Walt Bransford, SIGGRAPH 98 Conference Chair:
It is an honor, indeed, to introduce Jim Blinn, from Microsoft Research,
as the SIGGRAPH 98 Keynote Speaker. For almost three decades, Jim has coupled
his scientific knowledge and his artistic abilities in the advancement of
computer graphics. His many contributions include
- the Voyager flyby animations for space missions to Jupiter, Saturn,
and Uranus
- the animation of Carl Sagan's PBS series, Cosmos
- the PBS education series, The Mechanical Universe and Project:
Mathematics!
Jim has been to every SIGGRAPH conference. He is the first recipient
of the Computer Graphics Achievement Award. He developed many fundamental
techniques, including bump mapping, environment mapping, and blobby modeling.
As Graphics Fellow at Microsoft Research, he is investigating the mathematical
underpinnings of image rendering.
Now, there are a lot of Jim Blinn stories, and I'll tell you what mine
is. It was at a session in the 1980's when he was explaining this bug that
had ruined one frame in an animation sequence. So he showed us how to fix
the bug by holding up the film and cutting out the frame with a pair of
scissors. And then that's when I knew, well this is the group for me!
Please welcome, Jim Blinn.
[ applause]
Jim Blinn, Keynote Speaker:
So, 25 years.
Since I've been to all 25 of these things, they seem to think that I
would have something useful to say at the keynote address here. So, in trying
to figure out what I was going to say, one of the things that occurred to
me was what my friend Alvy Ray Smith always complained about people saying
in situations like this, and I resolved that I'm not going to say that.
[ laughter]
If you want to know what that is, you're going to have to ask him.
I'm not the only 25-year person here, although there is not an official
record kept. I understand that Lou Katz and Jim Foley have also been to
all 25 of these. Tom Wright was a member of the club for a while, but I'm
not sure if he's still here. And there might be others that I don't even
know about, and if so, let me know afterwards and I'll apologize for not
introducing you at this opening.
Anyway, so what am I going to say? Well, in the first place, to believe
that I actually went to all 25 of these things, probably I should go back
and give kind of my little historical retrospective of all the conferences.
This is decorated by scans of the covers of probably one of the very few
complete collections of SIGGRAPH conference proceedings.
1974, in Boulder, it was held at University of Colorado. We all stayed
in college dormitories instead of luxury hotels, and we had assigned roommates.
I don't remember who my roommate was at the time, but I remember he was
one of the speakers, which I was very impressed with, because that was before
I got to the point of presenting anything myself. The conference itself:
One of the first papers of the conference was Fred Parke giving a presentation
of his talking face, which research he did at [University of] Utah. And
now we have an entire session on faces, but this is kind of roughly what
his talking face looked like.
There's actually something that I'm a little disappointed in face research,
in that it's got its own version of the teapot, which they've kind of missed.
Fred Parke's animation with talking face was reciting a poem by Emily Dickinson.
It goes something like,
How happy is the little stone, that rambles in the road alone, and never
cares about careers, and exigencies never fears. and... etc., etc., Emily Dickinson stuff.
And it seemed like, this would be a good thing for all the face researchers
to try to reproduce and see the evolution of how convincing that is in time.
Maybe they can pick up on that and do it from now on.
This is also the conference I first saw Alan Kay give his presentation
demonstrating the Xerox paint program. I didn't know Alvy and Dick Shoup
at the time, but I was very impressed by the demonstration, and I went off
and did something based on that a little later on. And Alan Kay was just
sort of sitting in shorts and a blue shirt and just kind of hopped up and
sat on top of the table in front of the room and gave this talk. So things
were a little less formal than they are today.
1975, Bowling Green. This is where I met Nelson Max and we agreed to
collaborate on some pictures of his everting sphere thing. He actually emailed
me the control points of this everting sphere over the Internet, which existed
in 1975, and that was kind of a fun project. Also the film Hunger
was shown. I think that was the entirety of the film show that year.
1976 happened during my stay as a summer employee at the New York Institute
of Technology, so we all kind of piled into cars and drove down to Philadelphia.
