UCLA ART&ARC100-1 “Find By Image: Machine Learning For Artists” Spring 2017

    Tues and Thurs 2-4:50pm

    Office Hours:Tues 12-1pm Room 4256

    Overview:

    This studio-based course aims to introduce machine learning—a complex and quickly evolving field—to artists, designers and performers. The goal of this course will be to unpack and familiarize ourselves with available machine learning tools, which we will use to plan and produce works of art. In-class labs will open a preliminary investigation into the conceptual and technical underpinnings of key machine learning methods, exploring their application through hands-on demonstrations. Readings and discussions will attempt to connect the theory, practice, and poetics of machine learning, and to place our efforts into wider ethical, social, and art-historical contexts. In the process, we will expand on the general phenomena of learning, experience, and creativity as subject matters in themselves. Students are encouraged to pursue a process that continues the development of their own specific artistic practice. However, this course places a strong emphasis on collaboration—because of the interdisciplinary focus of the course, and because realizing artistically viable projects with machine learning will require a bundle of specialized knowledges.Whatever our background, we are each in experimental territory!

    Course Goals:

    As a group, we will be conducting original research with the aim of uncovering artistic and poetic possibilities in the practice of applied machine learning. Each student will learn the concepts behind common machine learning techniques, and apply these ideas to artistic projects of their own design.

    Expectations:

    Students will be expected to collaborate - to share skills and resources; to document their work and make it available to a wider public; to properly credit and attribute components of collaborative work; to keep up with course readings, to come to class with questions prepared, and to take turns leading in-class analysis of the readings; to actively participate in group discussion and critique.The form and media of each studio project will be left open to each student - ideally we will explore a range of approaches.

    Assignments:

    Assignment 1: Art and Technology

    • To be completed by each individual student and presented in a rapid “show and tell” style.
    • Find an example (either contemporary, or from anytime in the last 50k years or so) of an artwork made concurrently with the emergence of a technology. identify the technology being used. “artwork”, “technology”, and “use” are all terms we can think about expansively. examples of artworks that use machine learning are of course welcome, but not expected. the aim is come up with a common critical and aesthetic framework for talking about art and technology.

    Assignment 2: Everyday AI

    • To be completed by each individual student and presented in a mode of the student’s choosing for in-class critique (video / photo / text documentation, graphic representation, interactive demo, online, etc)
    • Identify a way that you are already engaging with AI in everyday life. Develop and document a process of exposing (breaking? interrupting? re-directing?) the limits of this AI.
    • In this process, consider the following questions:
      • Who made this? (a company, or a government, individuals, or many people, what gender, age, class, location, etc)?
      • What do you imagine the motivation for making this was? What outcome was expected?
      • What artifacts (intentional or otherwise) does it produce (visual, textual, sonic, affective, etc)?
      • Can you point to moments when the technology performs in unexpected ways?
      • Can you point to moments when the technology feels human? when it feels inhuman?
      • What have you given to this technology (data, time, attention, money, etc)?
      • What new questions does your interaction with this technology elicit?

    Assignment 3: White paper presentations

    • In-class discussions led by groups of three.

    White papers (or, variously, “whitepapers”) occupy an interesting literary territory between scientific publication and marketing tool. They serve to explain a specialized topic to a more general audience. A technique or concept may be explained, a problem identified, or a solution proposed. White papers may present experimental evidence that is designed to be independently repeated by readers, or it may only hint at or speculate on possibilities.

    For this assignment, each group of three will choose one paper to present collaboratively in class. The goal of your presentation should be to attempt a deeper understanding of new and interesting projects in the field of AI in general, and machine learning in specific. Your choice of paper should be based on your own curiosities, rather than any prior understanding or specialized knowledge.

    some possible questions to address in your presentation:

    • what is the “problem” being presented by the paper? this may be presented as an “opportunity” instead of a problem - read carefully :)
    • what methods, materials, and concepts are involved in solving this problem, or pursuing this opportunity? how open is access to these materials and methods?
    • assume the paper is a marketing tool - what is the product?
    • how does the paper place itself in context (what can you tell from its tone, references to prior work, footnotes, etc?)
    • are there other keywords or tags that you would apply to this paper to help other audiences find it?

    walk us through the paper in detail. bring in supplementary materials / images / references if it is helpful. this is an exercise in literacy-building from multiple angles.

    as for where to find papers, i’d like to focus on AI/machine learning/deep learning, but this can encompass all aspects from math to social science to ethics to infrastructure to design… it’s a broad field for sure. i’ve listed a few sources here. also, arXiv is a terrific repository of scientific papers - try searching for terms that interest you. ideally, we’ll start with newer work: if the paper is older than 6 months or so, try to determine if newer work has been published that updates or revises the original publication.

