In her 2009 article “Looking at Shirley, the Ultimate Norm,” scholar Lorna Roth describes how film emulsion could have been designed with more sensitivity to. In reality, a large set of biased (gendered and even racist) protocols intervene in the processes of saving the face to memory. the continuum of yellow, brown, black and reddish skin tones. A choice that should be motivated by recognition of the need for an extended, dynamic range. After the development of colour film for cinema, Kodacolor (1928), and Kodachrome for still photography (1935), there seemed to be little motivation to acknowledge or cater to a market beyond white consumers.
It was actually only when chocolate production companies and wooden furniture manufacturers complained about the impossibilities they faced when trying to reproduce different shades of brown that Kodak’s chemists started changing the sensitivities of their film emulsions (the
coating on the film base which reacts with chemicals and light to produce an image) and to gradually extend the abilities of the film stock to include a greater dynamic range or ratio between the maximum and minimum measurable light intensities (white and black, respectively). Roth concludes her article with the statement that the habitual racism embedded
within colour reference cards did more than just influence major standard settings, such as the tone of hue, chroma, contrast, quantisation and lightness (luminance) values. To her, it is also responsible for the highly deficient renderings of non-Caucasian skin tones, which have
resulted in an ongoing need for compensatory practices.
(Rosa Menkman, Behind White Shadows)
in
--> Computer Grrrls Catalogue:https://hmkv.de/files/hmkv/ausstellungen/03_Archiv/2018/GRLS/05_Publikation/GRLS_Magazin_digital_mit_Cover.pdf
--> Menkman's project: https://beyondresolution.info/les-inconnues
https://www.youtube.com/watch?v=YeCaDVj9VVc
enter text here
dynamic label, like wikipedia?
1.https://haveibeentrained.com/
Check what's the data in the LAION5B dataset, used for training Stable Diffusion
2. Google Open Images Dataset
https://storage.googleapis.com/openimages/web/visualizer/index.html
3. Google Cloud Vision Image Classification
https://cloud.google.com/vision/docs/drag-and-drop
4. Quick Draw (google)
https://quickdraw.withgoogle.com/
5. Text-to-image generator:
https://www.craiyon.com/
6. Coco dataset
https://www.maskedcontexts.com/masks
https://www.maskedcontexts.com/
cultural-specific datasets could be more accurate and less discriminatory, cause they decide for their own cultural artifacts. But who decides in their culutre what's representative and what not?
Localized Narratives
(mouse over video image tagging)
https://google.github.io/localized-narratives/
Positive because people aren't incentivized to use it.
Negative because people are incentivized to get it somewhere else, other datasets that are equally or more biased.
At the end, they already affected the field so much that it doesn't matter if they closed their dataset
enter text here
enter text here
Negative: things are now happening behind close doors. But this is always a topic up for discussions.
enter text here
Generally a good idea, more data is always a good ideabut the data should still be checkt,but in the beginning the algorithms probably need training data following a clear template, so voluntary input might not be useable in the beginning
also, does the dataset need to be representative or could it also affect algorithims negatively
enter text here
enter text here
A dataset is only a selection and can't represent a fair and unbiased world view as the data itself is already loading with meaning and implications. For what reasons do these categories exist in the first place? Should there be such categories even? By giving name to an already unspecific and questionable idea it invites itself to abuse and further escalations.
I wonder how this would influence the pool of participants, regarding gender, age group, skin color etc.
Due to trust towards technology, big tech etc.
Due to media ability etc.
Could voluntary participants participate with malicious intentions? What kind of biases could the introduce into the system (loud minorities, whilst a majorty just spectates)
the dataset distribution of production(? how do i say). people in North America and Europe produce more data then most people in the developing countries doing the labeling work, as some data is considered more valueable than the rest, "dirty data"
negative++, cuz tons of applications has been bulit based on the previous model, and are being used in various social senarious. and now they just hide it up and pretend it nerver happended, rather then make it as an biased archieve for researchers and data scientists to investigate
Maybe. But it doesn't seem scalable as these dataset are qualitative only when there's an enormous amount of data. The bias in a crowd-sourced set could be also questionable, cause it all depends on who posts there.
No, datasets will always be subjective and biases. While it is great to have the permission of the subjects of the images, this voluntary participation would exclude people from the get-go, that do not have access and control to these tools, creating a new way of inbalances and distorted representation of the actual world.
enter text here
thoughts and ideas
4.
3. Would it be an idea that voluntary participate a dataset and generate image materiale?
2. Was it a positive or negative decision in 2019 to take the pictures from ImageNet?
-> why postitiv
->why negative
1. How do you think a dataset is not discriminatory?
What has to change, that the categories change?
-> the politic
->the society
->the thinking of every individual person
enter text here
enter text here
scale.ai calls it "human insight"
iRobot's outsourced data analysing company scale.ai's outsourced label workers leaked the still video images from roomba
The data center got more noisy after surviving World War II.
The data center always exceeds the minimum requirements by even providing marzipan smell.
The data center is a gym for fluffy water proof cats that provides multiple power supply resources such as hometrainers.
The data center receives and proceeds data in a physical bunker unknown to the public.
The data center was so unfamiliar to me so that I felt like I'm out of this planet.
Data center monitors precisely to provide protection from ground level in case of bombing attack.
The internet uses cookies, lol and sells users' data, ah ok.
The data center uses a rigid blackout to provide a seven-day generator for the power.
The data center controls oxygen levels in case of fire in the room that smells like christmas cookies.
