Difference between revisions of "Chihuahua or Muffin?"
(→Misc) |
|||
| (29 intermediate revisions by the same user not shown) | |||
| Line 22: | Line 22: | ||
What is the current state of AI? How do these technologies work? How does a neural network learn from data? And most importantly, can these systems distinguish chihuahuas from muffins? | What is the current state of AI? How do these technologies work? How does a neural network learn from data? And most importantly, can these systems distinguish chihuahuas from muffins? | ||
| − | == Learning | + | == Learning goals == |
*Demystify and achieve a basic understanding of a series of terms in the domain of this course, including: AI, bots, machine learning, neural networks, singularity and computer vision. | *Demystify and achieve a basic understanding of a series of terms in the domain of this course, including: AI, bots, machine learning, neural networks, singularity and computer vision. | ||
*Have an overview of the possibilities that these techniques offer for artistic creation. | *Have an overview of the possibilities that these techniques offer for artistic creation. | ||
| Line 83: | Line 83: | ||
GPT-2: [[GPT-2 | GPT-2 Language model]]<br> | GPT-2: [[GPT-2 | GPT-2 Language model]]<br> | ||
| − | == | + | == Recommended readings == |
| + | [https://en.wikipedia.org/wiki/Weapons_of_Math_Destruction Cathy O'Neil - Weapons of Math Destruction (2016)]<br> | ||
| + | [https://mediarep.org/bitstream/handle/doc/13258/Pattern_Discrimination_23-58_Cramer_Crapularity_Hermeneutics_.pdf?sequence=3 Florian Cramer - Crapularity Hermeneutics (2018)]<br> | ||
| + | [https://strelka.com/en/press/books/lev-manovich-ai-aesthetics Lev Manovich - AI Aesthetics (2018)]<br> | ||
| + | [https://nyupress.org/9781479837243/algorithms-of-oppression/ Safiya Umoja Noble - Algorithms of Oppression. How Search Engines Reinforce Racism (2018)]<br> | ||
| + | [https://www.e-flux.com/journal/72/60480/a-sea-of-data-apophenia-and-pattern-mis-recognition/ Hito Steyerl - A Sea of Data: Apophenia and Pattern (Mis-)Recognition (2016)]<br> | ||
| + | [https://www.excavating.ai/ Kate Crawford and Trevor Paglen - Excavating AI. The Politics of Images in Machine Learning Training Sets (2019)]<br> | ||
| + | |||
| + | == Inspiration == | ||
| + | [http://geertmul.nl/projects/match_of_the_day/ Geert Mul - Match of the Day (2004 - ongoing)]<br> | ||
| + | [http://www.sester.net/access/ Marie Sester - Access (2003)]<br> | ||
| + | [https://nos.nl/artikel/2256421-schilderij-geveild-van-min-g-max-d-ex-log-d-x-ez-log-1-d-g-z.html Obvious - Edmond de Belamy (2018) -> Painting generated with a GAN sold for 375000€]<br> | ||
| + | [http://www.helenknowles.com/index.php/work/the_trial_of_superdebthunterbot Helen Knowles - The trial of Superdepthunterbot (2016)]<br> | ||
| + | [https://www.youtube.com/watch?v=LY7x2Ihqjmc Oscar Sharp & Ross Goowin - Sunspring (2016)]<br> | ||
| + | [https://cvdazzle.com/ Adam Harvey - CV Dazzle (2010 - 2017)]<br> | ||
| + | [https://ahprojects.com/hyperface/ Adam Harvey - Hyperface (2017)]<br> | ||
| + | [http://imagenet.xyz/euronet/ Constant Dullaart & Adam Harvey - Euronet (2017)]<br> | ||
| + | [http://scottandbenorbenandscott.com/#/signs-of-the-times/ Scott Kelly & Ben Polkinghorne - Signs of the Times (2017)]<br> | ||
| + | [https://vimeo.com/298000366 Mario Klingemann - Memories of Passersby I (2018)]<br> | ||
| + | [https://www.thispersondoesnotexist.com/ This human does not exist (2019)]<br> | ||
| + | [https://thisfootdoesnotexist.com/ Mschf - This foot does not exist (2019)]<br> | ||
| + | [https://www.youtube.com/watch?v=NuGT_L13bQ8 Andrew Bujalski - Computer Chess (2013) - Trailer]<br> | ||
| + | [https://www.youtube.com/watch?v=8tq1C8spV_g Greg Kohs - AlphaGo (2017) - Trailer]<br> | ||
| + | |||
| + | == Other resources == | ||
