This week we got a closer look at what Artificial Intelligence actually means, thanks to Taller Estampa. The first day we learn step by step some basic concepts around the topic to be able to understand the functioning of this tool. When can we  say that a machine thinks? We would say that a machine thinks if for instance can calculate or play games that implies strategy, ches. This substantive definition got pushed further and further along history. There were some experiments runed with machines and humans in order to determine the capacity of the machine. An individual couldn’t tell when they were talking to the machine or to another person. Generating a new pragmatic definition of artificial intelligence.

Machine learning means automating tasks by providing examples, that is to say, training data instead of writing a set of instructions (coding). In order to understand what deep learning means we were explained what the functions of a neural net are. A neural network requires to be feeded with a well thought data set specifically related to the outcome we want to get. We as humans decide a lot of things about these tools so it won't be fair to say that it is 100% automated. We decide what we want it to do, what data we use and neural network we need to use. The only thing that we don't decide is how the neural network self-configures, this last process is based on the data set is called training. Considering that humans provided the examples that neural nets will work with, we can not ignore all the biases that this implies. We discussed as class some examples like using machine learning to classify criminals or how google photos associations would many times be wrong.

Data classification

We were asked to generate a data set of images from poblenou as well as from our phone’s libraries. Later, Pau provided it to an image classification neural network and like that, we understood in a more visual way how the net finds connections. I found interesting the relation that net found between Vikrant’s Magic machine and the Capitel of an ancient column.  

On the following days we got to play a bit with neural networks in google collab. Finding ourselves on the latent space and changing the truncation value to generate unexpected portraits.

I personally had a more romantic vision of what artificial intelligence would be. I’m a huge fan of science fiction movies and series and always ask myself how robots would be able to develop feelings. I remember an assignment that I had to do for university about the concept of technoromanticism which actually contemplates this idea of when the machine stops to do what it has been programmed for and when it starts to think and feel itself. I would have loved to talk more about this concept in class, but to be honest I was feeling very sick and it was hard for me to interact in class. Nevertheless I very much enjoyed the conversation that came up and it was really interesting to see the uses of AI in the artistic world. I would personally like to use an image generator neural network to create new ceramics pieces based on ancient ones and new modern and optimized ones.

Speculative project

As a final task for the seminar we needed to conceptualize, program and present a speculative ML project. I was part of the group formed by Paula B. Nikita, José, Busi and Chris. Thanks to the previous experience of Chris using OpenAI’s GPT-3 language generator model, we were able to develop our idea a bit further. The idea was to use the network to rewrite the UN's Universal Declaration of Humans Rights for an animal species, optimally for one that humans easily disregard. This could be a way to empathize and provoke feelings towards ants and finally propose an entry point for interspecies legal frameworks. 

Ultimately we decided to create a website to present the output co-written with GPT-3, the Universal Declaration of Ants Rights. Please click on the link below to also sign the declaration.