Monday, May 28, 2007

Pediophobia and the Uncanny Valley


The "uncanny valley" is one of my favorite concepts. It describes humanoid objects. As things look more and more human, we are attracted to them more and more. But there is a limit-- at some point they look real enough to look like human bodies, but the way they act is not realistic enough for them to look like they are healthy human beings. The attraction turns to disgust and revoltion. Then, as the creatures get more humanlike, the attraction rises again. Graphing this, with human-likeness on the x axis and attraction on the y, we see a dip just before realistic human behavior. This dip is called
uncanny valley.

Examples of things that are, for many people, in the uncanny valley
include


  • Wax statues of human beings. They look real, but are not moving
    at all, so they look dead.

  • Realistic characters in video games and computer-generated
    movies.
    The facial expressions are not quite right. The way they
    move is not quite right. They look sick, or like walking, talking
    corpses.

  • Mimes. People doing pantomime train to move their bodies
    in unusual ways. This can look kind of eerie. This effect is
    exacerbated with white face paint (which is part of why people fear
    clowns, I expect). Doing "the robot" can be
    frightening for the same reason. What's wrong with these people?

  • Puppets. They sometimes have enough superficial
    similarities to human beings to make us expect human behavior, but
    then they don't come through. Note the use of scary puppets and
    dolls in the Chuckie movies like Child's Play and the clown
    in Poltergeist.

It was first described by Masahiro Mori.


Pixar deliberately tries to make their characters appear cartoon-like, specifically because they want to avoid the uncanney valley. Note that Geri in Geri's Game is like a caricature of an old man. Compare this to the zombies trying to act in the Final Fantasy: The Spirits Within movie. It's hard to see the valley without these things in motion.



Cynthia Braezeal's "Kismet" project tries to do the same thing. By making her robot look like a cute toy, rather than a human being, she avoids the valley.

One man is trying to build robots that are super-realistic. He's trying to make robotic heads that can emote so well that it looks real-- it's trying to cross uncanney valley! Nobody else is trying to do this in robotics. This guy believes the uncanny valley is pseudoscience: http://iiae.utdallas.edu/news/pop_science.html
and there's an impressive video here:
http://iiae.utdallas.edu/projects/hanson.html
Though this one looks much creepier:
(Thanks to Daniel Saunders for pointing me to these things.)

I read about an exhibit recently that featured people's fears. They had people write in fears. The artists categorized them and made different rooms based on the fears. They grouped all fears of "false representation of sentient beings" together under the name "pediophobia" (Schulman, 2006). I think they might have made up this word, since I can't find anything about it on the internet. But it's basically an extreme fear of the uncanny valley.

Schullman, M. (2006). Worst Nightmares. The New Yorker, October 30, p38.

Tuesday, May 22, 2007

"Pug Hunt" Video Online at Google Video



I have put a funny documentary online in which Montica Pes and I interview pugs and their owners at the 2004 Pugfest in Lawrenceville, Georgia. It's called Pug Hunt, and it's about 15 minutes long.

http://video.google.ca/videoplay?docid=-7808576521830907638

Bahama Lynch did the camera work and helped write the pug hunt song, which I sing to pugs in front of their owners on the video:

Goin' on a pug hunt:
Chop, chop, chop.
Gonna find a pug and
lop, lop, lop.
Find a little pug and
snap, snap, snap.
Grab him by the head and
break his neck.

Monday, May 21, 2007

How to Break Up Your Day


I'm a professor, and much of my time is unstructured. I have a lot to do, but aside from the occasional meeting or class I have to teach, my schedule is open. Which means I have to make decisions about how to spend my time. I have this and that project, but I find myself
thinking about how I can best break up the day at a high-level. Some examples will help clarify.

