The Infinite Adventure Machine (TIAM) is a proposal for a computer program which generates fairy-tale plots.
While fully automatic story generation remains an unsolved problem for computer science, this project explores the links between imagination and computation. Tales and myths; the core narratives of human culture, have been transmitted for generations through various technologies and media. What new forms might they take through digital formats and Artificial Intelligence?
Based on the work of Vladimir Propp, who reduced the structure of russian folk-tales to 31 basic functions, TIAM aims to question the limitations and implications of attempts at programming language and narrative.
Because the program is unable to deliver a finished story, rather only a crude synopsis and illustrations, users have to improvise, filling the gaps with their imagination and making up for the technology’s shortcomings.
TIAM was commissioned by:
Microsoft Research, Cambridge UK
Microsoft Office team
Design Interactions, Royal College of Art, London UK
as part of the Future of Writing project
In this interview David Benqué will talk about the project.
Institute of Network Cultures: Can you introduce yourself: your background and what you are doing now?
David: I’m a designer and I work in London. I received a BA in typography and graphic design, which I studied in the Netherlands. Then when I moved to London and did an MA in design interactions at the Royal College of Art which shifted my practice: I moved from a more traditional design to a practice where I look at the implications of science and technology, instead of their direct application. Right now my work consists of speculative projects where I work with scientists to try and think about scenarios of how technology and science impact our society and culture.
INC: Can you tell me about TIAM? How does it work?
D: TIAM – The Infinite Adventure Machine – is a computer program that tries to generate fairy tales, so it prompts the user with fragments of plot elements and some images, and basically the idea here is that the user has to improvise from these elements. As it turns out, computers are not able to generate stories, and this was the main issue that came out of the research I was doing, so the project turned into an improvisation tool.
INC: So does the user have to fill in the blanks of the story to complete it?
D: Yes, I guess it’s important to say that there is a timer, so there is pressure and the user has to make decisions quickly.
INC: Can you better explain the whole process? Do you visualize some parts of the story, some phrases, and then verbalize them?
D: Yeah, exactly. It’s based on a theory by a russian structuralist called Vladimir Propp who came up with a sort of DNA of fairy tales, he reduced the structure of russian fairy tales into 31 basic functions. For him that was a tool to analyse the tales, but for me it was a prefect opportunity to reverse it and to make a generator, since it was such a tightly structured system.
INC: Can you describe the process that you followed to develop the concept and then the software?
D: I should first say that this project was initially commissioned as part of a bigger project, it was a collaboration between Microsoft Research in Cambridge and the RCA, the name of the project was “The Future of Writing”. Five designers including myself were asked to look at the concept of authorship and go beyond the “tablets vs. paper” debate that is so frequent nowadays, and try to look at this topic in a more tangential way, pushing these ideas much further.
I started off with the TIAM project, having a strong interest in narrative. As a matter of fact it always plays an important role in my work, as does fiction. Fairy tales are good examples because their functions are very identifiable, they address the transition from childhood to adulthood, and all of the psychological aspects that come into play.
So my question, when I started the collaboration, was if we could find a DNA for stories. Once you have this DNA, the set of parameters, it’s a short step to think of what you can do with it computationally and the potential to automate storytelling, which is such a core human activity.
That was the starting point and from there I came across the work of Vladimir Propp which was a perfect fit and I also looked at a lot of different attempts ranging from artificial intelligence experiments and how we engage with computers in general.
INC: Can we call TIAM a digital book?
D: I don’t know if we can call it a digital book.
I think that the underlying question was more about authorship and how it operates nowadays or how it will possibly operate in the future.
Another inspiration was a book by Neal Stephenson called “The Diamond Age”, a story about a nano-tech future where a sort of magical technological book educates a young girl. The book adapts to her experiences and teaches her the core values and lessons of life etc.
So I was quite interested in this idea of the book being a fluid and flexible device more than something set in stone, so I guess in that way TIAM could be called a digital book, but for me it is essentially a computer program.
INC: What have been the parameters, limits and rules in the design?
