Another combination of topics I tried was ‘Machine Talk + Mythologies’. ‘Machine Talk’ referring to the communication between machines and humans (error messages, notifications, etc.) and ‘Mythologies’ being, in this context, the Roland Barthes’ theory, which I’ve written about before.
In a few words, ‘Mythologies’ is a collection of essays in which Barthes shows that images/concepts – or according to his terminology, signs – are stripped of meaning when removed from their context. (+)
Language is complex and ambiguous. Words have different meanings depending on their contexts. If we, humans, can have different interpretations of an excerpt when it is taken out of context, it is almost impossible for a machine to understand its original meaning.
It was around this idea that I started looking into machines, AI systems and communication barriers. I wanted to figure out why it was so difficult for machines to understand us.
I found a very interesting article in MIT Technology Review on deep learning, the AI’s current language problem and possible solutions. “There’s an obvious problem with applying deep learning to language. It’s that words are arbitrary symbols, and as such they are fundamentally different from imagery.”
I also looked into the Turing test and found some articles about the first AI system to pass it. These can be found here & here. I found it quite fascinating, tough a bit scary, how machines are becoming so developed and advanced that humans can actually mistake them for other humans.
There are a few riveting films that relate to this as well. ‘Her’ (2013) and
‘Ex-Machina’ (2015) both explore the concepts of social constructs between humans and AI systems and some of the consequences that may derive from it.