=== Image processing / Object recognition === | === Image processing / Object recognition === | ||
| Line 89: | Line 113: | ||
[http://www.nytimes.com/2012/06/26/technology/in-a-big-network-of-computers-evidence-of-machine-learning.html How Many Computers to Identify a Cat? 16,000]<br> | [http://www.nytimes.com/2012/06/26/technology/in-a-big-network-of-computers-evidence-of-machine-learning.html How Many Computers to Identify a Cat? 16,000]<br> | ||
[https://experiments.withgoogle.com/what-neural-nets-see what neural nets see]<br> | [https://experiments.withgoogle.com/what-neural-nets-see what neural nets see]<br> | ||
| − | |||
[http://www.labsix.org/physical-objects-that-fool-neural-nets/ Fooling Neural Networks in the Physical World with 3D Adversarial Objects]<br> | [http://www.labsix.org/physical-objects-that-fool-neural-nets/ Fooling Neural Networks in the Physical World with 3D Adversarial Objects]<br> | ||
[http://ssbkyh.com/works/cloud_face/ Cloud Face]<br> | [http://ssbkyh.com/works/cloud_face/ Cloud Face]<br> | ||
| − | |||
[https://github.com/CMU-Perceptual-Computing-Lab/openpose OpenPose]<br> | [https://github.com/CMU-Perceptual-Computing-Lab/openpose OpenPose]<br> | ||
[http://densepose.org/ DensePose]<br> | [http://densepose.org/ DensePose]<br> | ||
| + | |||
| + | === DeepDream & Style Transfer === | ||
| + | [https://www.garysnotebook.com/20190826_1 DeepDream explained]<br> | ||
| + | [https://www.vice.com/en_us/article/53dwg3/heres-how-google-deep-dream-generates-those-trippy-images DeepDream: How does it work]<br> | ||
| + | [https://towardsdatascience.com/a-brief-introduction-to-neural-style-transfer-d05d0403901d Neural Style Transfer explained] | ||
| + | |||
| + | === Reinforcement Learning === | ||
| + | [https://www.youtube.com/watch?v=gn4nRCC9TwQ DeepMind - Producing flexible behaviours in simulated environments (2017)]<br> | ||
| + | [https://www.youtube.com/watch?v=kopoLzvh5jY DeepMind - Multi-Agent Hide and Seek (2019)]<br> | ||
| + | [https://www.youtube.com/user/BostonDynamics Boston Dynamics]<br> | ||
=== GANs === | === GANs === | ||
[https://medium.com/syncedreview/biggan-a-new-state-of-the-art-in-image-synthesis-cf2ec5694024 biggan]<br> | [https://medium.com/syncedreview/biggan-a-new-state-of-the-art-in-image-synthesis-cf2ec5694024 biggan]<br> | ||
[https://tfhub.dev/s?q=biggan biggan implementations]<br> | [https://tfhub.dev/s?q=biggan biggan implementations]<br> | ||
| − | |||
[https://github.com/tkarras/progressive_growing_of_gans progressive growing of GAN's]<br> | [https://github.com/tkarras/progressive_growing_of_gans progressive growing of GAN's]<br> | ||
| + | [https://github.com/NVlabs/stylegan2 StyleGAN2]<br> | ||
=== Conditional GANs === | === Conditional GANs === | ||
| Line 113: | Line 145: | ||
[https://www.theguardian.com/technology/2017/sep/07/new-artificial-intelligence-can-tell-whether-youre-gay-or-straight-from-a-photograp New AI can guess whether you're gay or straight from a photograph]<br> | [https://www.theguardian.com/technology/2017/sep/07/new-artificial-intelligence-can-tell-whether-youre-gay-or-straight-from-a-photograp New AI can guess whether you're gay or straight from a photograph]<br> | ||
[https://openeth.com/ OpenEth Computable Ethics]<br> | [https://openeth.com/ OpenEth Computable Ethics]<br> | ||
| − | |||
[https://www.technologyreview.com/s/542626/why-self-driving-cars-must-be-programmed-to-kill/ Why Self-Driving Cars Must Be Programmed to Kill]<br> | [https://www.technologyreview.com/s/542626/why-self-driving-cars-must-be-programmed-to-kill/ Why Self-Driving Cars Must Be Programmed to Kill]<br> | ||
[https://www.theverge.com/2017/4/13/15287678/machine-learning-language-processing-artificial-intelligence-race-gender-bias AI picks up racial and gender biases when learning from what humans write: There is no objectivity]<br> | [https://www.theverge.com/2017/4/13/15287678/machine-learning-language-processing-artificial-intelligence-race-gender-bias AI picks up racial and gender biases when learning from what humans write: There is no objectivity]<br> | ||
