Tuesday, November 20, 2007

Computer Graphics and Photorealistic Human Animation in Beowulf (2007)

I just got finished watching Beowulf (Robert Zemeckis, 2007). It's an interesting film. It's completely CG (made with computer graphics), like Toy Story, but features (attempts at) photorealistic humans, like Final Fantasy: The Spirits Within. Unlike Final Fantasy, however, it attempts to model actual actors. That is, Malkovich, Hopkins, and Jolie are clearly recognizable in this film.

Zemeckis could have filmed live actors and enhanced it with CG, but decided not to. Ultimately, this was a mistake. The characters still don't quite look real enough to pull this movie off in terms of acting.

Ultimately I enjoyed the film very much, but more for it's story and monsters and graphics. It's only the acting I'm critisizing. And I'm writing a second blog post on this basic topic because ultimately all films will be completely CG. I will argue for this at the end of this post.

I have written previously on Jim's Law of Animation, which states that a film should only involve animation if it has some aspect of the otherworldly. This film qualifies, in that it has big monsters in it.

However, in part because of the uncanny valley, which I have also blogged about, the technology is still not quite ready for photorealistic acting. We saw the problem in Final Fantasy, and even in the first short of The Animatrix. Although Beowulf is getting closer, it's still not quite there. But mark my words, in the lifetime of most readers of this blog, it will be.

Let's distinguish the graphics from the animation for a moment, where the graphics refers to how good each frame of the film looks, and the animation to the apparent motion in the film. The graphics of this film are great. Oftentimes I honestly would not have been able to tell if I were looking at a real or CG person. The animation is the problem, the last step. And what I like about it is it's as much an AI problem as a graphic problem. It's the human mind and brain that controls the muscles of the face to generate facial expression, which is the core of the acting problem.

I appreciate the film as an attempt, an experiment. But it's a drama, and drama requires a belivability in the acting that photorealistic CG humans are not up for. Yet.

What about The Incredibles, you might ask. That was a great film! Yes, it was, but as explained in my uncanny valley post, Pixar deliberately does not render photorealistic humans precisely because we expect too much out of their acting. Also, I think it's important to note that The Incredibles was a comedy, which requires subtley, but more in the timing than in the genuine expression of emotion category.

That said, the film is engaging and well-directed. I particularly appreciated the silences. It's common for movies like this to be over-scored with sweeping music.

You can watch the trailer here and see if you agree with me:

Okay, so what is this business about all future films being completely CG? Before I answer, I will add a caveat: In the future, films with actual actors in them will be slightly more common than traditional cell animation is today.

The reason is cost and flexibility. Basically, when we can have CG actors that are indistinguishable from actual human beings, both in still appearance as well as acting, then it will be more cost-effective to use CG actors. It's as simple as that. Movie stars, or flesh-and-blood living ones, anyway, will cease to have a profession. They can all try to go audition for the theater.

If this sounds far-fetched, reflect on the incredible progress CG has made over the course of my (and probably your) lifetime. At some point in my childhood, the best CG was a game called "Pong." (you can play pong at http://www.csit.carleton.ca/~arya/games/pong.html)

Later came Pac-Man, Pixar films, the Titanic movie, and now Beowulf. This has all been in the span of about 20 years. You can't tell me that we won't be fooled by CG humans on screen with another 50 years of progress.

Sunday, November 18, 2007

Personal Genome Reading

Gene sequencing efficiency is increasing at an exponential rate. Every 1.9 years, the cost to sequence a base pair cuts in half (Kurzweil, 2005, p73). What does this mean for you? Well, just about everyone alive today in first-world countries, and much of the rest of the world as well, will have their own personal genome sequenced. This will allow you to know a lot about yourself, but the most important things will regard disease. You can know that you are more or less likely to get certain diseases, and eat and get screenings appropriately. See this for an example of what one might find:

As of this writing, November 18, 2007, the company 23andMe (https://www.23andme.com/) offers to sequence your genome for $1000. In two years it should cost half of that, and so on.

It can also tell you about your ancient ancestors. It's a fascinating new world we're moving into.

Pictured is an image of DNA. I wish it were rotating, but I can't find an animated one. A rotating DNA strand is a necessary part of any viewer-friendly interface involving life sciences!

Thursday, November 08, 2007

Do we understand ANYTHING about diet and health??

I've blogged before about the weak relationship between eating fat and death (http://jimdavies.blogspot.com/2006/08/ok-im-serious-this-time-now-does.html). It seemed to me that as long as you were not getting obese, you can eat however much dietary fat you want.

Well, the world just keeps getting weirder.


Turns out overweight people have a lower death rate than people with "normal" weight.

Now, I'm not a nutritionist. It just seems to me that the evidence for the relationship between diet and health is extremely tenuous. How confident can we be in the so-called facts when studies like this come out? Are we really so confident in our opinions about how diet affects health that it's rational to suffer as people do, eating the disgusting diet food they do, like whole-grain pizza crust and soy cheese?

We used to think that eating fat made you fat. Atkins changed all that.

Then we thought that eating fat made you die of heart disease. The evidence doesn't look good either.

And now it turns out that being overweight might not even make you die sooner.

And I have to eat these God-awful lean hamburgers at restaurants?

I'll give you my opinion on nutrition, for what it's worth: The relationships between diet, genes, stress, exercise, and health are so poorly understood that you should not drive yourself crazy trying to "eat right." That is to say that eating right for me might be different from eating right for you, and we have no idea how to know.

If you want to be thin to be attractive, fine. But in terms of health, yes, LOLCAT, you CAN has cheezburger.

Wednesday, November 07, 2007

Ideas Constrained by Biology

I've been reading "What's Your Dangerous Idea" (Brockman, 2007). It's
a book full of ideas from today's leading thinkers about ideas they
have that might have dangerous repercussions for society if true
and/or accepted. It's pretty interesting.

One of the ideas I came across (Trehub, 2007) is this: since all
concepts and thoughts are in brains, then our very ideas are limited
by biology. What an interesting idea!

On the one hand, this sounds incorrect. Or true but irrelevant. Our
brains are in a sense general-purpose machines, capabale of creating
new symbols and combining symbols in an infinite number of
ways. Saying that our thoughts are limited to biology is like saying
that a computer screen is limited in the kinds of pictures it can
produce. Well yeah, but it's not a limiting factor.

But on the other hand, we are limited in certain
ways. Certainly we can only imagine ideas up to a certain level of
complexity. The proof for the four-color problem was solved with the
help of computers and no human has ever checked it completely. We know
the theorem proved is true but we really don't know why (Strogatz,
2007). It's safe to say that the proof is beyond human
understanding. If this is not a limit of our biology, then what is it
a limit of?

The incredibly fascinating and hopeful book by Ray Kurzweil (Kurzweil,
2005) offers a possible solution to this in our future. According to
Kurzweil, by 2040 our minds and brains will be unrecognizably enhanced by
information technology, allowing us to understand things current
humans are completely incapable of imagining. Further, we are creating
AIs that will be able to undersand things for us.

If you doubt this, reflect on our understanding of the solution to the
four color problem.


Brockman, J. (Ed.) (2007) What's Your Dangerous Idea?. Harper

Kurzweil, R. (2005). The Singularity is near: When humans transcend
biology. New York: Viking.

Strogatz, S. (2007). The end of insight. In
Brockman, J. (Ed.) What's Your Dangerous Idea?. Harper
Perennial. pp130--131.

Trehub, A. (2007). Modern science is a product of biology. In
Brockman, J. (Ed.) What's Your Dangerous Idea?. Harper
Perennial. p234.