discussion on concepts, ideas, practices, and the broader interdisciplinary built of culture, art, politics, and environmentalism from various perspectives. And the School of Materialist Research was founded to Arizona State University, Institute of Social Sciences and Humanities, also the Academy of Improuvel, and the Department of Architecture, Theory, and Philosophy at the University at UBN. with the focus of the intersection of humanities, social sciences, and creative groups that address materials, but I mean through contemporary design,
mathematics, fields of the art, and many other disciplines. So we have a great pleasure tonight as a guest speaker to have Matthew Fuller. Matthew is a professor in cultural studies at Colesmith University in London. is for his writings in the field of the media theory, software studies, outdoor studies, critical theory. Some of his titles are media ecologies, further how to sleep in art, biology, and outdoors. And also something that I read recently is his collaborative book with our wise man investigative status.
So the story tonight is maybe the Society of the Building. Injil. Thank you. Thanks, Tihemia and Katerina for the invitation to come. And what I want to do is show an argument, there's a loose argument that's being currently developed around the idea of filters as a constitutive form in present society. So look at a number of ways in which, first of all, to explain what filters are in different ways, and then to talk through some of the ways they can be read as constituting present social forms. And this is done
often through aesthetic approaches, but also to try and address some commonalities between different disciplines. So sometimes the presentation will move through different kinds of discipline, from physics or mathematics to computer science to gender studies and so on, but to try and think think through the different ways in which these articulate filters in different ways. I should say also if any of the material is kind of annoying or boring or not understandable,
first of all embrace boredom. After the first 60 minutes it's going to get interesting. the first 120 minutes maybe or maybe by the first two minutes you'll be thinking it's 120 minutes already but just think of that as part of the kind of experiential process of your brain filtering all this noise but if you do have yeah questions if something's completely opaque do just interrupt me and ask a question because I'd sooner actually communicate something even at this distance this great distance in the room. So, yeah, do feel free to ask questions for clarification and just shout or hold your hand up or whatever. So just to talk about filters, I want to start with an artwork,
which is this work by Larry Bell, who's an L.A.-based artist. And what he does is work over a long period of time with glass treated in different ways. It's something that kind of resonates with this building in some ways, a kind of re-versioning of kind of modernist tropes of the square, of clear glass. glass. But this work involves filters in a lot of series of works that he has are about glass treated in different ways with thin layers of chrome in this case that are placed together that in their overlapping produce interesting kind of rainbow effects or interference
effects. So that's what I'm interested in is what these films are. Okay so this is an interesting example of a kind of overlapping and a kind of time delay acting as a as a filter. You always get one rude member of the audience, it's terrible. So yeah so this This production of filters overlapping producing this moiré effect or this interference effect or production of rainbows is what I'm interested in. The filter obviously has other resonances and this image of the contemporary filter
that we're most familiar with in a way is a way of living with the filters allow living with a condition. So the society, the filter proposition is a certain way of inhabiting space, spaces, collective spaces, individual spaces through through processes of filtering. can also say that the filter is also comes in different forms so as a form of spatial organization it arranges it is one of the ways in which nation states arrange themselves this is the visa form that you have to fill out as a student entering the uk for instance so the filter has harder and
softer versions and harder and softer kinds of rhetoric associated with it, which is not to say that all filters are the same. The filter rises in physics as a way of understanding, in this case, how heat passes through a space. So Joseph Fourier's theory on the analytical theory of heat from the early 19th century was something that allowed physics to move from a mechanical notion of the universe to something that involved thermodynamics, that involved the movement of
energies through substances. Fourier described the movement of energies in terms of waves. So those of you who took part in a performance just now where there was a description of a wave by the movement of human bodies holding football gave a very clear example of the kind of wave that Fourier was describing. Fourier analysis is fundamental to digital signal processing and most electronic music, well pretty much all electronic music for instance, will be composed by software that runs a version of these tools. So what he allows for is an understanding of
energies moving through physical systems and of physical systems being things that filter energetic flows. So it's a way of understanding physics as systems of interlaced and overlapping waves that filter and material objects that filter structures of energy. Fourier's analysis also goes into further into the analysis of data and the way in which data is distributed in contemporary terms in more or less in distribution in distributions of what is possible what is actual what is potential and so on so it's core to the description of big data
and the politics of virtualities that are integral to these kind of techniques, some of which I'll look at in a little while. One of the things that Fourier's work first set out is a description of the way the Earth's climate holds heat in. So it's the first description of the greenhouse effect, So from 1822, the first description of climactic heating via the impact of the sun's rays on the planet. So when we think that the present crisis of climate damage is presented simply as a contemporary phenomenon, phenomena and that knowledge about it has only been recent, we can remember that this
was known from the early to mid 19th century in the communities of physicians or physicists rather. Fourier's work was taken by Claude Shannon into a very famous piece of work that most of you are probably familiar with, the grounds information theory and theories of communication in this famous diagram as a way of calculating noise. partial differential equations that Fourier relied on were also deployed by Shannon in order to calculate the way in which information can pass from point A from an information
source to point B, a destination with the maximum economy. So how much interference, how much noise could be received before a message became uninterpretable. So it's a of thinking about how the medium in which a message is passed, the medium in which the message is transmitted, is participated in by, in this case, the copper wire that Shannon was calculating the parameters of, but also how much information can be lost. So Shannon's theory was about the way in which communication always involved the participation of a medium,
so the specific case of telephone technology, and the way in which the medium reduced and filtered the signal, and the way in which the information source and the information receiver also acted as participatory filters on the information that each other was providing. So there's a sense in which the construction of communication is also the construction of obfuscations to communication, techniques of hearing, of listening, of passing information from one place to another also construct forms of noise. They construct forms of the way in which the media itself interferes with, participates in the formation of that message.