D: It’s a pretty simple program. The 31 functions identified by Propp are the core database of possibilities. The program runs through them and selects which ones it’s going to use and prompts the user with that.
There are two main rules: one is the function of linear progression, which starts at the top and ends at the end, and you can’t go back: for instance, things like “the hero leaves home” it’s a typical node in the scenario and happens near the beginning.
The second rule is that some functions are linked to each other so if there is a particular quest there has to be a resolution for that quest. For example, if there is a quest for a magical agent, that quest has to be solved – or not – and that happens later on in the plot structure. It’s the same with the departure: if there is departure, then there has to be a return.
But overall it’s a pretty simple set of rules, which is another reason that I was attracted to this particular theory, Propp’s theory, as a basis for the project.
INC: How are the images generated in relation to the content? What do you think about this relationship?
D: The images are actually made by hand and it’s all pretty linear.
I wanted the program to reflect and in a certain sense criticize the way that a lot of these so-called “story generators” have been made: just taking pre-written bits of stories, shuffling them and selecting them in a random combination, which is not at all a story generator, it’s just a shuffle algorithm.
I intentionally made that tool so that the images, which are in 3D, always build themselves up, so it suggest that the computer is building them.
INC: Is there a particular way to deal with the content structure (for example, indexes, search options, different levels of visualization etc.)?
D: There is very little interface because I wanted to maintain that sense of mystery of what’s actually going on behind the scenes. So there are slight delays and other things that suggest that the computer is working, but in terms of input from the user it’s basically only about starting and choosing the speed, there are no other options. This is intentional because I wanted the user to focus just on pressing start and then have to deal with whatever the program throws at them.
INC: Is it possible to create story in a collaborative way with TIAM? Does the application interact with other media platforms (for example, a web site, social media, etc.)?
D: In terms of collaboration I think the example video that is online shows it well: it was actually shot live, Matt and Tilly were actually collaborating and that became really interesting because they had to deal with each other as well as with the program. They were both steering the story in a certain way and had to make some decisions, interpreting the flow of elements while the other didn’t always agree, so there is also this kind of challenge, on top of the normal challenge of reacting to the program in time.
In terms of linking to social media there is nothing built into it at the moment and I don’t really see how that would add anything interesting, because it is really all about the live moment of the experience.
Is there a way for me to talk and create the story while the software types what I’m saying? Can I have, in addition to the live experience, a sort of output and then something like a recording of the story that I have created?
That’s a great question because for me it’s the next step. I’ve been thinking about putting it online as a kind of web app but I don’t want to simply put the program online as it is, because I think that the most interesting thing are the actual stories that come out.
The way I’m working on it now, I think it has to be done with the audio, by recording what the user says and the output of the program at the same time, and then showing that as a record of a generated story.
Could it be something like a flow of information or a flowchart? So that if I say something, the application gives me some options to choose from. So that the software follows what I’m saying and there is something like a dialogue between what I’ve created and what I’m saying and what the software proposes to me.
Of course that is one way, but that would require artificial intelligence which doesn’t exist yet and is not likely to happen any time soon.
I rather think that the dialogue is between what computers throw at us, and how we make sense of that as humans. I’m not that interested in creating a feedback loop but more in finding a way of recording human response to this quite simple program and see which narratives come out of that.
INC: You’ve already told me about that in the previous question, but maybe you could add something if you’d like. What were your main goals in the beginning and did you reach all of them? Can you do a self-critique? What works and what does not?
D: As I’ve mentioned before, the project itself is actually a story about a computer program telling stories if that make sense. So the thing about having the users fill in the blanks and the computer not being able to finish the story is a failure that is used in the project to make a point. As I told you the only part that isn’t quite resolved yet is a way to build a library of these stories to possibly show them online, my idea is to open up a big catalogue of all the outcomes.
INC: Did you get positive feedback from the users? Did this feedback give you further inspiration to solve some of the project’s problems? Are you still working on it?
D: The project gathered a bit of attention online: it was blogged about a few times and it was shown in exhibitions as well. The feedback has been really positive and I think people really engaged with it and not just in a playful way, it really brought about some questions and conversations on the topic of generative narrative in the digital environment so I was really happy about that.
