes modify that structure to produce complex structure at a lower scale. Some work has been done in this area with high power lasers, but there are many possibilities.

  • ardenthebibliophile

    Could you do this on a larger scale (e.g. 1ft square) by applying point sources of heat the a thin metal sheet with a liquid suspension on it?

  • S S

    Reminds me of some of the work of Alan Schoen when he was using soapy water to find minimal surface shapes.
    Also reminds me of a study I saw not too long ago about coffee nanoparticles and the way they dry at the bottom of a mug.

  • Dan McNeill

    Why not just let people pay to go to the school they want and have a real distributed algorithm? Anyway, you might want to check out "robust first computing." Also, there may be a way of using the cellular automaton "Life without Death" (or something similar) to produce circuits that could be used for computation (either traditional or as a kind of neural network).

  • Evgeny Selensky

    Nonetheless, it is a human (or its delegate, a computer PROGRAM) that is actually doing a computation. Physics is only providing the result of a process whereby a system is driven towards a state with min total potential energy, nothing more. How to INTERPRET this is altogether a different matter. For this you need a mind, a translator, a Peircean triple (sign, interpretant, referent). Fundamentally, computation is abstract, not physical. Living systems are different because they too have a purposefully built-in translation apparatus.

Leave a Reply

Your email address will not be published. Required fields are marked *