This work by Shannon then went on to develop aesthetics of its own. The project that he worked on following was the technique of the vocoder, which was first used as a way of encrypting live speech in the Second World War between different locations, mainly kind of very concentrated state locations because the technology at that point was exceptionally expensive. So the vocoder, as you can see from this diagram, the schematic circuit of the vocoder takes a speech signal and you can see it takes the signal from something inside the head called
the brain, which something will come to in a while. But it's interesting that the brain is seen as part of the actual circuit and the circuit diagram. And what it does is take the waveform that reaches the microphone and then transposes different parts of that waveform by random methods. And it produces this very kind of robotic voice sound that some of you are probably familiar with from electro music and hip hop music. something that is the basis for contemporary aesthetics of music with things like auto-tune. So the kind of controversial way in which pitch correction is used by singers to create a robotic sounding voice.
The key innovator in this is obviously the R&B singer T. Payne. And this is, as you can see in the interface to this pitch correction, what it does is take particular fragments of the waveform and move them up or down in terms of the register of pitch. So it's a form of filter that is derived directly from calculations by Shannon of the economic value of a telephone call. So how much information can be pushed down a thin copper wire at any any one point becomes then this calculation becomes a means for determining meaning within a message.
It then goes on to determine, to be used to describe ways in which voice messages can be encrypted in the vocoder. And the same technique then goes on to describe the way in which the human voice can become robotified, can become expressive in different ways in collaboration with particular pieces of software. Another form of filter I want to look at is a particular argument from Ludwig Wittgenstein's Tractatus Logico Philosophicus, which is one of the kind of key texts for thinking about the logical formation of language.
So to think about how language, construction of meaning in formal and logical descriptions can be a grounds itself for a kind of filtering. And one of the things that Wittgenstein does is to argue in this book, in a series of passages that I've listed at the bottom of this slide, that one way to understand the accuracy or to analogize the accuracy of a logical statement is to compare it to the interaction between a mesh, a network diagram, or sorry, a network and a picture. So it says, imagine a piece of paper with a series of black dots on it.
And that piece of paper is read by a mesh. Very fine. The mesh could have square shapes. It could have triangular shapes, hexagonal shapes. But it's made of wire. And you look through the mesh at the image and determine whether that image, by its match to the mesh, by matching the image to the mesh, you determine whether that image contains a logically true statement. So the mesh is an analogy to the axioms, the logical ordering system that describes whether that image can be true, said to be true or not. But we can also say that Wittgenstein's analogy or Wittgenstein's story in the present phase of digital technology can be read differently.
So each image that Wittgenstein argues can create a logical statement in this mechanism could also have different readings depending on what kind of mesh you put upon it. So you have a different sized mesh, a different mesh with finer or grosser holes, and you can produce separate readings from it. And this is why I use the term steganography to describe it. Steganography is a technique in encryption in which an image, so that image there, is used to hide a secret message. so that image may say that each tone of grey
there are 26 tones of grey say and each of those messages each of those tones of grey corresponds to a letter in the alphabet or the Roman alphabet and that image can then be read off as reading reading as a series of letters those letters can then correspond to a series of statements or they can be further encrypted what Wittgenstein's proposition allows us to think though is that there are different ways of representing logical statements different ways of representing systems of truth that the the same statement can be encoded as a series of dots or it can be encoded
is an alphabet, and systems for reading truth can be encoded in something as simple as a wire mesh, depending on the structure of the argument being made. So here we have logic and a highly formalized description of truth encoded as a technique that is essentially equivalent to to a software filter or any other kind of filter. Wittgenstein's argument about the grid comes up again in, or comes up later in this building that he co-designed with the architect, Paul Engelmann, which some of you will be familiar
with in Vienna. This building had highly detailed grid structure that was built up, if you think of the building as a series of pixels rather than as a grid, the basic unit was a gesture like the door handle and the movement within the building can be read as a series of movements through a pixel structure or a voxel structure, like a three-dimensional cubic pixel. Here you can say that each gesture within the building, each movement of bodies within the building is already encoded by, just as Wittgenstein's grid space for descriptions of truth
or true statements or the allocation of true statements, described, could be described by that mesh. The building describes all the possible movements that could be made within the grammar of action allowed by that building. So it's a highly formalized, extremely pedantic building. Many of you will know the famous anecdote where there was an argument between Wittgenstein and the builders. They were asking him, do you really need that ceiling to be three millimeters higher. It was a big scandal. The builders were arguing, no, no, it's utterly irrelevant. For Wittgenstein, it was absolutely crucial that that three millimeters was, that the ceiling was raised by the necessary three millimeters. So the precision of the grid
It became the grid that was able to enable the correct enunciation of the axioms for movement, the axioms for living that were set out in that space. So in a way in this talk, this kind of encapsulates the movement between systems of logical representation and systems for axioms for living, axioms for movement, axioms for becoming that are set out in different systems. A related system, of course, D'Albrecht Dürer's work on perspective, the grid as a mechanism for describing physical space and the introduction of techniques for the construction of perspective.
