Menaquinone, Logo Daedalus, Nick, Kantbot, TV KWA
Welcome to KWALY, a new hot start-up
"...that will change forever how the ideas the rest of humanity are exposed to can be controlled by a morally superior enclave of progressive thoughtleaders.
Here we will manually deconstruct every problematic post and tweet on social media in a way that harnesses the power of crowds, of microtransactions, of organizational psychology. No rank fascist tweet or hateful crypto-nazi anime post will be left unblemished, all will be defaced by the hands of a loving, caring human being.”
INSTRUCTOR : TV KWA
The challenge of moderating hateful, hurtful, and harmful comments posted on social media platforms has become a job that not even the most advanced deep learning AIs are up to tackling.
"According to Kantbot, an artist in the exhibition, “This show explores moral entrepreneurship and what it means to deconstruct and control thought in an age when ideas are completely divorced as digital entities, from any tangible reality as objects.”"
"..recent studies have shown however that computers are haunted, and that their so-called ‘logic’ boards function by means of necromancy. The dead spirit of white male Aristotle is actually conjured in the tangling rivulets of electricity, emerging into the cold exterior of the computer to entice it to life using the siphoned strength of oppressed POC who hail from once proud cultures unable to resist the dominion of Western Logic."
Generative Adversarial Networks were first proposed by Goodfellow in 2014. It had been very difficult to develop good generative models for images before; the results were either blurry, excessively boundaried/pixelated, or totally incoherent. The genius was in defining the direction of ‘learning’/gradient descent via an ‘adversarial’/opponent system that would attempt to determine the ‘realness’ or ‘fakeness’ of the output.
The concept is pretty simple actually. ‘Normally’, you provide a set of ‘examples’ and optimize a generator such that, given a few parameters, ‘on average’ it does a good job of constructing an example that fits those parameters. This works ‘ok’.
In the GAN, you set up a ‘competition’ between two agents: the generator, as before, but also a ‘discriminator’ that attempts to determine generate examples from ‘real’ ones; optimization is performed on each in turn.
The results have been unnervingly effective; but this shouldn't be surprising: natural selection doesnt possess a concept of or even acknowledge ‘truth’ - it only recognizes conflict. To the human mind, the product of such competition is infinitely more ‘real’ than the product of a program of imitation of truth.
The battle between generator and discriminator has its genesis in the peculiar contingencies of anisogamy itself, the thermo-game theoretic equilibrium of Apollo and Dionysus, male and female. Apollo builds, orders, structures; while Dionysus deconstructs to perceive, renders essentialities, and delivers the capricious cruelty of the Universal Judgment, and such the generator and discriminator dance. He builds structure upon structure through dense matrices of pulsating, radiant connections and primitive operations; she absorbs his presentation in full and digests it, putrefies it, and once its constituent elements can be apprehended she gives voice to entropy and provides her verdict. This dance of striving/deception, choice/elevation/induction is the source of all the vigorous complexity of metazoa; the results are truths that we can understand, provided in god’s tongue of ceaseless struggle. An alien, machinic mating ritual written in our language of slime.
Date: 21 - May - 2017
Time: 6.37pm GMT
Location: 2-4 Tottenham Rd. N1 London, UK
N suspects: 3