The data center is quiet and loud at the same time and cooled down with water.
The data center holds and savekeeps information like a fortress.
The data center allows for complete communication. Spaß.
The data center has a compact oxygen level, just like its server racks.
enter text here
Stack Overflow User Survey result 2022
https://www.vice.com/en/article/pkp47y/metaverse-company-to-offer-immortality-through-live-forever-mode
HOMO + PHILIA
same, common + friendship, love
Parable of the Polygons - Vi Hart
https://ncase.me/polygons/
http://groups.csail.mit.edu/icelab/chimeria-gatekeeper/conv.html
Chimeria Gatekeeper (AIR project)
https://dev.to/reggi/stack-overflow-bans-answers-from-chatgpt-1d30
https://gpt3demo.com/apps/chatgpt
MOVIES
Her
Ex-Machina
Black Mirror
2001: Space odissey
I am Mother (2019)
Alien
Prometheus
Hitchikers guide to the galaxy
What's AI? 12/12/22
Cybernetics movement asked the question of can machines think and learn? They were optimistic.
ELIZA was an experiment that simulated conversations with a machine from the 60's.
AI taking over our human decision-making process.
Artificial intelligence is a diffiutl thing to define, because intelligence can't be clearly defined.
Strong AI is this idea of a machine knowing sort of everything, AGI, and Narrow AI is what we already see implemented in many places, very specific machines in specific settings.
80's AI winter where there was no funding and interest, from 2008 it came back again. In US it's heavily funded by military.
Loebner Prize (for Turing Test):
The system that performed best over the last years is called kuki. https://www.kuki.ai
Machine Learning a lot of data, which needs to be classified somehow. Either prepared manually be humans or automated. ML is a subcategory of AI, where it is learning on-the-go. It's always probability based, never 100% sure.
Supervised learning means building a math model with both input and output data shwoing it how to go from A to B.
Unsupervised learning doesn't give any output, the system by itself finds a structure in data to make the output by itself. The machine connects two things for causalities, but it's only a correlation.
deep learning is based on neural networks. Can be supervised or unsupervised. DL systems simulate somehow how the human brain works. GANs is a DL system where two different agents compete against each other.
Structural conservatism is how deep learning produces it's data, because it is based on historical data.
https://www.aiweirdness.com/
https://qualityland.de/
WMD <--> Weapon of math destruction
"Models are opinions embedded in mathematics." p.21 (Weapons of Math Destruction: What is a Model?, O'Neil)
Different Dimension Me is an AI based website that generates an anime version of images. Launched in November 2022 on the Tencent QQ platform, the website gained virality as social media users used it to create anime portraits of themselves and other people and characters. https://knowyourmeme.com/memes/sites/different-dimension-me
https://rlfbckr.io/work/rechnender-raum/
Project called Rechnender Raum by Ralf Baecker
https://www.volte-films.com/film/quantum-creole
QUANTUM CREOLE / Filipa César
enter text here
enter text here
enter text here
Jaquard Loom Programming cards
On 2021 May 2, a Tibetan woman in Sichuan province’s Heishui county carries components of cryptocurrency mining machines that were sent from the Xinjiang Uyghur autonomous region. Such machines are now being deactivated following China’s recent ban on cryptocurrency mining
Is The Universe a Computer?
Zuse, Rechnender Raum
He did not take words,
images or numbers as counterparts for objects as was usual, but
rather allocated digits to numbers for the first time: “Numbers can
be used to express all kinds of true sentences and deductions”
(Leibniz, De progressione dyadica, 1679)
Er nutzte als Entsprechungen für Gegenstände nicht nur wie üblich Wörter, Bilder oder Zahlen, sondern ordnete den Zahlen erstmals auch Ziffern zu. Leibniz (De progressione Dyadica, 1679): „Man kann durch die Zahlen alle Arten von wahren Sätzen und Folgerungen darstellen.“
Excavating AI
https://philpapers.org/archive/ZUSRR.pdf
Molleindustria’s “Phone Game”
https://vimeo.com/36775369
compiler
Abstraction
Willard Van Orman Quine
VirtualReality
The medium offers complete visual immersion; it not only opens
a window, as framed images do, as Leon Battista Alberti claims
in his treatise On Painting (1435). The Art of Immersion, however,
actually pulls the observer into the image and not only opens a
window as painting and framed art works do. VR is a gateway
through which viewers in the real world enter and leave the virtual
world. VR literally opens a door into another reality.ext here
Morse
Leibniz
Foucault
Wittgenstein
Galilei
Newton
Russell
Boole
Turing
Magritte
Wiener
Shannon
Hertz
Llull
Babbage
Lovelace
Open Codes is only revealed in the process of observers physical
interacting with the works. The participation of the audience is
the moment when the works come into being materially. Participa -
tory and analytical engagement with the works therefore includes
new forms of concentration and meditation as well as divertisse -
ment.
Ramon Llull - Ars Magna
https://adacad.unstable.design/getting-started/
VirtualReality
The medium offers complete visual immersion; it not only opens
a window, as framed images do, as Leon Battista Alberti claims
in his treatise On Painting (1435). The Art of Immersion, however,
actually pulls the observer into the image and not only opens a
window as painting and framed art works do. VR is a gateway
through which viewers in the real world enter and leave the virtual
world. VR literally opens a door into another reality.ext here
https://www.yourworldoftext.com/
Janelle Monáe — Dirty Computer
https://player.vimeo.com/video/268498567
JavaScript is turned off.
Please enable JavaScript to view this site properly.
Salon.io