You'll see by the male figure on the right, keeps their eye specifically on the tip of this obelisk. So you imagine it's quite precarious position for your eye to be in at the very point of this sharp needle. And the recumbent figure, who we can imagine is not the artist in this case, is described by the grid that the artist looks through. You can see this also as a kind of bureaucratic transition, people sitting on one side of a desk or another and being subject to different forms of scrutiny, so related back in a way to the image of the visa application form, except when you're applying for a visa, the
bureaucrat doesn't sit there with his eye next to a needle and you don't have to expose yourself in quite the same way but just via different bureaucratic categories. This object that is very similar in some ways to Dürer's system of optics is the first system for for visual recognition or for recognition of visual input by a computer. It's a 1950s computer called the Perceptron Mark I. The size of this computer was probably about an eighth of the size of this room. And what it could do was to recognize basic shapes.
It could recognize a triangle, a square, a circle, and some letters, but not all letters. And you can see that it has a number of different features. On the left, you'll see that it has what are called retinal units. And these are basically light-sensitive diodes that would receive light. This then would be passed through to a circuit that would then evaluate the weight of the strength of the light wave coming through the retinal units and evaluate whether it was in light or dark. So whether it was black or white.
So it's a very simple form of what we would now understand as feature detection. This would then go through a very thin layer of association units which would further evaluate the weighting of the stimulus, whether it corresponded to black or white, black or white, and then give a reading that would allow the operator of the computer to determine whether the computer was looking at something that was corresponded to a shape of different kinds. What this network is, is designed to be a logical equivalent of the retina, of the eye and the brain.
So we've had the brain and the voice. Now we have the eye and the brain as logically described. So this work builds on work by Warren Pitts, Pitts and McCulloch's work on a logical description of the brain as a calculus. And this piece of work sets up the basic operation of what we now know as the fundamental techniques of artificial intelligence. So the neural network in this case was hard-wired, so it's a collection of wires. You can see possibly the size of the cabinet, which contains thousands of wires.
thousands of wires. And if you look online for the perceptron mark one, you'll see this enormous wealth of spaghetti of wires within these cases. And the complexity and bulk of this machine made sure that it was never really followed up. And now we have what followed was logical descriptions software of these networks. Contemporarily, these move to systems such as this. This is a diagram from a textbook on machine learning, the particular set of techniques around deep learning networks.
So you're probably familiar with these kinds of diagrams where there's an input layer that makes a similar kind of calculation on a set of inputs, say a camera looking at a street to see if it can recognize faces, for instance facial recognition technology. And each pixel, so several thousand pixels in every image, is evaluated for its luminosity, the kind of color it has, and those Those pixels are then aggregated to see if features can be detected across them. And these networks have often many, many layers, not just three layers, maybe thousands of layers.
And what they do is take a numerical feed from one layer, feed it back to the next one, carry out calculations to work out the relation between layer one and layer two. And gradually, the machine is taught to discriminate significant features from insignificant features. So the basic techniques of machine learning. And once it's trained, the machine can then be turned on to any data set to see if the data corresponds to the features that it's been taught to recognize.