| + | [https://www.youtube.com/watch?v=fMym_BKWQzk Kate Crawford - The Trouble with Bias (2017)]<br> | ||
=== Natural Language Processing + Speech Recognition === | === Natural Language Processing + Speech Recognition === | ||
[http://www.algolit.net/index.php/Algoliterary_Bibliography Algoliterary Bibliography]<br> | [http://www.algolit.net/index.php/Algoliterary_Bibliography Algoliterary Bibliography]<br> | ||
| − | |||
[https://owow.agency/bots/ owow bots]<br> | [https://owow.agency/bots/ owow bots]<br> | ||
[https://research.googleblog.com/2012/08/speech-recognition-and-deep-learning.html Speech Recognition and Deep Learning]<br> | [https://research.googleblog.com/2012/08/speech-recognition-and-deep-learning.html Speech Recognition and Deep Learning]<br> | ||
| Line 126: | Line 157: | ||
=== Online Courses === | === Online Courses === | ||
[http://ml4a.github.io/classes/itp-F18/?fbclid=IwAR0NxhgFYyqb-FbKAdwdGMvxPxyoGzEkOoNOp7Sv6jFjBHFw11ulNPxQ0AY#syllabus Machine Learning for Artists]<br> | [http://ml4a.github.io/classes/itp-F18/?fbclid=IwAR0NxhgFYyqb-FbKAdwdGMvxPxyoGzEkOoNOp7Sv6jFjBHFw11ulNPxQ0AY#syllabus Machine Learning for Artists]<br> | ||
| − | [https://www.coursera.org/specializations/deep-learning coursera deep learning classes] | + | [https://www.coursera.org/specializations/deep-learning coursera deep learning classes] |
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
=== Misc === | === Misc === | ||
| − | |||
[https://www.youtube.com/channel/UCbfYPyITQ-7l4upoX8nvctg 2 Minute Papers]<br> | [https://www.youtube.com/channel/UCbfYPyITQ-7l4upoX8nvctg 2 Minute Papers]<br> | ||
[https://experiments.withgoogle.com/collection/ai ai experiments]<br> | [https://experiments.withgoogle.com/collection/ai ai experiments]<br> | ||
Revision as of 16:00, 12 January 2020
Elective - Chihuahua or Muffin?
Tutors
Brigit Lichtenegger
Javier Lloret
Arjen Suijker
Special Guests
Florian Cramer
Lotte Klompenhouwer
Yves Gellie
Description
The recent developments in the field of Artificial Intelligence and Machine Learning are opening up new possibilities but also presenting new challenges and ethical dilemmas.
AI systems are now capable of synthesize realistic portraits, psychedelic imagery, harmonic melodies or intriguing narratives. These systems are also used to make predictions based on data in wide ranging fields. They guess the words we are about to type, the songs or tv shows we might like or the content of a picture. But they are also used to determine whether we would commit a crime or pay back a loan, often perpetuating human prejudices.
In this elective, students will be introduced to Artificial Intelligence and Machine Learning through a series of lectures, presentations, hands-on workshops and discussions.
What is the current state of AI? How do these technologies work? How does a neural network learn from data? And most importantly, can these systems distinguish chihuahuas from muffins?
Learning goals
- Demystify and achieve a basic understanding of a series of terms in the domain of this course, including: AI, bots, machine learning, neural networks, singularity and computer vision.
- Have an overview of the possibilities that these techniques offer for artistic creation.
- Reflect on these current technological developments and their application to other fields connected to scientific and technological innovation.
- Develop a critical mindset questioning the impact of these developments in our society.
- Get familiar with a broad range of machine learning open source tools and frameworks that have been extensively used in recent years.
Final assignment
- Work on a project in which you inspired by the current developments of AI and Machine Learning.
