Thoughts on CGI conference

People have been really challenging me in the same way, "I think you're not making a tool. You're making a game", "What are the fundemental, building blocks of the functions such systems should possess? How would you broaden such functionality, and broaden the behaviors that the users can edit?", "Can the users edit the existing behaviors? Or they have to delete the previous one and recreate it all-over again?" "Can the behaviors be represented like video strips? If it would be better to represent the behaviors in tree structures instead of linear representations, such as video strips, whose power of expressiveness is over-simplified, what would be the ideal solution for this problem?"


These have always been the questions people are asking be about, but I've never found the correct answers by this time - even those proposed by others. Someone came to me after my presentation and talked about these problems with me. He and I both agree that it's definitely not a simple task - honestly, a delimma - to allow users to achieve complex task in a simple way. "The current behavioral structure is relatively too simple for users to make up humans' arbitrary behaviors," he said. Indeed, my proposed work would never be really practical untill it becomes powerful enough and generalized enough. But the problem is, once we make the model complex, how are people gonna do the manipulations in an easy way? How should we visualize the representation such that people will feel comfortable - but not scary? How can we reserve their controlability over it?


We made no conclusions. Afterall, no one ever gives clear conclusions yet. The delimma is basically the major problem that the whole End-User-Programming area wants to solve - "to make hard things easily achievable". I think it is important to find a way out that the sooner the solution is found, the sooner how users can deal with computers can be different, and all computer users will be benefited. I made it clear that to me it is important to have the interface "stay simple". Although it has been keeping me I trapped and making no progress, I believe the insistence is undoubtedly correct.


So I'm proud of myself that I'm facing something people have turned their faces away from. And I'm actually pretty confident that the solution is not so far, after reviewing a huge portion of the commonsense computing work done by the Media Lab, MIT. Because of the deep belief, "Human-Computer communication cannot be simple for complex task because there are so many things that computers don't understand; once computers know all that we know about, we can communicate with them easily," they have done so much work on making computers even closer to how we humans think, and their beginning stage on applying commonsense seem to be successful. For now, I wouldn't say that novice-edited behaviors that are finite-state-machine-complete is impossible, now that Hugo Liu can already turn natural language descriptions into python.


The founder and executive of the company, "digital fish" also came and talked to me after the presentation. He is searching for exactly the same solution as mine - making a behavior-authoring tool that is really easy for novices. As I talked about the possibilty to him, I felt so glad that someone seems to be really caring about what I care in the research area. It seems that I am starting to be a researcher and people will discuss about my research, rather than that we always have to follow people else's footsteps.