So this system of input, of feeding through layers, of adding a numerical value to the transition from one layer to another, to the next layer, gradually is a system of training the machine, training the network to recognize features, to allocate the numerical values, and then to produce an output at the other end, which supposedly corresponds to the trained features. So a trained network might be used to look for a certain kind of tank, for instance, in a field, or it might be used to look for certain faces in a crowd, for instance,
and it would be trained to look for those specific kind of features. The neural networks are also, as we know, at the core of things such as social media. So many of you will be familiar with the debates around discrimination and bias within systems such as Facebook and other mechanisms which run on these basic technologies. So one of the ways in which these work in relationship to questions of computing more broadly and how computing acts as the kind of filtering machine par excellence is through a notion from computer science or a key conceptual layer,
a key concept in computer science, which is the idea of the abstraction layer. So the abstraction layer is a term that describes the way in which a computer represents what it's doing at different levels up to the level of the user interface. So at one level, a computer is simply a series of plastic and metals that is organized in a certain way in order to allow electrons to pass through it. The next layer up will be a logical description of that structure that allows the computer then to control that process of movement of electrons and to balance
for any errors that the electrons make. Another layer will be, for instance, the operating system that allows the software to talk to the user. Other layers will be the graphic user interface. So an average computer contains around 14 layers of abstractions. So each layer sends different signals to the other, different kinds of signals to the other, depending on the language it's using, systems of encoding that allow for one level compute to communicate with another and these all act as different levels of the different levels of abstraction allow for different kinds of filtering so when you're
writing in a programming language you're writing in a way that is closer to the way that the computer experiences the world the way the computer undergoes the world but you're never communicating directly with the metal or the electrons even if you're writing in assembly language which is a very simplified form of encoding. So the abstraction layer is fundamental to contemporary technology and that means to contemporary culture. If we think that most cultural activities we engage in in the present we at least um at least partially digital in terms of their their formation or at least partially digital in the way we interact with them
it means we have to understand uh the architecture of computers is fundamentally cultural fundamentally social and so the types of abstraction that the computer unfolds or the computer puts in place a core to our understanding of contemporary culture. Here we see an example, a crude example of analysis of a face and the association of different tones and different regions of the face with different numerical values, which is then turned into a matrix of numbers. So this image on the left of a kind of mosaic is logically equivalent in this structure to the structure on the right, which is simple matrix
numbers. We see the correlation between this kind of image and the argument made by Wittgenstein about the mesh, that the mesh would be a way of disclosing or disproving a truth in a statement if the statement was written out as a series of dots. This mode of description of the world expands into more and more types. Many of you will be familiar with some of the recent artworks generated in artificial intelligence that have created different kinds of controversies around authorship, different kinds of controversies around questions of creativity, of where the locus of art is. And one of the arguments is that the
locus of art now becomes the construction of filters, the construction of systems for the weighting and compressing and uncompressing, the coding and encoding of numerical values in systems such as this. This is a diagram of a convolutional neural network, which is a particular kind of network which takes a series of numbers, combines them to produce new results, takes those numbers combines them again to produce new results very much like a weaving process but these kinds of process also have results that we're very familiar with in the present and produce certain kinds of conditions
of possibility for subjects, conditions of possibility for cultural processes, and in turn are also things that magnify, amplify, and buffer or weaken other kinds of cultural process. So when we're talking about filters, we're not simply talking about logical processes, We're talking about filters that also enfold and engage with other processes in the world. And obviously a key system that interacts with filtering is the question of gender and the representation of subjects as gendered subjects.
Over the last years, one of the debates around filters as cultural objects has been around representations of faces. So the idea of the so-called beauty face that's built into a lot of apps. For instance, this is the Samsung version of Android and the beauty face filter that's built into the Android camera allows you to modify your face. in this way in order that your eyes become bigger, your cheekbones become more prominent, your face is smooth, your chin becomes more blended in to a thinner point, and your face
corresponds more finely to a series of geometrically described norms which are described described or considered to be corresponding to other notions and codifications of beauty. Another company, FaceApp, has deliberately taken a controversy courting position so that has let's say even more extreme filtering techniques that allow people to manipulate their face in many different ways but also what what we have is not simply
that these filtering techniques correspond to actual truths I mean this is not your future self in the top right hand corner for instance, but it's a projection, it's an inhabitation of a virtual, it's an inhabitation of a selected set of probabilities which according to a series of axioms may be projected on your existing, the way your face reflects light under certain circumstances. These filters have also generated numerous controversies.
This is an example of one in which FaceApp was within its beauty filter, or its hot filter rather, was whitening faces, producing faces in this case with a nose that's made more slender. So, you know, changing faces, recognizing the face as an artifact and then modifying that face to conform to certain geometrically described proportions that correspond to specifically cultured ideas of a certain kind of hotness or beauty or handsomeness and so on. something that's immediately picked up by users of the software and recirculated in social media
to create a further articulation of it. One of the early cases that created an interesting scandal in China was a case of a vlogger who was working on beauty products, an influencer selling beauty products, makeup. And she had commissioned a custom filter for her vlog. And she looked like the person on the left with the blue icon. The software running the filter broke for a moment and it glitched.
And her actual face, which is the one in the center, was revealed. So this filter was, as you can see by comparing the two faces, was very strong. Let's put it that way. and this vlogger uh it became a kind of national scandal or a course celeb and in fact this vlogger came back uh even stronger afterwards uh with an even stronger filter that you can see on the right uh so this this the audience for for beauty products uh in this case actually accepted uh the filter as part of the the entertainment service as part of the marketing service uh that this blogger was providing.
So it's interesting to see that the filter is not necessarily seen as something untrue or dishonest, but it's seen as part of an entertainment or a set of media effects that are interesting to play with. Filter also moves out into the real world. Many of you'd be familiar with this phenomena, Instagram face, which describes a form of makeup that attempts to replicate the way in which things like Samsung beauty face or the beauty filter in Instagram produce or rework the
faces of users. So in this case, a reporter for the BBC followed a series of internet or YouTube tutorials to see if she could apply makeup in a way that would allow her face to look as if in real life she was working, she had the Instagram filter on her face. So not only does it act as a filter between phone users, between Instagram users, it also starts to move out. The conventions of beauty that is described by these filters start to move out into three-dimensional spaces, into other parts of the world. This routine of preparing her face took about between 30 minutes to an hour.
so it's a bit slower than using a filter on your smartphone. So you'd have to work this into your daily routine and it's a big commitment. But as you can see, it has quite strong effects. So what I want to move to from the movement of filters from logical description in Wittgenstein to artificial intelligence and the various stages of artificial intelligence from the perceptron through to deep learning through to convolutional neural networks and then their implementation in smartphone apps such as Instagram or FaceApp.
