Me, myself and AI

Posted on Jan 5, 2008 by in artificial intelligence | 0 comments


The only thing we can state as a hard fact in the {en:artificial intelligence} field, is that its definition regularly changes with time. It is very likely that the one we give it today will fully differ in a short period of time (typically, we can say that its definition changes every decade). Could you imagine a domain like mathematics, or even one of its sub-divisions like calculus, being redefined every decade?

All this because of a single truth : we don’t have any axiom to rely on which has been proven worthwhile over a long period of time. In other words, we don’t have a clue of what intelligence really is. And I am not being theoretical or picky here : any valuable core specification will do, even if it is neither the only one, nor providing miraculous results : any basis of axioms supporting the field in a sustainable way : meaning a way we can build on for good and which is not {en:domain specificity|domain specific} .

Considering this, if you want to work in this field, the only rational thing to do is to chose your battle : Should I try to help the creation of a solid basis for AI, or should I try to do something with whatever is considered as useful in the current period? If you are like me, you will choose the second one immediately, not because the real holy grail quest is pointless, but because you have an unstoppable urgency of creating something which works and to put it on the market.

Does it mean we should have no faith in anything which has been developed in the AI field? Of course not. The commercial development of the current results of research in the field of AI can change the face of the software market, and the results of tomorrow will probably enable us to revolutionize the entire market place. This is why, for me, the real question is: What are we waiting for?

Well, what the market – and especially the venture capitalists – are waiting for is pretty simple : they are waiting for it to make money. Now, as we can see that the VCs are very reluctant about the term AI (so much money has been already lost), as well as customers (who are systematically disappointed that what they bought is not as smart as they thought), we can conclude somehow that the market for AI itself is pretty much doomed. Indeed, anything coming out of this field which happens to work is very quickly renamed as a non-AI technology, in order for companies to raise money and make their customers less disappointed.

My favorite definition of AI is the one of Bill Bulko: “Artificial Intelligence is the art of making computers that behave like the ones in movies”. One could argue that this definition is a joke, but in fact, it is not worse than any other and it has two major advantages : first, it is funny, and second, it defines AI as an art and not as a science, which allows us to do stupid things and yet claim they make sense.

Now, to go back to the title of this post, what do I have to do with AI? Actually pretty little: I have studied it, especially for my Master’s degree, and have made several trial and error moves in the field of machine learning, expert systems and intelligent user interfaces (click to see their definition by the Association for the Advancement of Artificial Intelligence).

I have quickly realized the opportunities and difficulties of the domain and enjoyed them immensely. AI is a domain which is fundamentally both rejected – as the resulting products and technologies don’t have well defined behaviors – and needed by the market – as it is clear that the behavior of humans is often even less well defined. This duality is fascinating to me and I believe that working on innovative technologies needs this type of background to be worthwhile.

I also realize the difference between a technology which works and a technology which makes money when I created the most efficient children’s Internet protection filter on the market – way ahead of McAfee and Norton – (see results of April 2005 of Action Innocence) and made absolutely no money from it.

At Boxalino, we have been focusing on technologies which improve the user experience for a while now, and I am convinced Intelligent User Interfaces will become one of the most important software markets within a decade, especially if the technology is set-up as a platform delivering service (like an intelligent portal, or search engine). One of the reasons I believe it will become a domain in itself is that you can look at any software product from a user experience point of view in order to evaluate it; the same way we do by looking at its features, performance, database structure, etc.

All this is quite theoretical and fuzzy, but I wanted to make this post an introduction for this blog. Of course – as it has been already done many times – complete books should be still written about AI in general, as many things still have to be explored, discussed and defined, but I don’t believe to have any ability to bring value at this level. Instead I will try to be much more domain specific and speak of global problematics, but from a quite specific perspective.    Send article as PDF   

Submit a Comment

Your email address will not be published. Required fields are marked *