By Project. I use the method in the book Getting Things Done, by David Allen, and I recommend you do too. Anyway, anything
you want to change about the world is a "project." Writing a paper, getting reimbursed for a conference, doing your taxes, raking the
lawn, etc., are all projects. At school I have several projects. One way to break up the day is to look at the next to-do items on each project, and using the urgency, schedule hours in your day that reflect this. One hour reading for this paper, one hour grading tests, etc.

By Practicality. I've heard it said that a scientist should have "one hand in the theoretical, one hand in the applied." One could break up the day spending some specified percentage of time working on applications and some percentage working on theoretical stuff. I have some problems with this conception however. One is that I most scientists are often funded through public funds, and I don't think taxpayer money should be spent on things that private businesses can take care of. The public money should be maximized to do the long-term
basic research that business is, often necessarily, too short-sighted to do. The other is that I feel basic (unapplied) science is more important, but that's an essay for another day. Suffice it to say that a scientist should have "one hand in the theoretical, one hand in the theoretical." Or better yet...

By Risk. "One hand in the safe, one hand in the risky." In science there are relatively "sure things" or "low-hanging fruit" that will yield incremental results. What I mean by this is that the outcome of the scientific work is not a breakthrough; it adds a bit to what we know but is not revolutionary or, perhaps, even surprising. Some of your scientific time can be spent on this stuff: it's particularly good for your career, which you want to keep to be able to do the risky work. Risky work might not work out-- that is, you have a theory that is probably wrong but, if right, would be a very big deal. Personally, I think the scientists I know should be trying to do more risky work, but in any case, a scientist can decide to break up his or her day according to which projects are safe and which are risky.

By Task. Yet another way to break up the day is by task type. For example, spend some of each day reading, programming, writing, etc. This is a method I've used at points where there are task types that I really didn't like doing, such as reading hard papers or programming.

Can anyone think of other ways to break up the day? I'd love to hear about them, either by email (jim@jimdavies.org) or in the comments.

Unlike many of my blog posts, I don't have a firm opinion on this issue. I see value to several, and I can't totally choose one, but at the same time I can't see how to combine them in an effective, practical way. Suggestions?
Photo: This is a fire hydrant outside my house. In Ottawa we have these yellow plastic things rising a few feet above each hydrant so that they can be found when the hydrant is submerged in snow.

Monday, May 14, 2007

A Better Way To Format Text: Live Ink




This article is very interesting. A company made a program that can format text into meaningful chunks. "Early results have been encouraging. According to Walker, a study funded by the U.S. Department of Education found that students who read text books in Live Ink are adding 10-15 percentile points on nationally standardized reading tests."

You can click on the picture above and see what it looks like.

Sunday, May 13, 2007

Cell Phones In Cars


Cell phones are dangerous to use in the car, right? Right. Studies show that you're more distracted and have slower response times when on the phone in a car. About 1500 people die per year because they were on a cell phone.


Cell phones are less dangerous when they are hands free, right? No.Studies show they are equally distracting. This is because the impairment is due to your attention being tied up, not your hands.


Isn't talking on the cell phone in the car is just as dangerous as talking to a passenger in the car? No, because the passenger in the car is in your context. You will notice that when traffic gets crazy,or someone runs in front of the car, your conversation partner, if she's in the car, will stop talking. She knows you need the attention to drive. On the other hand, someone on a cell phone demands answers from you as though you were sitting at home. They don't stop talking no matter what happens in the car.


If cell phones are dangerous to use in the car, then this practice should be outlawed, right? Not necessarily. There is a gain in productivity by being able to talk on the phone. To give a simple example, there have been many occasions where I got directions to a place, while I was driving there, on the cell phone. If I had to pull over every time I wanted to talk, I would have wasted several minutes.


Can you imagine if everyone on the highway tried to pull over to the shoulder, stop, and turn their flashers on every time their celly rang? Fully eight percent of people driving are yakking on the cellphone. 60% of minutes sold are used in cars. The shoulders would be full, and we might even have more accidents with all the people rushing to pull over.