So as always when you show a PowerPoint you always find that you've done one spelling mistake and you only notice it at the point you present it. So you know, luckily this is the slide in which the description of information theory as a description of communication is also applied to the description of the brain. So you can see that mine is non-optimally filtering this slide. So some recent work over this, in this century really, in cognitive science,
starts to describe the brain itself as a filtering system. So we go from the filter as a description of logical systems, from the description to the description of filters as mechanisms for shaking culture and participating in culture, to the description of filters as cognition. So in Carl Friston's work and a range of people such as Andy Clark, Emil Seth and others in cognitive science that are looking at the way the brain interacts with sensory input. words. What they what they argue is that the brain and the body connected, obviously, interacts
with the environment and interacts with the environment in a way that is predicted. So it interacts in a way that assumes that certain kinds of responses will come at it from the the stimulus that it produces. So the brain is constantly filtering interaction with in their models. The brain is constantly interacting with the world in order to infer meaning, in order to gain a response from stimulus. And the brain generates this feedback loop between an inference, what happens in the world, an inference of what it means or how the brain recognizes what's happening, how it learns what's happening, and then feeds that back
into further processes that allow the brain to then work with minimal effort in calculating what next to do. So the model of the filter starts to move into the model of, into the understanding of the systematic interaction of humans and other animals with the spaces they're in, the social systems they're in, the symbolic orders that they're in, and it implies the filter as a general system which allows for the understanding of all these processes. Additionally, with the additional qualifier, the brain itself is something that is recognized through systems of filtering.
So a lot of what these cognitive scientists will do will be to put someone in an fMRI machine and feed them stimulus, which may be sound and maybe images, and then to use the fMRI machine to map the response of the brain to that stimulus. fMRI of course is a system that is based around around the basics of information theory and it uses some of the same the same mathematics that Fourier proposed in the early 19th century to describe thermodynamic activity to describe electrical and energetic activity in the brain so we have the kind of
feedback loop between models of the brain, the understanding of the brain, and the way the brain is modeled through the filter, through the idea of the filters. So the filter becomes the kind of ruling metaphor for understanding cognition, but also information transmission of all kinds. So what the brain does, according to Friston and others, in terms of the so-called free energy principle is that it attempts to minimize the energy it needs to expend on calculation so when you're walking for instance or you're reading a text you're trying to calculate your brain is calculating how much variance this this stimulus
is has from what you're what you're normally going to expect what you're normally expecting from the act of walking or from a text. For certain kinds of behavior you want more variance, from other kinds of behavior you want things to be highly predictable. So this culture in a sense is wrapped up in this envelope of cognitive difficulty or cognitive, the energy expenditure based around the cognition or cognitive processing of the stimulus that is fed to the brain. So if we think that the brain as a filtering system in this model, or is understood metaphorically as a filtering system, it goes some way to understanding the way in which filters
expand out into the world as a wider system of formation, not just of language, of meaning, of interaction between people, symbolic elements, but also of the way in which they're received and enacted. To loop back then to the question of gender that was raised by the Snapchat, sorry, the Instagram filters and face app. There's a strong correlation between the model of information theory and the cognitive science use of information theory to describe iterative process of learning
and what many working in gender studies, and in this case the artist Vajinal Davis, who who works in trans and drag performance to reenact the traumatic formation of gender and the repertoire of actions and the repertoire of predictive or predictable actions that structures produce, is in a sense something that reads quite strongly in relationship to this quote from Judith Butler's bodies that matter. So a key argument in Butler's
gender studies or gender theory is this idea of iterability, that we're not born with gender, that we learn simply to iterate it, that is to repeat it until it becomes a learned set of behaviours, a learned set of features. And she uses a vocabulary that is very close in a way, but entirely distinct from the work of the cognitive scientists that I mentioned earlier, and also of machine learning, when she talks about a regularised and constrained repetition of norms becoming the basis for the formation of the gendered subject. So repetition is what produces the training that sets gender in place.
And repetition is also in machine learning what enables that learning to take place. So there's a strong convergence of vocabulary and conceptual structure that is interesting. And we can see that the subject that Butler is talking about can be seen as a form of performing filtering system. You know, so that this iteration of an act or a process of regularized and constrained repetition is what produces us as subjects, what produces us as gendered subjects. This process in cognitive science is what enables the brain to come about as a learning mechanism or a learning organ.
And in machine learning, it's a description of the way in which neural networks learn how to read and predict the world. So there's this interesting convergence, which I'm trying to map in this, between the formations of cultural norms, for instance, gender, and the technical tools in computing, information science, and machine learning in the filters that we use to communicate on our apps, and which govern our interactions through social media. but also in the way in which these are understood by discrete disciplines, by computer science, information science, artificial intelligence and gender studies in these cases.
