Ken's Blog

I attended a technical talk today that I enjoyed very much. The speaker was showing a scientific result that was very practical and useful.

Yet at the same time, the mathematical theory underlying the technique was also very beautiful and elegant. And that is not usually the case.

Most of the time, people invent very practical things that are not at all formally elegant. The invention gets the job done, but in a way that is very messy and unaesthetic.

But every once in a while I encounter a beautiful and simple theoretical framework that also produces a truly practical result. And when that happens, it takes my breath away.

And it reminds me, all over gain, why I love math so much.

I was on a panel yesterday where the topic concerned the future of communications technology. At some point, someone in the audience asked a very interesting question.

“Do the panelists think,” he asked (I’m paraphrasing), “that all of this advancement in communication technology is going to make society better? In what ways might it actually make things worse?”

I replied with the first thought that entered my head. “Societies have always been dysfunctional,” I said. “They’re just getting better at it.”

I am currently reading about dinosaurs. And I find myself thinking about the fact that they managed to stick around for more than 150 million years.

In contrast, we humans have been here for a mere blip of time. We pretty much just arrived here on planet Earth, and there is no particular guarantee that we will be here much longer.

Of course dinosaurs were not just one species — they were many species. If you take the long view, there is a reasonable chance that humans will give way to other human-like species, and then still other species after that.

All of those species will likely look back on us as a fascinating experiment, a necessary if failed step on the ladder of evolution. We will be seen as exhibiting the beginnings of higher intelligence, albeit in primitive form.

The species that will evolve from us, the ones that manage to last longer than a mere few hundred thousand years, will likely have traits that we lack. For one thing, there is a reasonable chance that they will have a very different relationship to their emotions.

An intelligent species that is capable of lasting millions of years will probably not be prone to sudden bouts of rage the way so many humans are. They will not find it so difficult to value members of their own species who seem different.

Unlike us, they will likely not lose their ability to think rationally when experiencing sexual attraction. And they will almost certainly have a much greater respect for the health of the planet that sustains them.

All of this is conjecture of course. But who knows — maybe one day, many millennia from now, these words might be read by a being far more intelligent than me.

And with any luck they will think “Well, he got some of it right.”

This past weekend somebody explained to me the concept of ego death. It is usually defined in the context of mythic stories or psychedelic drug use.

But it occurred to me that the concept fits neatly into the difference between lean-forward entertainment experiences, such as game play, and lean-back entertainment experiences, such as being told a story. The former preserves and even enhances the presence of the ego, whereas the latter temporarily annihilates it.

Our sense of self is in the very center of game-play. In a way, it is all about us. We place ourselves at risk in a safe context, and thereby learn our strengths and our limits.

But when being told a story, our self disappears. We are not in the story, and (other than in experimental works) we as individuals are not called out or referenced. It is only afterward that we apply our individual self to analysis of the narrative that we have experienced.

In a sense, good storytelling is all about ego death.

Following my recent blog post about being able to use A.I. to synthesize actor performances in the future, what would come after that? I think a next logical step would be to create actor mixes.

For example, for a particular role, we could start with a base of Humphrey Bogart, blend in some Jimmy Stewart, then add in a dash of Brad Pitt. And maybe just a hint of Bette Davis to top it all off.

To be clear, this is all we could do, given our current approach to A.I. Convolutional neural nets are only able to scrape existing data, and then find new points in the space created by that data.

What makes the original Humphrey Bogart a genius will continue to remain mysterious and unfathomable. And all we will be able to do is tap into that genius.

The creation of a new actor who goes in a completely different direction — a true original, will remain beyond reach. A.I. will not create the next Merrill Streep or Christopher Walken or Cate Blanchett.

But will, eventually, be able to effectively place mixtures of those actors into new scenes and new situations. And that limitation is important to understand.

Because it would be a mistake to believe A.I. can replace great actors. On the contrary, future A.I. actors will require the performances of the originals. Without them, it would have nothing to work with.

I am attending a lovely wedding this weekend. Lots of hellos, lots of food, lots of little kids running around.

But the main event, of course, is the ceremony. The world stops, everyone goes quiet, and a union is solemnly joined

Physically, the world is the same afterward, but on a human level everything is different. A mysterious door has been opened, entered, traversed.

It is one of the defining joys of the human condition. We weave together a consensus within our collective minds. And explicitly, beautifully, it becomes reality.

It's swell to make a poem rhyme
But often when I have more time

I like to add an odd constraint
Somewhat weird and maybe quaint

Perhaps I'll make a poem's form
Something not quite in the norm

Today then just for fun I might
Make lines line up on the right

But is it art? Hmm, not so easy
The very notion makes me queasy

Some notions can be good to try
Yet still quite hard to justify

And everything can go all wrong
If the last line ends up running long

When you are on the phone with an airline company, or any other outfit that requires confirmation codes, sometimes you need to spell out a six letter nonsense word, like DMXRFU or ZYTPSW. There is no great way to do this, so most of us resort to saying things like “D as in Denver, M as in mammoth, …”

But sometimes this doesn’t work as planned. A friend told me recently that she was on the phone with an airline representative who was trying to give helpful hints.

The representative explained how you could choose words that made the process easier. “For example,” she said, “you could say Q as in ‘cucumber’.”

My friend was too polite to correct the nice lady on the phone, but hearing that story got me thinking. What are the various ways that people might screw this up?

So here are some helpful suggestions for how one might sow chaos and confusion, in the course of trying to helpfully spell out the letters of those mysterious six letter codes:

        A as in eight
        C as in sea lion
        E as in Igor
        G as in jeep
        I as in eyeball
        K as in candy
        Q as in cucumber
        S as in ace
        U as in pew
        X as in accent
        Y as in wine
        Z as in peas

This morning I spent a lot of time in my computer graphics class talking about Bela Lugosi. The reason is that I was teaching character animation.

As an example of beautiful character animation, I showed Chernabog, the dark god of Bald Mountain, from the 1940 Disney film Fantasia. That character was animated by the great Bill Tytla.

I had learned that Tytla had based his animation of Chernabog on the movements of Bela Lugosi in his role as Dracula. In that film, Lugosi moves in a very specific and hypnotic way — which you already know if you’ve seen his mesmerizing performance.

It turns out that we can approximate some of that style algorithmically in an animated character. One trick is to use sine functions with progressive offset phases for the movement of the character’s arms, from shoulders down to fingertips.

The lecture went well. I think the students really appreciated the beautiful clip from Fantasia that we watched in class.

But then — just now in fact — I found out that today is Bela Lugosi’s birthday. As I said in another recent post, I would like to think that somehow, in some way that I don’t yet fully understand, the Universe is speaking to me.

A lot of my communication with colleagues is through diagrams. A well made diagram can convey a lot of useful information that is difficult to get across in text.

One of the cool things about diagrams is that you can use them to subtly express things that are hard to say in words. Your choice of font, of color, how you group things or align them, a judicious use of arrows, all of these add to your story.

It’s not just that you use diagrams to show all the parts of whatever you are talking about. You are also using them to show which relationships are really important, and to get across the proper sense of hierarchy and flow.

It’s still amazing to me that in school they don’t teach you how to make good diagrams. Some people manage to pick it up on their own, and it’s a shame that those people need to be largely self-taught.

Why don’t we have proper diagram-creation literacy in our education system? I don’t have a good answer to that question. Do you?