This is the year that I showed the first teapot paper with the reflections
on it. And also the film show at that time was still pretty impromptu. We
just ran off a video tape of the teapot being rendered the night before
driving down there, and when they asked for submissions during the show,
I just hopped up on the stage and plugged it in and narrated it live. Also,
again things were pretty informal, I remember Ephriam Cohen gave his paper
by just kind of getting up and scribbling on the blackboard in the room
rather than even having any prepared slides. This is the conference I met
Steve Levine at, who encouraged me and Martin Newell to put together a course
for the next year in San Jose.
This is the cover that was designed by Maxine Brown, in fact. But this
was also the issue that had the first color pictures inside the Proceedings.
This [image] is done by Don Greenberg and his crew, I didn't do that one
myself, but it's a sample what the color images looked like at the time.
So, Martin and I put together this course on, as Steve Levine put it, the
course was supposed to contain "how you do all that cool stuff that
you guys do at Utah." Which we did. And this was the year that I met
Andy van Dam who came up to me in one of the course receptions and complained
bitterly about how crappy the course was. [ laughter]
Um, this is also the year that the film show died in the middle because
the power went out. Also during Martin's talk, one of his slides got jammed
in the projector and some helpful person from the audience came up to try
to unjam it and turned the slide tray upside down to get at it. [Dumping
all his slides on the floor.] This is also the year that I met Turner Whitted.
We both discovered we were working on similar problems of rendering patches
and we got into such an excited discussion at the back of the auditorium
that somebody came around and told us to shut up.
The next year, '78, the first time it was in Atlanta. And this is the
year that SIGGRAPH was held concurrently with some sort of fashion show
and the meeting hall that we were supposed to be in was preempted and they
kicked us into, I don't know, a parking garage or something with very low
ceilings. And so they had to go around and make three or four duplicates
of everybody's slides so they could have several slide projectors around
the auditorium, simultaneously, so people could see it because they didn't
have a big throw that they expected for the real slide projector.
SIGGRAPH had to work its way up to be respectable in the computer exhibition
business. This is the year that the first color picture [cover] came out
which Dick Phillips asked me to design. It was the year that I talked about
bump mapping, and, in fact, I got Lance Williams and Gary Demos on both
ends of the coast to help me a little bit on the rendering of that. This
is also the year that I brought a roll of exposed film to the conference,
hoping I'd be able to get it developed in time to give my talk, and, in
fact, I did, and that's what the picture of the orange was. This is also
the year that I met Pat Cole, who later came to work with us at JPL.
The next year, the first time it was in Chicago. This is the first time
I showed the Voyager fly-by movies by Jupiter, and, again, things were sufficiently
informal that I just kind of brought it with me to the conference thinking
there'd be some place I could show it. Then they were looking for things
to put together the film show at the last minute, and they stuck that in,
as well as other things. Also, this year, Pat Cole did a film and video
retrospective of earlier computer graphics films. I think is interesting
that, even in 1979, there was enough stuff and enough interest in the history
to put together retrospectives, even then.
1980, in Seattle. This is the year I remember, one of the things, was
that I forgot to pack socks when I went onto the plane, so I wore the same
pair of socks for the first half of the week until I found a place where
I could buy some more. This is, of course, the year that Loren Carpenter
showed his movie, To Fly, which was the hit of the show there. It
was one of the first fractal mountain animations, and it was kind of one
of the first pictures that I looked at and said, "that can't be a computer
image, that looks too good!" I'm easily fooled, I guess, but it did
look too good at the time, but, nowadays, you can look at it and, you know,
sort of tell because your expectations are increasing.
That year the film show, actually, didn't have enough room for everybody,
so they actually had two showings of the film show consecutively in one
night. The second showing, since this is all kind of done live and people
narrated their things live, we sort of abbreviated. I kind of felt sorry
for the people who came to the second show because they only saw half of
it because everybody was too tired to do the [whole] show the second time.