What this process of interaction between different systems of learning, different forms of filtering and different iterative processes produce is in Butler's work, particularly this idea of the subject that emerges through the interaction of different systems of repetition. And these different systems of repetition can be described by patterns of interference, moiré patterns, between different overlapping systems of patterning. Here we have an example, and I'm coming to the last few slides now.
A glitch produced by a camera interacting with software as it flies along the River Thames in London to describe the riverbank. bank you'll see the difference portions of the image are more or less photorealistic others uh the sample rates um mesh messes with the rate at which the information is coming in there is uh a lag of the information at certain points the information being more complex at certain points in the image takes more time to process the filter slows down and you get these these very interesting
glitch effects where the system of the sample, that is the rate at which it takes a quanta from the world outside of the camera, the way it's then composed by the filters within within the computer on board the drone and on board the camera, there is a lag in those processes So the same image is taken over and over again and composited on top of a prior version of that image, which then creates this very kind of interesting uncanny layered effect. There you have filtering systems overlapping and producing kind of chaotic visual effects.
But they also have very interesting correlations to other forms, other patterns of interference, for instance, in geology, sedimentary rocks in this case. but you could also think of other forms of geological layering and interference in igneous rocks like granite or the kind of marble that we see on the floor, in which different chemical thermodynamic effects produce differently textured material formations. So there's a correlation that I'm arguing for between the way in which matter filters other matter and the way in which information processes filter other information processes.
Close, I want to kind of argue that this argument in some ways is quite an old argument. To make a comparison to a book that some of you are probably familiar with, John Epstein's The Intelligence of a Machine from mid-20th century, a film director that looked at that the apparatus of film, and for instance, taking an image at 24 frames a second, the sample rate of cinema, and described it as a system, as an intelligent system. So it's interesting to compare Epstein's arguments
about cinema as an apparatus, and indeed his work is the basic founding text of apparatus theory that later influenced, for instance, Deleuze's writing in cinema one and two, Foucault's writing on a dispositif, and other work that then takes the idea of the way in which machines, systems of sampling, systems of machinic operation, and production of images of texts and so on, produce certain kinds of systems of enunciation. Epstein relates these to the cinema and argues that the cinema
itself is a filtering system that produces a certain kind of knowledge about the world and that itself produces a space of interaction between different world systems, whether they're cities, economies, cosmic forces, metaphysical or objects that are generally described by metaphysics or by the then emerging theories of relativity in physics. So here what the kind of echo or the resonance with Epstein that I'm proposing is that these objects like filters or these technologies such as filters are deeply cultural, that they have a fundamentally
a fundamental relation to our systems of thought and how we also then reflect on our systems of thought but also have strong correlation to some of the kind of critical philosophical and theoretical work that allows us to then engage with these, engage with these materials. So the filter becomes the mechanism by which we also understand the operation of filters in the world, and they're able to critically reflect on them. So that means something, I'm not sure what it means, but it means that we perhaps need to develop a critical vocabulary of filters more broadly, which is what I've hoped to point a little towards in this talk.
So thanks for your patience. So, the filter on this slide has the two input now. Then we run the mode, some counselors present the counselors. Ever since we know that in the human prevention,
not as a problem with the filter because I'm honored to be here to welcome our city and our families, supporting the stages of uh I didn't have the top-com sort of . I guess the
yeah i mean that is interesting i think this uh and I can't should know, is part of the most reality while it's a new way to counter-politics and all the reasons.
Yeah, I guess I think it's an interesting argument. I think in the material I'm looking at, I guess what I would argue is that the culture is much more polycentric than the kind of high culture, low culture model. And this, you know, for instance, singers like that I mentioned, like T-Pain or one of the people that appears on one of the slides, Ramal Zee, who's a hip-hop artist from the 70s and 80s, who's an early use of vocoder in his work. I think that they were able to explore very intense areas of cultural creativity,
of production of novel forms of voice, and to understand the voice and to work with the voice as a technical artifact, to understand the human voice as something technically produced in a way that was outside of the capabilities of so-called high art, for instance. No one was doing that. So I think that I would tend to look for a much more, and I think this is a kind of classic cultural studies move anyway. It's not especially something I can claim an originality for by any means. that culture is much more polycentric and that, you know, grounds for experimentation and rigorous experimentation can be found, you know, in many different kinds of places, including things that just don't fit in the high culture, low culture model.
And social media, as you say, is part of that, but is also a recognition of culture being formed by multiple myriad kinds of forces and not simply mappable onto a high or low continuum in a way. So more multidimensional. Thank you. Yeah, we are now in the example of a subculture, like the cultural culture, which was originally in the UK, especially in the UK, about a different time, which is a non-authentic culture, which
You said it very well, I have to say. It sounds great. There's a question from the school room from Paul, Ohio. It says, you mentioned a convergence of vocabulary. Do you think any emergent discoveries may lead to divergences in language, in brackets, industry versus theory? And thank you for this very informative lecture and important reading, John.
