Ken's Blog

I often do the Spelling Bee puzzle in the New York Times. It’s nice way to exercise my brain, while pleasantly filling the time when I’m waiting for this or that.

When you do a puzzle like that, which requires you to find words, you quickly realize how many perfectly plausible words don’t exist. Many combinations of letters follow all of the proper rules of spelling, but they simply aren’t in the language.

I just today’s puzzle, for example, I came up with brini, intort, thiron and hibortion, all of them perfectly pleasant sounding, and none of them in the dictionary.

I wonder whether we could create a dictionary of alternate words. We could probably make it happen if we all worked together.

People could suggest meanings for various alternate words and then we could all vote on the most popular meaning. If nothing else it would be a fun activity.

As far as I know, there is no name for such a project. Maybe we should come up with a word for it.

I spent the last week hanging out with old friends at the ACM/Siggraph conference, seeing technical papers, and generally immersing myself. And I started to better understand the nature of the bond that we computer graphics people share with one another.

We are people who have a passion for math, and also a passion for art and visual expression. And we recognize other people who share those two passions, which in our minds are inextricably linked.

We don’t necessarily love computers — they are just a means to an end. The computers are basically plumbing. You don’t love your house because it has plumbing, but you definitely want it to have working sinks and toilets.

There is a particular kind of shared language that people have when they love both math and visual expression, and are excited about creating things that embody both of those loves. Siggraph is the best place to find those people, and to start a wonderful conversation with them.

For me, some of those wonderful conversations have now been going on for decades.

I admit that can be hard to focus the delightful ideas that I encountered at Siggraph. Alas, on the national stage an old jackass is trying to start a race war.

I wonder who he is going to go after next. The Italians? The Irish? Probably not the Germans.

I wouldn’t put it past him to start attacking left handed people, if he thought it would inflame his base. After all, most people who would vote for him are right handed. And that may be as near to logical thinking as we are going to get from this moron.

In any case, I will try to put this ugliness out of my mind and circle back tomorrow to discussing the wonderful things I saw at Siggraph.

Siggraph just ended, and now my head is full of a million ideas. I need some time to digest.

I am sure it will all make sense in another day or two. Now I need to get some sleep.

Continuing from yesterday.

The other side of the coin is that What you get for all of this limitless power is not at all the same thing as what you get by doing it the old-fashioned way.

There is no actual human intelligence inside that magic box. Yes, it can endlessly reconstitute human knowledge and know-how, but it does so with zero judgment.

This is all perfectly fine when simulating tornadoes. But it’s something else altogether when simulating a person that you would like to get help from.

In the latter case, you get what you pay for. Free advice from a realistic but judgment-free bot is one thing. The actual thoughts of a caring human being are another thing entirely.

What I realized, after thinking about the statement “AI doesn’t care where it is trained,” is that modern generative AI is essentially a kind of magic box to store up know-how. It can take a tremendous amount of energy to get know-how into that box. But once that has been done, new applications of that know-how can emerge from that box an infinite number of times — all at low cost.

Another time that this kind of thing happened was the creation between the late 1970s and the mid-1980s of desktop publishing. Before then, if you wanted to produce a document with high quality type-setting, you needed to hire a professional to do it for you.

But starting with TeX in 1978, and software that followed, the knowledge of experts could be encapsulated in software that you could run at home on your personal computer. Essentially, that was the beginning of “expertise in a box.”

We are now entering an era in which all sorts of expertise can be stuffed into that box, given enough prior examples, after which it can be cheaply accessed anywhere and everywhere.

More tomorrow.

Today at Siggraph, Jensen Huang, during his keynote address, said “AI doesn’t care where it is trained.”

And all of a sudden I got it, what generative AI really is about, why it is actually eco-friendly, and so much more.

I’m still working it through in my head. More tomorrow.

I am at the Denver Convention Center, sitting the very from row. About 2000 of us are about to experience the Siggraph Fast Forward.

Everyone who is presenting a technical paper this year will be giving a short synopsis of their work. There are several hundred papers this year, so I think each presenter will be getting less than half a minute.

I was a participant in the very first ever Siggraph Fast Forward in 2002. So I know how thrilling and nerve wracking this brief but critical live performance will be for this year’s presenters.

I look forward to the next few hours of having my head filled with wonderful new ideas. Wishing everyone good luck!!!

For some reason I have a new-found fondness for coconut trees.

I realize now, more than ever, that I exist in the context of all in which I live and what came before me.

Today I showed a demo to somebody, and right afterward, I realized things that I wanted to improve. Some of this was due to the reaction of the person I was demoing to, and some of it came out of thinking “Gosh, I wish I also could have shown …”

And I realized, for the first time, that all of my years of showing demos has not been just to show demos. It has also been to learn from the experience. And then to use that knowledge to make the demo better.

It turns out that giving demos is an iterative experience: Learn from doing it, make it better, repeat.

Which means, in essence, that a good demo is something that is organically grown, not from one person, but from an entire community. I kind of like that.