AI < IA > AI
A short version of a talk given at The TekLab London Seminar 2018.
The title indicates some initial confusion and different perspectives on AI. My perspective is partly based on theory, but I consider my role and experience as a content producer to be the most important.
From this perspective the meaning of AI is somewhat diverse. I have little experience with AI in the meaning Artificial Intelligence.
In the middle I switched the letters, because I find more familiar ground in what can be characterised as Intelligence Amplification, understood as various ways in which computers amplify our abilities to utilize information.
There is a shift of focus here, away from the computer, towards bringing the human actor more into the equation. This leads to another meaning of AI as Augmented Intelligence. That does by no means leave Artificial Intelligence out in the dark, but it has a stronger focus on the turntaking between human and machines.
What should the AI know about storytelling techniques in order to make really good and valuable stories for the human users?
When addressing this question I am most interested in the word “make”. However, I am not only interested in how machines can make valuable stories, but even more how machines can help me to make and tell more valuable stories.
Dialogs between humans and machines
How these stories are made and told does not concern me that much, as long as I am left with a sense of taking an active part. However, the combination of spoken words and physical motion seems particularly powerful. This means that my communication with an AI-agent will have to be dialogic, through language, touch and motion.
Emotion - move feelings by moving the senses
In the science fiction novel Snowcrash, Neal Stephenson describes a conversation between Hiro Protagonist and "The Librarian". This conversation is a deep dive into knowledge about Sumerian culture, a conversation that is possible because The Librarian is a piece of software that can quickly search through large amounts of information and make it available through speech.
I like the idea of having a conversation with a machine friend that makes me know more about ant topic.
What should this friend do?
The AI-narrator should of course adapt to me, and help me figure out which characters and events that are the most interesting in a specific situation. I would like to come back to short stories having these told from another character's perspective. The AI-narrator should also keep track of what I have been already told, what I liked more and less, and influence on how new stories might challenge me and at the same time blend in with things that I already like.
The AI-narrator should also help me when it comes to make connections between stories and the space that I am currently in – physical or virtual spaces, or a combination of both.
I reflect upon a story that still have some physical traces in Broadwick Street, in Soho, London. When the Google Streetview car passed by this spring and important part of this story was left out. We could easily pass by, as in any other street in London.
However, the AI-narrator know about my interest in visualisation through maps and it also knows that I am in London to talk, learn and discuss about AI. Therefore the AI-narrator will make me aware of the sign on the wall, why the pub on my right hand side is named John Snow, and the fact that there is a replica water pump, which was not there in the streetview images from April.
At this corner many potential stories begin. I am most interested in the map and how this was an important part of a coming transformation of London. Others will perhaps be more interested in the science behind epidemic diseases. Yet others will be pleased to learn about John Snow’s dot map, and how this is considered the first success of the nearest neighbor algorithm.
Today it may be the nearest neighbor algorithm that helps the AI-narrator figure out that it should make me aware of John Snow and the cholera epidemic in London, 164 years ago.
Linear - and spatial plot
If I am in a hurry the AI-narrator will figure out that it can tell me a narrative that is completely linear, like listening to a podcast. Just briefly make me aware of the pub and the water pump.
If I have more time available I would stop, and the AI-narrator will give me several options that invite me to move around at this location – exploring the story as I move in space.
If I want to the AI-narrator will help me to create my own story from this place, adding my pictures, thoughts and comments, maybe for others to explore
The AI- narrator will of course also be able to invite me to explore related stories.
I will all the time move between general space (platea) and specific places of meaning (locus), both in physical and virtual space. This also interact with a social space, where my perception of the narratives I am told unfold.
Story and discourse
The AI-narrator has to handle all the elements that can be part of a story (the potential building parts that can become a narrative) and discourse (the way a specific narrative is told in a medium).
Data – Interface – Representation
The last slides do of course not provide a complete picture of the complexity. There is no clear border between “system”, which is the AI-narrator’s domain, and the “user”, being both a consumer and a producer. The different models do, however, make me able to think somewhat more clearly about the tasks and how they can be distributed between machine and producer / user. As a producer I would be most concerned about the actual representations, but I would be happy to get some help and suggestions from the AI-narrator.