Yeah, I think I'd be cautious in using the word disruptive, I guess, in that, you know, is a kind of language derived from economic thinking and that already I guess clues you into how I would respond to your question. But I think like the high culture, low culture distinction. There is also this subsumption of the language of avant-garde, for instance, such as disruption in economic vocabularies. So there are other kinds of appropriations and
convergences that we have to play with in this kind of multidimensional or hyperdimensional space. it would be interesting to think okay what where do where do particular kinds of break come about and I think the the break will be will come about partly in terms of what kinds of values are attributable to technologies you know if we think of the the mainstream culture of computing Silicon Valley, if we think of it as equivalent to Hollywood. Silicon Valley is the Hollywood of computing and we need other kinds of computing, other
kinds of way of thinking technology that corresponds to more artistic, to different kinds of cultural imperative, to more experimental forms. One of the breaks will be around what kinds of values are attributed to technical systems. So whereas I'm arguing that they have implicit cultural, political, economic and metaphysical consequences, Silicon Valley approaches would say, no, they're simply about enabling people about making the world a better place. So they try and simplify the number of factors which can be taken into account.
So there you could say there's a kind of a filtering of vocabulary or filtering of a filtering of the register of address at which a technology is said to be operative. So the terms by which something has to be taken into account or not. So, for instance, in much computer science work at present, there is very little account taken of energy consumption. The energy consumption of particular algorithms, particular kind of computing types, the way in which these in turn produce different kinds of loads on the climate.
so they're productive of different kinds of effects in terms of climate damage or of carbon production. This in computing is usually compartmentalized so that these things should not be taken into account. And there are, you know, normally the work is seen as very modular. There are people that design algorithms, there are people that design operating systems, there are people that design the hardware, there are people that design the energy, and they're the ones only that are responsible for energy consumption. An argument would be, and this is one made by a tendency in computing called permacomputing that takes the values of permaculture, gardening and agriculture into computing that says all aspects have to be
taken into account at every stage, and so they have a much more multi-dimensional approach to the problem to say that every every act of computing um is integrally connected to its interaction with the environment and its production of uh carbon for instance so that i guess that's where the break i see and this kind of multi this break would be around what constitutes the mode by which you're able to evaluate something so filtering is about uh cutting down what one has to take into account in order to save energy uh a critical theoretical response is about opening it up and saying actually you have to take more things into account you have to be abundant with your
Thank you. I think we can imagine this for sure. I think one of the key documents that attempts to imagine such a system of communication is Wittgenstein's Tractis Logico Philosophicus, which, you know, in this book, that's exactly what he tries to describe, is a system of perfect communication.
And there are many precursors to this, of course, in the history of logic or the history of philosophy. But what Wittgenstein arrives at ultimately is that this system fails. That there is an impossibility there that's foundational to language, that's foundational to the idea of communication. and um you know he then moves on to a totally different understanding of language which is based around language as it's used in everyday practice so it becomes much more anthropological simply kind of observing what people do and recognizing the the gaps the difficulties the problems uh in that but but he also brings this very intense kind of mystical uh relationship
to the idea of communication which is I think partly what you're talking about and it's this there is this kind of yearning for an unfiltered access to truth and a self the constitution of a self and a thinking self that would be able to gain access to the truth of itself and therefore be able to communicate to others uh the consistency of that truth and this you know this thing that tortured big in time his entire life. You know, you can see it in many aspects of modernism. For instance, there's this kind of torture. But there's also this, there's another sense in which the impossibility of that condition
is also something that drives culture on, drives politics on, drives this kind of utopic idea, which is impossible, that creates a kind of attractor within culture that pulls people's action towards it. And it becomes kind of interesting, at least, to experiment with the ways we can fail in communication. Thanks. Thank you very much for the thanks for the content of the
Thank you very much. I'd like to let you here also to the university in South Africa, and then as, should I say, useful to go to the energies and creating some benefits for the society, keeping the environment as well. computer in a way of easing our life and filtering information and coming in the, should I say, MPM, practical tools, recognition, recognising cases and all of this. And this is such a great, should I say, purpose of filtering, to help the society and people
having better communication, perfect communication, I don't think so, because perfect communication is something very difficult to define. But when I was invited by my friend Catalina for this lecture yesterday, I was really triggered by the very title, and I started to feel what I triggered for the filtering. And it's exactly what you came with the example of social media and practically all of this filtering of the bases. And this is the phenomenon where I'm seeing that filtering becoming a kind of running away from
ourselves. Filtering ourselves in something which we are not. And what is strikingly with the Chinese lady who was, what was, she was, invoicing and sending products. At the end of the day, this fight was disclosed that this filtering was practically untruthfulness and there's any bias, people accepted. And this is where I'm seeing in society the filtering query very dangerous developments in the society, especially among youngsters, because we are
having more artificial creatures with the filtering rather than coming to ourselves. What do you think about that? Yeah, no, I think it's an interesting it's an interesting kind of poll in the debate, one would say. I think I think there's a lot of debate around filters and so-called filter bubbles or echo chambers, the fragmentation of communication, I mean, and the use of filters by political agencies or commercial agencies, the debate around the neutrality of search engines, for instance. I think that's the question of turning away from oneself.