1981, in Dallas. This is when I first met Yoichiro Kawaguchi, who showed
me a book that he'd published with some of my pictures in it. And this is
one of the first times that we had an art show that had images viewed directly
onto a framebuffer display. Usually, computer graphics images had been photographed
and then you displayed the photograph of them. But photographs always looked
washed out or the wrong color, or something like that, and when you're producing
the pictures, you're looking at the display, so I figured seeing the actual
display in the art show would be a great idea and so James Seligman helped
me out with that. And, primarily, the images were ones that I had produced
and that David Em had produced. We felt a little bit self-conscious about
that, but those were ones we could get at easily (rather than trying to
promote only our images.)
Anyway, 1982, in Boston. Showed the Voyager-Two Saturn movie there, and
this is when I did the blobby paper and some cloudy light reflection papers.
This is where I met Vibeke Sorensen, who later came and worked with us at
the Art Center [College of Design].
1983, this is the award year. First time they gave out awards, and I
was happy to share the honors with Ivan Sutherland for the two awards they
gave out. Also, the framebuffer show, we had one there, wasn't working very
well, so I spent like an entire day behind the curtain reprogramming the
thing to set the [color look up table] registers properly for the framebuffer
so it would look good.
1984, in Minneapolis. This is when I showed the first Mechanical Universe
excerpt/demo, which was the physics series. This was about two hours worth
of stuff I'd done that year, packed into a three minute little movie-trailer
version with a few appropriate comments. It was my first attempt at comedy,
and fortunately people recognized as such. In fact, a lot of people came
up to me afterwards and said, "gee, Jim, I didn't know you had a sense
of humor." [ laughter] This is also when The Magic Egg
was shown. This is the collaborative effort of a lot people doing an Omnimax
movie because they had an Omnimax theater in Minneapolis. This is the year
that Tom DeFanti made up a bunch of kind of fake ribbons saying, "SIGGRAPH
85 - 89 Attendee." And was kind of the beginning of the ribbon phenomenon.
1985, in San Francisco. Showed more Mechanical Universe images,
and this was the year that they had the course on image rendering tricks
where I demonstrated my skill debugging hardware [that Walt mentioned in
his introduction]. And this was the year that Rob Cook handed out ribbons
that said "Party Jury" on them. That was to help us to get into
all the parties, you see.
Second time it was in Dallas in 1986. This is the year that I saw, what
I think, is still one of the most incredible computer graphics films of
all time, by John Lasseter, called Luxo Junior. Afterwards, I went
up to him and asked him, "gee, John, this lamp thing here, is that
the mother or the father?" But I don't remember which one he said it
was. I think it's up to your own imagination.
1987, Anaheim, first time. I did a course in 1987 on the production of
The Mechanical Universe, which is first and only one-man show that
I did for an entire day. And, unfortunately, I was so exhausted from three
weeks worth round-the-clock, making five or six hundred slides for this
that I practically collapsed on stage in the middle of it. But, I made it
through. This is also when they finally came up with something useful to
say about ray tracing in Paul Heckbert's Ray Tracing JELL-O®
paper.
Second Atlanta conference. Showed excerpts from a production about the
Theorem of Pythagoras that I'd been working on. We were still looking for
grant money to produce a series on high school level mathematics. We didn't
have much interest, at the time, from the US, but we had some interest from
some people in Japan. And so, in 1988, there was a big protest organized
by Tom DeFanti and Todd Rundgren saying, "Keep Jim Blinn in America,"
where they went around wearing signs in front of the film show. And it worked
because SIGGRAPH gave me a nice grant as seed money for Project: Mathematics!
and we went on and produced it.
1989, in Boston, the second time. Just before the conference, in fact,
Todd had organized it so that his band was performing in a rock club, and
he invited me to come on stage and play the trombone in one of his pieces.
I expanded my SIGGRAPH repertoire a bit that year. That was the year that
John Lasseter organized a session called "Bloopers, Outtakes, and Horror
Stories of SIGGRAPH Films," and so I was able to show a few of those
in that. And that year I showed just the opening sequence to the Project:
Mathematics! I had actually something in the film show practically every
year up to this point, and this one was only like five seconds long, but
at least I got it in.
1990, in Dallas. This is the first time we had the SIGGRAPH Bowl, which
I was sort of the mascot of. Basically, I waved applause signs when we wanted
people to applaud. And I showed to SIGGRAPH the first results of the Project:
Mathematics! thing. Maxine Brown did a nice thing for me. She made up
a hundred ribbons for me to hand out with the notation "I hugged Jim
Blinn." [ laughter] But only people who earned them got the
ribbons.
