My presentation entitled 'AI, Creativity and Generative Art', presented at the annual symposium for AI students (CKI) at Utrecht University, Fri. June 16th, 2017
3. eelco den heijer
â˘Developer/ Architect/ Scrum
Master
â˘Part-time Researcher
â˘MSc Cognitive ArtiďŹcial Intelligence
(Utrecht)
â˘PhD at Computational Intelligence
Group, VU, with Guszti Eiben
(2009-2013)
â˘Dad, living together, Amsterdam
4.
5. What is Creativity?
⢠Margaret Boden - Creativity and Art: Three
Roads to Surprise
⢠Create something 1) new, and 2) of value
⢠Combinational - unusual combination of
usual ideas
⢠Exploration - like local search in concept
space
⢠Transformational - transforming a concept
in concept space, âchanging the rulesâ
14. â˘Algorithmic Art, Computer art
â˘Process
â˘Autonomy, Independence from
author, creator
â˘Randomness
Generative Art
15. â˘Process Art very old
â˘Sol Lewitt, Instruction Art
â˘Mozart, WĂźrfelspiel
â˘Computers, software, AI are
excellent tools for processes
Process
21. Randomness
â˘Necessary to create something
unexpected
â˘Will almost certainly lead to new
ideas
â˘Will potentially lead to bad ideas
â˘Stochastic vs. deterministic
â˘Edge of Chaos
22. Computer Art
â˘Started around 1965, Germany
â˘Primitive software, primitive
hardware
â˘No advanced graphics cards, no
5K screens
â˘Drawing Machines
23.
24.
25.
26.
27.
28.
29. Outline AI Art System
Generate Evaluate
âArtistâ âCriticâ
Feedback
*
30. How do we generate art?
â˘First, we need a process to
generate an image
â˘Next, we need a process to
evaluate them
31. Image generation
â˘From scratch; a model, math tools
(fractals), CA, live coding
(Processing)
â˘Existing image, manipulate it
â˘Existing images, blend them,
manipulate them
33. A-Life
â˘Based on principles from Biology
(vs. Psychology)
â˘Very simple, low-level processes
â˘E.g. Replication
â˘Lindemayer Systems (L-Systems)
â˘Cellular Automata
34. Outline AI Art System
Generate Evaluate
âArtistâ âCriticâ
Feedback
*
35. Image evaluation
â˘Evaluate what exactly? What are
you looking for?
â˘Beauty?
â˘Shock? Conceptual art?
â˘Manual or Automated?
â˘Computational Aesthetics
36. Computational Aesthetics
â˘Tries to compute beauty (but what is
beauty?)
â˘Focus on parts:
â˘Complexity
â˘Symmetry
â˘Colour harmony, contrast
45. Scalable Vector Graphics
â˘XML based vector ďŹle format
â˘Used as a genotype
â˘Custom Crossovers, Mutations
â˘More ďŹgurative than expression
based genotype
â˘Initalization is diďŹcult
46.
47. Glitch
â˘Niche computer art movement
â˘Manipulation of binary encoding of
images
â˘GIF, BMP, PNG, JPG
â˘Process: existing image, basic steps,
recipe=> ďŹlter
â˘Very basic byte array operations
â˘Beware of image death
52. Functions
â˘Create functions; starting point
â˘Unary: one input image, ďŹlters
â˘Binary: two or more input images; blend/
merge
â˘Higher order: iterators, non-linear
functions
53. Create/Get Image
â˘Using APIs; Yahoo, Google Images,
Pexels, Pixabay, Google Streetview,
Flickr, Tumblr
â˘Generative: line art, graďŹti
â˘Local ďŹle system
65. WorkďŹow ex. 2
Get Image
Get Image
Blend
Recolour
Mirror
Left/Right
Recolour
66.
67.
68.
69.
70.
71.
72.
73. Conclusions Arfunkel
â˘Combinational and Exploration
Creativity
â˘Parametric Design and Generative Art
â˘Will it ever âbreak the rulesâ?
â˘In theory: it might
â˘In practice; diďŹcult to âďŹndâ balance
novelty and quality
â˘REPL - Read Eval Print Loop
75. Whatâs next?
â˘The search space for generative art is
(too) large
â˘Fundamental question: what are you
looking for?
â˘In other words; when is an image âartâ?
â˘Automate the search process!
â˘Computational Aesthetics
76. Too bad, nobody wants itâŚ
â˘Itâs made by a computerâŚ
â˘It doesnât mean anything
â˘There is no authorâŚ
â˘You didnât go to art school
â˘Two Cultures (C.P. Snow)
82. Word pairing
⢠âThe quick brown fox jumps over the lazy dogâ
â˘âThe quickâ + âquick brownâ +
âbrown foxâ, etc
â˘Create index in memory
â˘Create phrases starting with last
word, using Markov chain index
â˘Create indices use diďŹerent corpora
==> remixing
83. 'bout 'bout it bout it is een sappie
ben dan hij komt vjezze oeh Jappie
boy vaderlay low
sappie opgelucht
plat ja ik heb ondoeltreffender
alle schoon prijs aan t beter zangers
rijk juweel in t hare kostgangers
in liefdes almacht
herdenkt de haardracht
gegeven hun hulde aan windvangers
Source: William Shakespeare (Sonnetten)
Source: Jeugd vanTegenwoordig
84. Formats & Functions
⢠DiďŹerent templates for poetry forms Limerick,
Haiku, SMS, Snowball, etc.
⢠Easily implemented in Python
⢠Small functions:
⢠Syllables (language speciďŹc)
⢠Groene Boekje (Dutch Word List)
⢠Anagram
⢠Rhyme
85. is de club huilen
lucht kwam ineens changde z'n
holler at ya boy
aan de stervling steeds
aan den gloed en zinneloos
de schuld en voorproeft
Source: William Shakespeare (Sonnetten)
Source: Jeugd vanTegenwoordig
86. Dichtfabriek: conclusions
⢠Still Proof of Concept
⢠Real fun to do, very short development time
⢠Generative Art, Combinational Creativity
⢠Python is excellent for Creative Coding
⢠Currently: Dutch only, 100% form, 0% Content
⢠Many points of extension for future work
87. â˘Processing, VVVV, Python, etc.
â˘Music: Chuck, SuperCollider, etc.
â˘Creative Coding Utrecht
â˘Creative Coding Aâdam
Get Started Today!