Forum for Artificial Intelligence — Fall 2003

Better late than never: Matt MacMahon and I finally have the fall schedule lined up for the Forum for Artificial Intelligence. The main cause for the delay was that Pat Beeson and I both wanted to “retire” from FAI this year, I having just completed my dissertation proposal and he trying to propose this semester. Generally only one FAI organizer retires at a time, to ensure continuity. In the end, I decided to stay on, and Matt replaced Pat. Since all three of us are taking a class friday afternoon during the traditional 3-4pm FAI timeslot, we’ve decided to run the series as a brown-bag lunch talk/discussion series, though sometimes speaker schedules will require that we depart from that format.

Anyway, I’m pleased to announce that the first speaker will be Tony Cassandra, from St. Edwards University CS, talking about Partially Observable Markov Decision Processes, a mathematical model for reasoning with uncertainty. The talk will be at 11:30 a.m. in Taylor Hall 3.128 at UT.

The computer says I’m a visionary?

So I finally took the Meyers-Briggs personality test. It said I’m an ENTP. I’m often dubious of these kinds of things, but this and other ENTP characterizations seem to be pretty accurate. Some telling quotes:

ENTPs … generally love to argue … and enjoy playing devil’s advocate. They sometimes confuse, even inadvertently hurt, those who don’t understand or accept the concept of argument as a sport.

Brave new associations flow freely from the unconscious into the world of the living. Making, discovering and developing connections between and among two or more of anything is virtually automatic.

This portrait calls the ENTP the “Visionary” type. Not so sure about that, but it does contain this:

ENTPs are less interested in developing plans of actions or making decisions than they are in generating possibilities and ideas. Following through on the implementation of an idea is usually a chore to the ENTP. For some ENTPs, this results in the habit of never finishing what they start. The ENTP who has not developed their Thinking process will have problems with jumping enthusiastically from idea to idea, without following through on their plans. The ENTP needs to take care to think through their ideas fully in order to take advantage of them.

Hmmm. I guess that’s why I’m in grad school: To “develop my Thinking process.” Ben says that one of the purposes of doing a Ph.D. is to show that you can pick a project and stick to it for 2 or 3 years.

One of the reasons I’m typically dubious of personality tests like this is that I often have a hard time choosing a yes-or-no answer to the questions. I either am unable to evaluate myself in that particular way, or I don’t see it as a yes-or-no question. In the test I took, for example, there were 17 of the 72 questions that I was unsure about. That’s almost 25%. After answering them to the best of my ability, I went back and experimented with different answers to the hard questions to see how it changed the results — I even tried answering all 17 “yes”, and then all 17 “no”. It always came out ENTP. All that changed was the strength of expression for some of the traits. That’s a pretty robust result.

In-car Navigation: Another Slow A.I. Step

This wonderfully anthropomorphic NYTimes article describes the author’s experience using a talking in-car navigation system in his new Acura. This is another step in the incremental rise of A.I. It’s easy to see how Yvonne v.2.0 might seem a little smarter, and v.3.0 smarter still, each revision with more knowledge of routes and destinations, and better language capabilities.

Also interesting is how much this system resembles Manna Marshall Brain’s fictional AI-based fast-food management system that plans and makes decisions and then uses humans to carry out its actions, instructing them at every step what to do. “Yvonne” has no such user feedback system, but it would be an interesting feature to add.

Read the rest of this entry »

Cool Sunspot Pix

Doc Searls says the sunspots that caused this weekend’s solar storm are visible with the naked eye when viewed through the smoke of the southern california wildfires.

Read the rest of this entry »

Great Essay on Richard Feynman

Thinking Machines Corporation founder Danny Hillis has a fantastic essay on the great physicist Richard Feynman. Turns out Feynman actually worked for Hillis at Thinking Machines for a summer, and Hillis gives us a glimpse of the insights he got into the workings of a great mind.

Read the rest of this entry »

Sony’s QRIO humanoid robot

Sony has launched QRIO (pronounced “curio”), their humanoid robot, cousin of AIBO, the robot dog. In addition to bipedal walking, it has stereo vision and hearing, as well as (supposedly) the ability to recognize faces and voices. It’s motion capabilities look impressive. I’m very curious to see its perception and cognition abilities firsthand. While I’m sure it’s far below Cmdr. Data, I’ll bet it’s far beyond any toy anyone has ever seen before. Another step in the slow rise of A.I.

It will be interesting to see how people react to this robot given its humanoid appearance, and particularly it’s two eyes. Anthropomorphism is a powerful factor in the perception of intelligence. When I first visited the UT Intelligent Robotics Lab as a prospective Ph.D. student, they had a visual-tracking demo running on the lab’s intelligent wheelchair, Vulcan, that would follow a moving object by rotating the chair’s two “eyes”, or the entire chair if the object moved to far from the centerline. Seeing two eyes fixating, and the “body” moving face the center of attention gave an astonishing and enduring illusion of cognition, even when I knew that the behavior was driven by a simple (though well-written) reactive algorithm.

Update: See QRIO run!

Read the rest of this entry »

Posted in AI. 8 Comments »

Instant Runoff Voting

A couple days ago I mentioned the Irish ballot counting system as a more fair means of running elections. I’ve since learned that this is called Instant Runoff Voting, because if there is no majority winner, the candidate with the least ballots is eliminated and each of his ballots is recast for its next preferred candidate. Like a runoff, it ensures that the winner will have a majority of the votes, but doesn’t require a costly and annoying runoff. There are excellent arguments for IRV, well summarized in these talking points:

  • Ensures majority rule, in contrast to plurality voting.
  • Saves money compared to costly two-round runoff elections, which often have low voter turnout.
  • Increases voter turnout by giving voters better choices. Experience around the world shows that voter turnout goes up when voters have a wider range of choices.
  • Promotes positive, issue-based campaigns because candidates will seek 2nd and 3rd choice votes.
  • Creates a clearer mandate for a winning candidate’s agenda, giving better direction for policy-making.
  • Solves the problem of groupings of voters splitting their votes among similar candidates, which allows a candidate with only minority support to win.
  • Minimizes “wasted” votes, votes that don’t help elect a winner. To the fullest extent possible, your vote will contribute to electing a candidate that you like.

A very simple argument for it can be tailored specifically for Democrats or Republicans: If we had been using IRV in 2000, Al Gore would have won, since most Nader voters would have chosen Gore second. If we had been using IRV in 1992, George H.W. Bush would have won, since most Perot voters would have had Bush second. [UPDATE: According to this analysis Clinton would still have won.]

Another argument for IRV is that it will reduce political polarization. America’s current extremely polarized political climate is an emergent phenomenon rising out of the prevalence of simple plurality voting. Under plurality voting, any candidates beyond the first two are simply seen as spoliers and wasted votes, and thus they get few votes. In this kind of system, all elections are seen as contests between a favorite and a challenger. The need for campaign funding will require that all favorites and challengers be aligned with a political party. Since almost all races are effectively two-candidate races, the two most popular parties rapidly gain political currency at the expense of all the others. This causes all politics to be viewed in a false one dimensional space: left vs right. Many voters values may in fact mix values from the two dominant parties, but most of these “centrists” or “moderates” slowly get pulled to one party or the other as they see that choosing the lesser of two evils is the only way to prevent the greater evil from prevailing. Eventually everyone is either us or them.

IRV will correct this situation, because it will actually allow third party candidates to get more votes, increasing their political currency and visibility, even if they still lose. Many more people would have voted for Nader had they known that when he lost, their votes would go to Gore. Likewise for Libertarians. Then us/them campaigns must become campaigns on the issues.

Read the rest of this entry »