Month: August 2023

Today I sleep.

There was a lot of excitement at Siggraph this week, but a surprising lack of vision into the future. I guess that makes sense for a technical conference.

For the most part, people were focused on the next thing, whatever that may be. So they weren’t generally thinking about what computer graphics might be like in another ten years.

Well, ten years from now computers will be a hundred times faster than they are now. So things are indeed going to be qualitatively different.

Computer graphic imagery (CGI) will be completely integrated into our everyday life, as wearables become cheap and ubiquitous. The real and the virtual will be seamlessly intermixed to the point where the distinction between the two will start to become meaningless.

CGI will be continually created by generative AI in response to our casual conversation and gesture. And we won’t even think about it, because it will all just be normal.

But at Siggraph this year, they weren’t really talking about any of that.

As I was saying, one experience yesterday morning at Siggraph jumped out at me. At the nVidia booth, they were showing how they can use machine learning to turn a single photograph into a 3D model. The interesting part was that the process takes three seconds.

My first thought, after all this week’s talk about Moore’s Law, is that in another ten years this process will take one hundredth as long. This is because every ten years, computers become 100 times faster.

In other words, by 2033, we will be able to turn a single photograph into a 3D model in real time. At that point it will seem instantaneous.

Like Turner Whitted said, certain effects go beyond the quantitative, and become qualitative. When we can create 3D worlds instantaneously from single images, that will be a qualitative change in our ability to manipulate reality.

This was the 50th anniversary of the conference. And therefore there was a lot of talk about Moore’s Law.

Roughly speaking, every 10 years computers get about 100 times faster. Which means that after 50 years, computers have gotten about 100 trillion times faster.

Turner Whitted made the observation in a session this week that we shouldn’t be focusing on the quantitative effect of Moore’s Law. Instead, we should be focusing on the qualitative effect.

Just morning I had an experience that reaffirms this. More tomorrow.

Second day of the conference, and — not surprisingly — far too many things to be able to wrap my mind around it all.

But one thing is clear: The people who have been coming here the longest have the best insights about the future. I suspect that this is at least partly because it takes a few decades or more of living with Moore’s Law to be able to use it predictively.

Completely wonderful, and completely overwhelming.

It’s going to take me some time to process.

Today I read a fascinating article. It seems that the standard process for creating nanosensors, which can electronically detect the presence of many different types of materials, involves creating connections between extremely thin layers of silicon together.

This is done by “curing” them — subjecting them to very high temperatures for 12 hours. Which is an extremely expensive process.

Apparently, while cleaning one of these sensors prior to curing, a researcher accidentally spilled a little ethanol — that’s regular drinking alcohol — on one of them. And the sensor started performing better than any of the cured sensors.

The research team then figured out that adding just the right amount of ethanol to uncured sensors– not too much, not too little — resulted in sensors which were more effective than the cured sensors method. And also a lot cheaper to make.

So apparently if you get these sensors a little drunk — but not too drunk — then they become very sensitive.

Which is a principle that Chinese poets knew centuries ago.

If I had an idea many years ago
That was just sitting up on a shelf
And today if I take it and actually make it
Is that stealing ideas from myself?

Maybe it just wasn’t ready back then
Or it felt like an end in itself
But now I can use it. Is it kosher to choose it?
Or am I stealing ideas from myself?

I have been playing with the video pass-through mode of the Quest Pro. And one thing has started becoming clear to me.

Namely, that there is something very compelling about miniature virtual worlds that are within reach of your hands. Not that there’s anything wrong with immersion in full-size worlds.

But there’s something particularly charming and compelling about miniature worlds. I think it’s because we have a more intimate relationship with things that we can pick up and touch within our own personal space.

I predict that as mixed reality eyewear becomes something common and everyday, a lot of our virtual interaction will start moving to tabletops. Things like schedules and maps and organization charts are all going to begin showing up as three-dimensional structures that we can reach out and point to.

These will be things that you and I will simply expect to see on the table between us as we talk to each other in casual conversation. And of course we will eventually forget that it was ever any other way.

I’ve recently noticed an important strategy in my programming. It’s something I’ve always done, but until recently I wasn’t conscious of it.

This strategy could be called “finding the simplest thing”. When I need to extend the functionality of one of my interactive software systems, I look for the absolutely simplest possible case.

The idea is that if I can initially get the simplest possible thing working, then I can add fancier things later on. Because if I were to instead start out by trying something more ambitious, I might end up spending needless hours debugging.

So if I’m extending something from two dimensions to three, or adding the ability to respond to a mouse events, I look for the absolute minimal way to do that. I won’t draw two lines if one will do. I won’t respond to mouse drags if I can get away with responding only to mouse clicks.

This strategy sort of reminds me of something I learned in school about how people used to build bridges. First they would fly a kite and wait for the wind to blow it across the river.

Once they had a string going from one side to the other, they would use that to pull across a heavier line. And then use that second line to draw across a still heavier line.

Eventually they would have a thick rope going across, strong enough to support the weight of people. And then they could begin in earnest to build the bridge itself.