Contempora intearther -machine interprets my dissertation and end of the year musings about AI
One of the big things in digital technology this year has been the advancements in AI and in particular in machine learning (ML). A lot of great and exciting things has been written about them this year. From the almost incredible mind-reading AI to dystopian visions and further. Despite, AI and ML are producing novel ideas and applications, in some way the current hype of AI and ML reminds me of the hype of big data, or digitalisation, just a while back. The idea remains the same, but the buzzword has just changed.
Art has not been left out of the AI-ML discussion. From questioning the role of art to discussing the vocabulary of AI art. (See for, e.g., Here and here AI has stirred the art world and given new possibilities, methods, and medium. One of the big things seems to have been the advancement of GAN’s, generative adversarial networks, that are capable of producing original images derived from broad sets of other images. One can, for example, discover new pictures with the help of Ganbreeder, that uses Google’s BigGAN-algorithm.
The use of AI, whether in art or in general is not without problems. Unfortunately, the hype and coolness of many of these discoveries drown the needed discussion of, for instance, application, ethics, and sustainability of these technologies. I have enjoyed reading Dan McQuillans takes on AI from Rethinking AI through the politics of 1968 to The poetic losing your voice: as well as the manifesto for Algorithmic Humanitarianism. These articles do a great job questioning the role and position of AI in our society.
To get to understand better the potential in ML, I too wanted to use machine learning for something. However, I have not been that interested in image creation with ML, such as using GAN’s or other similar algorithms. So, instead, I tested out the LSTM-algorithm by feeding it my dissertation and taking a look at what can be done with text. Such practise is quite easy for everyone to test out by using p5 and the excellent ml5-library.
If you want to see how does AI interpret my dissertation, you can try it here. Just change the seed word, to hear about a topic that might interest you and see what happens. You can also adjust the length of the produced text and the temperature; the closeness to the original text.
As one can see, my AI is not very bright, but maybe it can give you delightful predictions for next year!