8 thoughts on “Watson, Artificial Intelligence, and a male-female Turing test”

  1. Though it is very hard to define what “intelligence” really is, it is much easier to tell what it is not.

    I’m occasionally listening this AI-gaga, since 25 years. The idea that a traditional program, tailored to follow a specific task, will ever achieve something we could name “AI”, sounds hilarious. At least to me.

    Intelligence is about to cope with situations which are completely new and unforeseen. If a computer will be able to “learn” chess, without being instructed in its program code what a “game” even is, then I’ll be impressed. Not earlier.

    I support your analogy regarding google. Could also be Wikipedia or anything else. It is like the difference between a man with good knowledge and mileage in repairing cars and one who runs a guidebook section in a bookstore. The knowledge is all there, but not “owned”. Well, may be in a legal sense. 🙂

  2. Watson solved yet another symbol manipulation problem better than the best competing humans. This is not a new result and does not reveal anything new about the potential abilities of computers.

    As you noted, early AI researchers thought that skill at symbol manipulation (e.g. chess, calculus) was the defining characteristic of intelligence, since the people who were best at those tasks were clearly very intelligent. As we have found since then, computers can be programmed to do well at symbol manipulation tasks, even when they are incompetent at anything else.

    My view, though I am not intelligent enough to prove it, is that human intelligence is characterized by its ability to receive continuous sensory inputs and identify manipulable symbols within them – as you say, “convert a continuous stream of audio into distinct words”, and then manipulate them in sometimes symbolic, and sometimes non-symbolic(this is where the magic happens) form. Neuroscience will eventually show us how neurons represent the immense variety of phenomena in the world outside the skull. When we learn how to copy that mechanism, we will have intelligent computers. Ironically, they may not be good at symbol manipulation tasks.

    In summary, I agree with everything you said.

  3. For an alternative approach to AI based on the pattern-recognition facility of the human cortex, I suggest “On Intelligence” by Jeff Hawkins (creator of the Palm Pilot and Graffiti), now being commercialized at Numenta (www.numenta.com).

  5. Hey Phil,

    That Turing test comes up quite often in conversations these days. It really was great fun. For completeness, the mighty Golems also won the next round with a single question of mine, “What is a ‘Fighting Falcon?’ ”

    Female answer: “Um…a falcon that fights? A big bird?”
    Male answer: “F-16, airplane, cornerstone of America’s international fighter exports.”

    Some thought obviously went into the advanced selection of those questions, mostly around likely contexts of historical learning unique to either sex, as well as around questions laden with context unrelated to the direct question. Lateral thinking and all.

  6. Hey Phil,

    I think you are selling it short. Even “Watson” which is in its infancy in terms of AI is a huge leap forward. A computer could hardly beat a human at chess 15 years ago and now it has won Jeopardy.

    In terms of applications in our society, as with any new technology and invention, the people who create it often never see most or all of the possible applications. Sure, google could possibly do all of these things but it requires your help. For example if you just type an entire question into google, it will usually just take you to ask.com or something which then will lead you to comments and all sorts of other places and maybe some other human eventually will have a link to some factual information. Watson will eventually be able to take a vague question and direct you to factual and reliable information immediately.

    In addition, the often cite the medical field as an application and to me that makes sense. At any given time there are millions of doctors all over the world doing research, seeing new conditions and problems and treating them ect. Watson can look at all the symptoms and virtually instantly compare it with all information available. It could help prevent misdiagnoses and also allow for quicker referrals to proper specialists. I know many of the doctors I talk to know nothing of some of the medical breakthroughs I am aware of as they just aren’t interested in things like that.

    There are also many other application I can think to set Watson to such as trying to find associations with news and the stock market and or analyzing economic information and correlations to build a proper model.

    It’s a good first step IMO.

    Jesse

  7. Jesse: A computer could hardly beat a human at chess 15 years ago? http://en.wikipedia.org/wiki/Chess_(Northwestern_University) says that a computer program was winning tournaments in 1976, i.e., 35 years ago. By 1977 the same program was rated stronger than approximately 4 billion out of the world’s 4 billion humans (and weaker than a few tens of thousands).

    AI has been in its infancy for more than 50 years now. When the AI labs did not give the U.S. government the promised results, the standard response was “AI is in its infancy; this is a 20-year project”. After 20 years, the standard response became “this is a 30-year-project”. Now we’re halfway through a 100-year quest! Don’t stop the funding now!

    Check http://en.wikipedia.org/wiki/Artificial_intelligence for the history of AI, with some optimistic quotes from the 1960s and ending with Watson.

    I do think that Watson is interesting, but I’m not sure how much closer it gets us to the computer that can write a high school term paper or carry on a conversation.

  8. I agree Phil and I do not think the government should be investing or rather “spending” any money at all on AI. However, I think that Watson actually has some practical applications in the private industry that add value to society. It is only at this point (IMO) can any real advances in an industry start. Once it become a self sustained private enterprise.

Comments are closed.