Thursday, July 28, 2005

Making Women, Making Robots, Making People

Auntie has a story about a Japanese team that has constructed a realistic looking female android (via /.). This looks like the long expected convergence between Sony's Aibo and the uber creepy Real Doll. (link not safe for work.) If you want to see a real shitty attitude to women in action, check out the slashdot discussion.

Discussion question: Why did the engineers first make an android that looks like a 5 year old girl, and then make an android that looks like an attractive woman?

I have been playing around with lego robots, in an effort to develop some good classroom demonstrations to accompany Leiber's *Can Animals and Machines Be Persons.* (I've mentioned this before.)


My main goal is to stop students from saying "Computers can only do what you program them to do" by demonstrating a wide variety of machine learning, starting with robots simply recording information about the world around them, moving on to hypothesis testing and Bayesian algorithms, and finally full fledged evolutionary algorithms. I'm spending two weeks on the dialogue, so I'm not aiming to make them proficient in any kind of programming. I just want them to have a sense of what AIs can do, a vague sense how they do it, and how it relates to what we do.

And I want them to stop saying "Computers can only do what you program them to do." Sometimes I think that I will be a successful teacher simply if I can disabuse my students of a few simple statements that they repeat like they are truisms, when really they are false. Some other examples:

* A theory is an unconfirmed fact (corollary: evolution is just a theory)
* The only one-hundred percent effective way to prevent pregnancy is through abstinence (Yes, but the failure rate of vows of abstinence is between 61% and 79%)
* Animals can only act from instinct.

Some other persistent myths that I deal with in talking politics, but I don't have to deal with in teaching:

* Anarchists advocate chaos and everyone doing what they want.
* Pacifists advocate letting people walk all over you.

Creating AIs poses all sorts of ethical difficulties. Most of which are right at the surface in the decision by Hiroshi Ishiguro and colleagues to make an attractive female android. Unfortunately, we can't even begin to talk about these issues until we acknowledge the possibility that robots may one day deserve moral status.

The flipside of granting machines moral status is holding them morally responsible. In 1998 I met a woman named Cari who was working on a dissertation on how we can tell when to hold a machine responsible for its actions. (Christ, was that 7 years ago?) Google lets me know that she finished her dissertation, and has been publishing. I'm adding her stuff to the list of resources for the unit in the course.

I'm also going to use the Beeb story and my discussion question in the course next term, I think.

No comments: