Still debating the merits of AI in 2007?
Thursday, May 24th, 2007Here we are in 2007 and we still debate the merits of AI! AI is over 50 years old as this nice review of where AI is describes. http://www.post-gazette.com/pg/06002/631149.stm I was speaking on the Ferris Emergent Technology Webinar last week http://www.ferris.com/2007/05/16/emerging-messaging-technologies-3/ and amongst the topics we got to debate was the merit of AI. Specifically, I am still asked why should anyone trust AI? As a practitioner of AI for over 25 years now (the half life of AI!). I have often explained to people where AI works well and where it does not. AI is very effective in specific well-bounded applications. You can find many great search tools based on AI techniques of statistical and linguistic analysis. Indices are used to point to bodies of text for search based on the probabilities that the text actually relates to something very specific! If search engines used even more structured AI such as ontologies of relational concepts, we would even get better results! We have also developed great data mining applications based on AI. Let us also not forget about all the AI in games these days as well as the bots that scour the net while we sleep! Another well-bounded area for AI is content analysis for outbound email content control. We effectively implemented this for compliance applications to prevent Securities, SOX or HIPAA violations or company secrets from going onto the Internet http://www.amikamobile.com/news_events/pr/press_release190504.asp Now at Amika Mobile, the challenge is to pick our key points from email and maximize the use of the 2×2 inch micro-screens of mobile devices such as the typical mobile phone that most people walk around with. We do this by again using statistical and linguistic analysis on the content and ranking what is most important. To me this is not rocket science or magic any more, it is simply good old algorithmic software engineering!