    Assignment 4: Studio Project

    FINAL PROJECT

    This is the primary project that each student will complete for the class. Each project should start with an essential question related the poetics of machine learning, and progress towards becoming an artwork that makes critical use of machine learning techniques. What form this artwork takes - an intervention, a toolkit, a model, an exploration, a generative result… etc - is up to you.

    • to be worked on in groups of three

    • to be presented in three iterations, each of which should be considered as a standalone artwork in its own right. we can look at examples of speculative, ad hoc, and contingent artworks that might give context for how this might work:

    PART ONE: proposal (due 5/2 and 5/4)

    considerations can include technical and critical considerations, without necessarily getting into the practicalities of “how” it will work just yet. to be clear - emphasis should be on a speculative generation of ideas, without worrying about “how” to implement these ideas just yet.

    at this stage, the artwork you present should take the form of a presentation, a brainstorm, a sketch or a maquette. at minimum, upload presentation materials (text, image, slideshows, videos, etc) to the course github as a blog post. supplementary (offline) materials are very welcome.

    include thoughts about:

    • the data that you will use
    • the form in which this input data will be used
    • the quality or type of learning that will be employed
    • the model of learning you are interested in using
    • the intended outcome (behavior, material, engagement, use, audience, distribution… ?)
    • how to evaluate success or failure of this specific project?

    PART TWO: proof of concept (due 5/16 and 5/18)

    • “proof of concept” means getting concrete with the ideas and curiosities you presented in the first stage.

    Expectations:

    • You have identified the core question or objective that your collaboration seeks to address. This core objective can be as abstract or poetic as you like, but it should be concise - a 1-2 sentence question: a what-if, a prompt, or a call to action.
    • You have identified what the role and workflow for each collaborator will be.
    • You have each begun to test out ways of forming a response to this objective, and in the process, generated a small body of work. Bring evidence of these tests (models, scripts, sketches, hacks, mock-ups, plans, maps, drawings, notes, recordings, etc). “Response” could be interpreted as carrying a question further and further, or trying to solve a problem, or just feeling out what’s possible - it situates the objective in actions. In this sense, failed attempts are more useful than plans yet to be enacted.
    • Bring questions for us! Think of this as time to test things on a very forgiving audience.
    • You have a general sense of the final form(s) your project will take - what will it look / sound / feel / smell / taste like? where will it be? how will it act? will it be material or immaterial? permanent or temporary? public or private? will it take time to experience, is it instantaneous, or does it emerge? basic questions, broad possibilities - good to set a path; experiments with form are also valuable to bring to this presentation!
    • as before, please (in addition to any other presentation materials you bring) make a post in the class repository

    PART THREE: final presentation (due 6/6 and 6/8)

    The final form these collaborative projects take is completely up to the groups. It can be one artwork made collaboratively, one prompt responded to in three different ways, three completely autonomous artworks that share a space, or two separate things with one intermediary, or a chain of processes (one person makes something for the next person to react to) - think expansively about what collaborating means here.

    Final presentation of projects for this class should include the following components:

    • an explanation video
      • screen-capture with voiceover, or documentation, demo, etc
      • uploaded to vimeo or youtube for easy sharing
      • if you’d prefer to record this live in-class, please let us know! we will make arrangements
    • any code that you’ve made use of, uploaded to a folder within the project_code directory in the class repository, with a README file that provides:
      • attribution for any code copied or modified from another source
      • any instructions for building or launching the project from the source code
      • any links to datasets or trained models too big to host on github
    • a post on the class blog that includes:
      • a written explanation of the project
      • links to the corresponding proposal and proof-of-concept posts
      • supporting documentation (images, sounds, video, etc)
      • footnotes for white papers or other texts, artworks, code repositories—any IP used or referred to in the making of your project

    final presentation of the projects will be evaluated based on:

    • a demonstrated consideration of / relationship to the project’s context
      • themes, concepts, problems or ideas inherent to machine learning / AI
      • research being done in the field of machine learning / AI
      • the history of art, i.e. other artworks, concepts and theories of art
    • legibility
      • clear articulation of intent, process
      • ability to form essential questions that drive your project
      • when obfuscation is used, it is used artfully, to add value and/or surprise
    • effort / work
      • evidence of collaboration
      • evidence of original thought
    • resourcefulness
      • leveraging skills or resources you already had to work on new problems
      • ability to search out solutions or workarounds
    • experimentation
      • risk, surprise, failure, variation, and unanswered questions can add value as well

    In-class critiques at each stage can help guide the group to refine their initial proposal and working process, and to make full use of available resources.

    Additional Course Materials and Readings

    Throughout the course we will be reading and discussing critical texts, poetry, and speculative fiction, as well as looking at artwork for contextual and comparative examples. Students will be expected to apply an understanding of these materials in project critiques. A reading list can be found here. A range of additional resources for learning more about machine learning - online courses, tutorials etc. as well as API’s, frameworks, and projects of interest are collected here, and we may make reference to them occasionally as needed.

    Grading:

    Grading will be determined as follows:

    Assignment 2: Everyday AI - (20%)

    Assignment 3: White paper presentations - (10%)

    Assignment 4: Studio Project (proposal) - (15%)

    Assignment 4: Studio Project (proof of concept) - (15%)

    Assignment 4: Studio Project (final presentation) - (20%)

    Participation - (20%)

    Participation can be summarized as “contributions to discussions and critique, sharing skills and resources, evident role within collaborative efforts.”

    Studio projects will be evaluated through in-class critique, based on their artistic value, which we will describe as

    1. The extent to which the artwork addresses its presentation within the context of fine art: its relation to historical antecedents and future possibilities, its critical function, its virtue as an object.
    2. Legibility: how meaning is articulated, both in terms of how you as an artist speak about the work, and in terms of how the work itself speaks.This doesn’t always have to be literal - Open-ended-ness, obscurity and ambiguity are qualities that can be legibly contained in an artwork.
    3. Commitment to work: as demonstrated by the amount of time and care put into making the work (speci cally) and developing an individual artistic voice through practice (generally).
    4. Economy of means: how well does the work make use of its available resources (material, skill, time, attention, and so on).
    5. The experiment, the risk: what new possibilities have been opened? what is at stake?

    Attendance:

    If you are going to be late or miss a class, please email before the start of class. Each absence or lateness (more than 15 minutes) without prior notice will drop your grade by a half-step (A to A-, for example).

    Schedule

    (subject to modification):

    Tues April 4

    Thurs April 6

    Tues April 11

    Thurs April 13

    Tues April 18

    Thurs April 20

    Tues April 25

    Thurs April 27

    Tues May 2

    • Assignment 4: Studio Project - preliminary presentations / critiques (groups 1-5)

    Thurs May 4

    Tues May 9

    • Lab: Sequence Models - Considering time and other kinds of order using RNN’s

    Thurs May 11

    Tues May 16

    • Assignment 4: Studio Project - proof of concept presentations (groups 1-5)

    Thurs May 18

    Tues May 23

    • Assignment 3: White Paper presentation - group 5
    • Assignment 3: White Paper presentation - group 6
    • Lab: Latent Space - peeking into hidden layers

    Thurs May 25

    Tues May 30

    Thurs June 1

    • Assignment 3: White Paper presentation - group 9
    • Lab: work on studio projects
    • Read: [Paulo Freire “Pedagogy of the Oppressed” (Chapter 3)](https://github.com/publicityreform/findbyimage/blob/master/readings/freire.pdf
    • Read: [Jacques Ranciere “The Ignorant Schoolmaster” (excerpt)]

    Tues June 6

    • Assignment 4: Studio Project - Final presentations / critiques (groups 1-5)

    Thurs June 8

    • Assignment 4: Studio Project - Final presentations / critiques (groups 6-9)