Is it one way that we have to consider what that means? What this turning away from oneself means? but in culture and you know turning away from oneself is is also the kind of fundamental act of thinking in a way of trying to move ahead from what it is that you've established yourself as so to produce a condition in which one becomes other than oneself even as a kind of experimental conjecture. And I think it's not so much the turning away from oneself or turning away from ourselves that is problematic.
I think, you know, that kind of the alienation from oneself is absolutely fundamental to being able to think. But the way in which then we create that as an experimental space is what's important, not that we use filters as a way of closing down what we are. So we close down our access to information from others. We close down our access to other forms of experience, which is the problem with the filter and things like the echo chamber and the filter bubble. but we see the filter as something that is also part of this alienation.
So, you know, so for instance, the singers that I mentioned, like Ramosy or rappers that I mentioned, like Ramosy and the singer T-Bain are often attacked by others for having artificial voices, of sounding non-real or sounding inhuman. but what I think they're doing is adding something that's previously unheard that we can experiment with and to see what it feels like to feel the human voice in a different in a different register to feel it as something synthetic and I think that's that's a challenge from a notion of culture that wants to see humans as relatively stable
and known, that are known to itself and that are known to themselves, and an opening to something that is more unstable, more open to question, more of a hypothesis. I think we need to maintain ourselves as a hypothesis rather than as a known quantity and becoming alienated is a good, I mean, who wouldn't be alienated in the 21st century, right? So, okay, thanks. I just wanted to ask to check when you met you, if the mesh of the grid, any grid yesterday
we're talking about the same thing as the must stop in German. The Greek thing? Yeah. And it's skills and English. Isn't that the same thing? I don't know. Well, I read both the German and the English parallel. So I believe that it was translated as scales. There, the master board scales tend to actually be the perfect the omissioness of the master board scale,
the whole maybe, to be the perfect replica of the reality replica of the reality that is supposed to become a via language, through language, and he operates with the word image. So this replica, whatever object in reality or a situation where it's supposed to be observed So in the history we have this contradiction there, because there is this desire to
immediately through through this whole space out of the view of observation transformation into a message and through uh language and the other half yes he affirms that whatever is conveyed as true must go through the certain shelter which is must have or So how would you comment this? Do you also observe this, you know, this diet of slaughter meal, like the
of the kind of the immediacy of the missile, the part of the immediate issue with the structure on the one hand, and on the other hand, elaborate parts of the treatise of the spectra, where he also explains how it's inevitably, in fact, filtered in your terminology from the immediate image that is supposed. So do you also discern this, you know, like a contradiction in his initial premise? Yeah, no, I think that's right. I think, you know, you describe it very well.
Well, there is absolutely a contradiction. And, you know, he's kind of circling around this impossibility throughout the book in different registers and using different vocabularies, different analogies, different logical languages, different even mystical languages at times. And I think that's really interesting as a text because it attempts to be as clear as possible, but yet it recognizes its own failure. But it's kind of imminent, but by becoming imminent to its own failure, maybe also realizes something of what it attempts to set out.
So I think it retains a high degree of precision, but also poetic value at the same time. So, yeah, I think it's very resonant, I think, for that reason in some ways. But I think, yeah, the way you, I think you described the problem very well. yeah it says uh thanks um very interesting lecture uh can you elaborate on how in your experience observation the criteria in selecting the reduction slash augmentation that filtering does tend to be chosen.
Abundant or minimal approach in modeling, what could give adequate orientation in your opinion? Or, to pick up your last comment, how could one even know in an epistemological sense if one is failing in communication or succeeding? Thanks. Yeah, I mean, it's an extremely good question. I think it would really depend on the specificities of the filter and the setup that you're analyzing, rather than filters in general. In terms of machine learning, for instance, you would look at whether a neural network is trained or untrained.
That would give some answer to your qualification, or some answer to your question. Then I think there are key questions in relationship to these highly kind of cultured definitions of beauty that are visible in the Instagram face. to understand where they come from, to understand the kind of racialization that's embedded within them, in the conformation of faces to certain particular modalities of beauty that are very oriented around white or characteristically white or Caucasian features,
then I think you could look at the way in which they reinforce certain technologies of gendering within them and understand them on that basis. Within neural networks, there's a whole subfield called explainable AI, which, you know, under which bases itself on the idea that it would be possible to unpick the actions and decisions of a network in order to to understand its specific or to be able to explain its specific decision. This creates a number of problems in that many of these systems are working across myriad sources of data simultaneously.
It's very difficult to replicate the conditions under which they would make certain kinds of decision. decision. There are other systems being developed which allow for the state of a neural network to be recorded at every moment so that a step-by-step process can be used to analyze exactly the point at which a certain value was present in a certain node of the network and therefore and pick the specific way a decision was reached. So once you have a step-by-step decision, a set-by-step map of all the decisions that were made,
one can then show the way in which a particular flow of weightings or a particular sequence of weightings unfolded, and that might allow for a certain understanding of the criteria that were emerging in the network at that particular moment. I don't know if that helps. Okay. We can conclude the session. Thank you. Thanks.