We spend a lot of time in the car, and lot of time on the phone. Most of the time, driving doesn't require all that much attention, but we save a lot of time by talking on the phone in the car. I saw a talk by an expert in this field, John D. Lee, and he said that the analyses done show that with the increase in accidents and the increase of productivity, the overall good ends up being about the same whether or not you talk on the phone in the car (Josh Cohen at the Harvard Center for Public Policy apparently showed this, but I have not tracked down the reference yet).


Interestingly, the debates regarding whether cell phones should be legal to use in the car mirror those debates regarding whether or not radios should be legal in cars. Turns out that listening to the radio is almost as dangerous as talking on the cell phone. About a thousand people die per year because they were listening to the radio in the car. If you're listening to a fascinating radio program, just like a conversation partner on the cell phone, the voice won't shut up just because you're hydroplaning. You can't even say "hold on" and get them to stop talking. You have to turn the radio off and miss stuff.


It's not all gloom and doom-- cell phones and radios also prevent some accidents by keeping people awake. It's also been shown that drivers drive more slowly and stay more distant from other cars when on cellphones, probably due to risk compensation.


Pictured is a piece of graffiti from my neighborhood, centretown Ottawa. This artist writes "no air" all over the place.

Wednesday, May 09, 2007

Deep Blue Day 10 Year Anniversary Saturday, May 12, 2007


Every year on May 12 I celebrate a milestone in AI: Deep Blue's defeat of Garry Kasparov in 1997. Pictured is a photo of last year's celebration, during which Daniel Saunders and I dressed in blue and watched the documentary Game Over. We were rooting for the AI.

And this year it's the 10 year anniversary, so I'm going to live it up, party, and wear blue!

For those of you who don't know what I'm talking about: Deep Blue is a chess-playing computer and computer program created by IBM. It's name is a combination of "Big Blue," the nickname for IBM, and "Deep Thought," it's predecessor, named after the fictional computer in the Hitchhiker's guide books. It defeated world chess champion Garry Kasparov. As you can see in the film Game Over, Kasparov was a sore loser, accusing the other team of cheating!

There are those who say that Deep Blue is not AI. People who say this tend to have a mystical view of intelligence. If they can understand how a program works, then it just can't be real intelligence. It's "just planning" or "just search." This is called the "AI Effect."In response to people like this AI researchers have joked that AI stands for "Almost Implemented."

But the facts are that Deep Blue works by using algorithms and principles created by AI researchers, for AI reasons. Many of these algorithms were invented long ago-- only in 1997 were the computers fast enough to make them effective enough to beat the world champion. To me, this is enough to classify Deep Blue firmly as an AI program.

Sunday, May 06, 2007

What I'd Like To See in Electronic Calendars


Like many people I use an electronic medium for my calendar. I use a
program to keep track of my days. I use the Palm Desktop (pictured), which
synchronizes with my PDA, the Palm Treo. But it's problematic, like all calendars I've seen, paper included.

I would like to put things on there and somehow be able to see that I do not plan to do them. For example, sometimes there will be a talk I'd like to go to, but I can't go because I have a meeting. I want to be able to put it in my calendar anyway because the meeting might be
cancelled. As it is now, I have to look and remember which one is more
important to me, or else not put it in my calendar at all (and miss it
when the meeting cancels).

I'd also like to put in there other people's schedules. If a good
friend is at a conference, I want to know it, but I don't want it as
prominent in the display as the things I'm doing. As it is now, I
usually put in a note on the date of return that the friend is
returning, but I can't look at an arbitrary day and know instantly if
my friend is around.

Finally, I want to be able to put in my half-hour by half-hour
plans. I break up my day into half-hour chunks, and I do it on paper
because it would positively clog up my calendar if I put it in
there. Every day I break up my day into half-hour long project
sessions, but if I put them in the calendar, when I'd search for
something, if it happened to use a keyword from something I do daily,
ALL those would return as search results, which would be
overwhelming. I'd never find what I was looking for.

