August 2, 2005
Bio: Dr. Richard S. Wallace is the Chairman and CEO of the ALICE A.I. Foundation, Inc., a non-profit organization devoted to the development and adoption of the free AIML software for creating chat robots like ALICE. Dr. Wallace is the three-time winner of the prestigious Loebner Prize for .Most Human Computer. based on the famous Turing Test for artificial intelligence, and recipient of numerous other awards. His work has appeared in numerous international newspapers, magazines, television and radio broadcasts, and all over the web. He is the author of two books, Be Your Own Botmaster, and The Elements of AIML Style. You can support the development of AIML by joining the A.I. Foundation at http://www.alicebot.org/join.html
Botworld: Do you think pop culture, most especially their image in movies and literature, has hurt the image of artificial intelligence to the general public?
Wallace: Not at all. Hollywood is the greatest advertisement for A. I. and robotics in history. The problem is with academic scientists and engineers not living up to the public.s expectations. A system like ALICE, which has won the award for coming closest to passing the Turing Test, could never be built in a University research lab. The pressure to be politically correct and to confine one.s research to the areas approved by the establishment, not to mention the scale in years and manpower, would prohibit any kind of believable A. I. from emerging from a University, or any government-funded research lab. This is one reason we see more advanced developments in hardware robotics emerging from Japan, which is not to say that Japanese scientists are free from political pressures. But the kind they face are different than those in the U.S., which tend to stifle innovation and creativity.
ALICE, by the way, has been used to help promote several Hollywood movies and TV shows on their web sites, including ABC Alias, Lynn Hershmann's Teknolust, and the Steven Spielberg-Stanley Kubrick A.I. movie.
Botworld: What do you think is needed to advance the evolution of AI and overcome any negative perceptions?
Wallace: The two greatest advancements in A. I. over the last 15 years had nothing to do with A. I. The first was open source. We have in place a system of building and sharing knowledge, of taking advantage of an economy of scale provided by a large labor force of volunteers, to accomplish something that could once only be done by a large corporation or government research lab. The ALICE AI is a free software project, much like Linux or the Apache web server. Over the years, we have had contributions from hundreds of developers from all over the world. I have often said, how else could a person such as myself, with barely any working capital, create a software system that has captured, as some say, 80% of the market for chat robots, without the magic of open source?
But even more significantly, I believe that free software stands in historical significance to the proprietary approach, as the University once stood to the Church. We are living through a kind of Reformation, where the very mode of knowledge acquisition and dissemination is changing before our eyes. We can even see parallels to the counter-Reformation, when Microsoft agrees to open up parts of its code to their most trusted customers. But I am getting off the topic.
The other great advancement was the internet. Before the web came along, we never really had a chance to connect a natural language bot to a web page and collect a large corpus of data from people trying to have conversations with the machine. This was really important, because it told us what people thought the bot should be able to say. There is a statistical fact of language, called Zipf.s Law, that people tend to repeat themselves, or repeat what they hear other people say, over and over again, much more often than they say original things. So with the web, we could measure the the frequency of things people say. The problem of building a believable bot was reduced to attacking those things in order, starting with the most frequent and working our way down the list. By the time we covered the top 45,000 or so things people might say, we had built a fairly conversational bot.
Botworld: Do you think the first common forms of robots and AI will need to overcome some human intolerance?
Wallace: One of the biggest obstacles to human acceptance of chat robots is suspension of disbelief. A child can have more fun with a bot than an adult, because the kid will forgive the bot when it breaks down and gives an incorrect answer. Adults, especially highly educated ones, will tend to be more critical of the bot.s mistakes. There is actually a tension between part of people who want bots to be like super-intelligent machines, always accurate, truthful, and precise; versus the part of us that wants robots to be more human, which means something like the opposite: sloppy, lying, funny, hypnotic, charismatic, and maybe sometimes truthful and accurate. Robots might be telling us to get over ourselves.
Botworld: What do you think are some of the most positive images of AI?
Finally, if you feel the public's perception is negative, do you think it will hurt the advancement of truly complex AI?
Wallace: I'm going to sum up those last two questions in one by saying that no technology has ever been either entirely positive or entirely negative. Moreover, technology has a kind determinism, or at least a natural course of evolution, that appears to skip over the minds of individual inventors, despite their egos and individual passions. So I don't think you could do much to help or hurt the advancement of anything by manipulating public perception, not for very long anyway.
Chat bots have been applied to entertainment, teaching English as a Second Language, and selling cars. Under a darker scenario, one can easily imagine them being used as interrogation machines. The problem is not with A.I. or any other technology, but with our own human brains. No one has proved that the human brain is smart enough to solve all of the problems it has created.
Copyright © A.L.I.C.E. AI Foundation, Inc.