I’m still working on it and the next step will be setting up some workshops and finding a way to record these workshops and hopefully pushing the program online at some point.
INC: Now I have some more general questions: Do you see a narrative potential in all the software and social networks that we use to communicate? What do you think about Twitter as a narrative tool?
D: About the narrative potential I think Twitter is a good example, whereas with social media like Facebook the narratives are more like info dumps of people’s lives, like what you are doing for your holiday or what you are having for lunch, which is not necessary that interesting.
With Twitter some interesting things are happening at the moment. There are numbers of accounts that are re-exploring history, making historical re-enactments, for example, there is one for World War II.
I think these kinds of experiences can create an interesting dimension, combining both short bits of text and archive photos, so that you can be looking at Twitter and it’s like: “Oh! 70 years ago today this part of London was bombed” and you’re actually sitting on a bus near that neighbourhood.
I’m currently thinking of another project in that space of reverted narratives and historical re-enactments: it’s about telling a fictional story from the future, or a parallel world through Twitter, but I’m just at the starting point.
INC: What is the main difference between a book and an e-book? What about other digital editorial products (iPad applications, digital magazines, devices for reading in the web etc.)? What about calling them all “reading experiences”?
D: I’m not sure how useful it is to cling to the idea of the book, I mean, I’m not saying that books are disappearing, definitely they will be here forever in whichever form.
However I think the most interesting thing to notice here is that we are now writing and reading more than ever in our history, in a lot of different ways. This is interesting because it impacts the language that we are using, which is now evolving to fit these new modes of communication.
Furthermore, relating to the TIAM project, what I found interesting was that the notion of authorship in the digital environment is changing, for example with things like video games. What actually gets created is a set of rules for the people to play with in a non-linear way, as opposed to the notion of crafting a narrative which is linear, with a beginning, a middle and an end.
I think this is an interesting development and something to really observe, and maybe here is where the biggest difference in reading experience that you’ve been talking about is, because we are actually creating something which is again a set, more a defined territory rather than a line. During the research I did for the project, this was one of the most interesting findings.
INC: What do you think about the obsolescence of the digital world?
D: What I find interesting is all the stories, myths and core narratives that have been around for thousands of years, whether it’s the creation of the world or all the typical archetypes of characters that come up even in Hollywood movies, I think these will stay.
So yes, it’s a more ephemeral medium, but I also think that we’ve always told stories with such mediums. With oral storytelling for example, things are over as soon as you say them, there is no exact record but in any case, it was the vehicle for all of these stories to travel through the centuries. I think that maybe we are actually coming back to that in a sense.
Turning this around, now everything is also constantly archived, so the problem is more about limiting than about keeping and recording, or in making sense of a huge number of archives where every photo or every line of Tweets get recorded: how to make sense of all this information?
INC: We are living in the era of the dissemination of information. Do you think we are lacking concentration?
D: In many ways we are shifting into an environment in which skill isn’t about having knowledge but about finding knowledge, but I don’t see this as a bad thing. Obviously there is an ever increasing number of situations in which different things want your attention and you’re just jumping all over, never diving into anything with any depth.
There is an interesting cycle where technologies are used to solve the problems they themselves have created. So now some people are making apps that cut off Internet access or allow you to concentrate. Probably because to get anything done in your life you have to learn to be able to shut up the Internet access sometimes.
INC: Can art and design practices make a contribution to the development of proper structures, models and even technological devices for digital publishing, and if so, how?
D: I’ve studied typography so I’m quite aware of the history of design enabling the publishing of books and I’m sure this will continue in the digital space.
I think there is also a space for design to be used as a tool for exploration and critique of what we want to make of technology and the possible ways that we could use it. In these stories, scenarios or ideas, design is used as a medium to reflect critically on technology, instead of just a way to make products.
This is maybe not so much about digital publishing but rather as a way to engage with technology in general. I think this is also a role that designers can have, and for me that’s a way to contribute to digital culture and knowledge.