How could these needs be implemented? I'm not an HCI (Human-Computer
Interaction) but what I imagine is you can have calendar events that
are hazy, behind yours, and different colors. You can focus on these,
but they are not intrusive most of the time. Daily grind stuff,
described in the last paragraph, might only be foregrounded on the
current day.

Tuesday, May 01, 2007

Simple Shapes Over Time




I've recently joined a writing group here in Ottawa. It's a bunch of
English majors at Carleton who sit around, read each other's stuff,
and comment. I've been in writing clubs before, but I've never had
anyone bring in poetry. These people love poetry. I thought I wouldn't
have much to say, but it turns out I do, and I find these sessions
very rewarding. The interesting thing is that I find it more rewarding
giving feedback than getting. Anyway, this post is about
something I think I've discovered about time-based art as a result of
the comments I've made in this group.

I found that a great deal of my comments were along the following
lines: If you're going to do this kind of thing here, you need to be
consistent and do it throughout the work. Else it looks like a
mistake, or looks out of place, etc. Examples:

  • writer made up one word in a poem
  • writer has only one funny line
  • writer has only one image of gritty realism


I thought about this, and came up with a general rule, that in short
works conventions you set up should be used consistently
throughout. But then I thought this wasn't quite right-- you can also
have build. I came up with the five simple shapes that I think
would be acceptable. They are depicted in the diagram.

So, for ascending there must be more and more of something as
the work progresses. e.g.



dishwater

pancakes

strawberries

powdered sugar

cocaine



Now this scarcely deserves the title poem, but it's simple enough for
this analysis.
What's happening in the above set of words is an ascending trend of
phenomological wonderfulness. It grows, so it kind of
works. "Cocaine," however, is qualitatively different from the other
things in the list, in my culture anyway, so it kind of breaks the
flow. It's so jolting that it's funny. That's okay, because the first
and last parts of something this short can have a special
status. Shakespeare's acts ended with rhymes, as did his chosen form
of poetry, the sonnet. It somehow seems appropriate for the end to
havce something different. If there were only one rhyme in the middle
it would look weirder.

There are other aspects to this poem, though. Each line is
consistently long. That is, the aspect of line length follows one of
the simple shapes. We could play with this:


several sinks of cold, dirty dishwater with lots of egg pans in it

pancakes with syrup, stacked a mile high

a big bowl of strawberries

powdered sugar

cocaine



Here the poem is similer but the line length has been turned from
consistent to descending. We could do the U-shape:


several sinks of cold, dirty dishwater with lots of egg pans in it

pancakes with syrup, stacked a high

strawberries

two pounds of white powdered sugar

six perfectly sized lines of the greatest, uncut, columbian cocaine



Notice this one works less well. The impact of the "cocaine" at the
end is damaged by the U-shape, which emphasizes "strawberries."

The u-shape is a little complicated for a poem this short.
I am reminded of the discussion in Hofstadter's Metamagical
Themas
in which he talks about the grain size of a work of art. He
says the title of a novel is perhaps the most difficult thing to
translate, because you have so much meaning to transfer in so few
words. For example, translating the title of the novel All the
President's Men
into French is difficult because it's a play on
the Humpty Dumpty rhyme, which the French might not have.

Similarly, if you have only a few pixels, it's difficult to represent
a circle. With thousands of pixels, a circle is clearly
recognizable. But if you're down to around 16, it looks, well,
pixelated. Similarly with short works I think the shapes in the
diagram are the only ones that will work because of the grain size.

Novels and feature-length films are another beast entirely. They are
long enough such that you can have multiple climaxes in the same
work. Each scene can have one of the simple shapes above, and there
might be several scenes per chapter.

In conclusion, I think if you're doing short works in any medium
(music, theatrical improvisation, short film, poetry, short stories,
etc.) you should be aware of all aspects of the work (tone, humor,
imagery, tempo, excitement, etc.) and make sure each one fits one of
the simple shapes.