- Free Format
- The project presentations will take place on Friday (January 17).
- You will have 5 minutes to present your project + 5 minutes to answer some questions.
Planning
Week 1
Monday
AI and Machine Learning: Background history, philosophy & the Turing Test with Brigit Lichtenegger.
Tuesday
Theory Session on Neural Networks. Workshop Classification with Arjen Suijker.
Wednesday
Convolutional Neural Networks with Javier Lloret.
Thursday
Ethics + Datasets with Javier Lloret. With invited guest speakers Florian Cramer & Lotte Klompenhouwer.
Friday
Generative Models and Reinforcement Learning with Javier Lloret + Screening of the short film “The year of the robot”, with an introduction from the director, Yves Gellie.
Week 2
Monday
Natural Language Processing and Project proposal feedback with Brigit & Javier
Tuesday
Assignment time
Wednesday
Assignment time
Thursday
Assignment time
Friday
Presentations with Brigit, Arjen & Javier.
Tutorials
Image recognition with Google Teachable Machine
Image recognition with YOLO with Darknet
Deep Dreams: Deep Dream with Darknet
Terminal basic tutorial: https://maker.pro/linux/tutorial/basic-linux-commands-for-beginners
Generating images using GANs DCGAN
Pix2pix: Online demo of Pix2pix
GauGAN: Online demo of GauGAN
Training a Neural Network how to play Pong: Pong
Generating text using a "Recurrent Neural Network" RNNs with Darknet
GPT-2: GPT-2 Language model
Recommended readings
Cathy O'Neil - Weapons of Math Destruction (2016)
Florian Cramer - Crapularity Hermeneutics (2018)
Lev Manovich - AI Aesthetics (2018)
Safiya Umoja Noble - Algorithms of Oppression. How Search Engines Reinforce Racism (2018)
Hito Steyerl - A Sea of Data: Apophenia and Pattern (Mis-)Recognition (2016)
Kate Crawford and Trevor Paglen - Excavating AI. The Politics of Images in Machine Learning Training Sets (2019)
Inspiration
Geert Mul - Match of the Day (2004 - ongoing)
Marie Sester - Access (2003)
Obvious - Edmond de Belamy (2018) -> Painting generated with a GAN sold for 375000€
Helen Knowles - The trial of Superdepthunterbot (2016)
Oscar Sharp & Ross Goowin - Sunspring (2016)
Adam Harvey - CV Dazzle (2010 - 2017)
Adam Harvey - Hyperface (2017)
Constant Dullaart & Adam Harvey - Euronet (2017)
Scott Kelly & Ben Polkinghorne - Signs of the Times (2017)
Mario Klingemann - Memories of Passersby I (2018)
This human does not exist (2019)
Mschf - This foot does not exist (2019)
Andrew Bujalski - Computer Chess (2013) - Trailer
Greg Kohs - AlphaGo (2017) - Trailer
Other resources
Image processing / Object recognition
quickdraw with google
How Many Computers to Identify a Cat? 16,000
what neural nets see
Fooling Neural Networks in the Physical World with 3D Adversarial Objects
Cloud Face
OpenPose
DensePose
DeepDream & Style Transfer
DeepDream explained
DeepDream: How does it work
Neural Style Transfer explained
Reinforcement Learning
DeepMind - Producing flexible behaviours in simulated environments (2017)
DeepMind - Multi-Agent Hide and Seek (2019)
Boston Dynamics
GANs
biggan
biggan implementations
progressive growing of GAN's
StyleGAN2
Conditional GANs
Pix2pix
Vid2vid
CycleGAN
BicycleGAN
AttnGAN
StackGAN
AI & Ethics
New AI can guess whether you're gay or straight from a photograph
OpenEth Computable Ethics
Why Self-Driving Cars Must Be Programmed to Kill
AI picks up racial and gender biases when learning from what humans write: There is no objectivity
Kate Crawford - The Trouble with Bias (2017)
Natural Language Processing + Speech Recognition
Algoliterary Bibliography
owow bots
Speech Recognition and Deep Learning
neural network generated recipes
Online Courses
Machine Learning for Artists
coursera deep learning classes
Misc
2 Minute Papers
ai experiments
Deep Learning is not the AI future
Future of Life Institute
NPU chips in intelligent phones
The FRIEND-Browser
Googlism
mrpimpgoodgame.over