Thursday, June 3, 2010
Imbuing computers with "common sense" has long been a holy grail of computer programmers and users alike. This article detailed some of the various attempts by researchers at the MIT media lab to create such AI systems that could do things such as discern user meaning without direct statements of mood, or suggest images and pictures based on content.
As Stephen and I mentioned in class, while reading this article we kept picturing the loathsome Microsoft paperclip. The paperclip was designed to facilitate document creation, however users HATED it. The application was intrusive and insulting (not to mention smug).
I think it is important to note that not only must these common sense systems understand content and context they must also have the sense to recognize user need. In general people hate being offered help or advice when they were not seeking it or when they were intent upon solving a problem themselves, it makes the person feel inadequate, or less than.
As I type this blog post I realize a successful "common sense" application that users have been using and loving for well over a decade now; the spellchecker. The spellchecker is unobtrusive as it does not immediately demand my attention. It simply underlines misspelled words that I can fix in the moment or go back through my document and correct all at once. In addition the spellchecker does not pretend to be my friend in the way that the paper clip does, it simply is. I want my machines to support and expand my sentience, not compete with it.