Las Vegas in 91. Las Vegas was, actually, a little bit big for SIGGRAPH.
It kind of got lost in there and diffused through the entire town, so I
don't have a whole lot of recollections from that one. [ laughter
] That was the year, however, Loren Carpenter first did his red/green paddle
thing, and that was a big hit. And, I understand, it was dicey getting it
to work in time, but it did, and everybody loved it, and we have another
one of those this year.
1992, Chicago two. This is the year that I had the maximum number of
ribbons on my badge. The organizers of 1992 contributed to that because
they all made their own ribbons saying various odd things. And, also, being
sort of a student of presentations and so forth, I realized that, when you're
talking like this and they put you on the monitor, if you wear dark clothing,
you disappear into the background. So, instead of using the green sweater
that I had been doing, I went out and had my mother knit me another light green sweater so I would show up better against
the backgrounds. [laughter, applause]
1993, second year in Anaheim. This is the year they had the virtual sex
panel which, uh, everybody had a good time with. 1993 was also the year
that had, what I think is probably the most amazing SIGGRAPH party that's
ever happened, which was the party at the Richard Nixon Museum. I have this
image of watching Timothy Leary ranting and raving on stage at the Nixon
Museum, twenty feet from the grave of Pat Nixon, and it's... just really
cosmic, I don't know.
This is also the year that I got married the day after SIGGRAPH was finished,
and somebody produced ribbons saying "Bride" and "Groom"
for Amanda and myself at the conference.
1994, in Orlando. One of the main recollections I have of that is another
very amazing party, put on by Bruce and Carmi where they rented out
the entire MGM studios for us to go and cavort around at one o'clock in
the morning. I'm still impressed by that.
Los Angeles, 1995. I don't know what this picture is. This is showing
SIGGRAPH getting into pain or something. This was the year I was on the
Electronic Theater Jury myself with Alvy and David Em, and got the chance
to pick the films that were shown. This was the year that the movie Toy
Story came out, which was one of the major achievements in computer
graphics up to that point. And this is the session where there was a panel
session called "Ask Dr. SIGGRAPH" where we had a bunch of us up
on stage attempting to come up with funny answers to questions that people
would pose. Jim Kajiya and I were kidding on stage, and Jim said, "hey,
Jim, why don't you come up and work for me at Microsoft." And I said,
"well, I don't know, give me your number, I'll give you a call."
I had no idea at the time that, a month and half later, I'd be working at
Microsoft.
Okay, 1996. New Orleans. This is the time that Al Barr and I actually
got into some parties, without tickets, by crashing in and attempting to
look important - and it worked! [laughter] And I showed one of
the papers excerpts at the film show that year.
1997, Los Angeles. I don't remember a whole lot about the conference
because I spent so much time talking to friends in the area, and I didn't
see much of the conference.
And, 1998, I gave the Key Note address. [laughter] There was
a big history thing and what not.
And, 1999. [laughter] Well, I'm getting a bit ahead of myself
here.
So, let's go over and kind of see what this all means. First of all,
we have to examine the ribbon count. [laughter / applause] And
you can see the physical manifestation of that in the SIGGRAPH Time Tunnel
out there.
This is kind of an interesting thing, if I can work this in the slide
sorter viewer in PowerPoint.
[shows composite of his entire slide presentation ]
Gives kind of a nice little everything-all-at-once of the entire 25 years
of SIGGRAPH Conference Proceeding cover history here.
I want to kind of go through and give some observations on how things
have changed over the last 25 years. You can kind of see the difference,
this is the first Conference Proceedings. [holds up 1974] Actually,
this is actually just the abstracts. At the time, they didn't know whether
this was going to be interesting enough for people to see the whole papers
printed. And, I understand, actually, the papers were printed in some other
journal elsewhere, which I never got a copy of, unfortunately. So, you can
see things have gotten thicker. [holds up 1998 versus 1974] And
they use a lot heavier paper. This thing weighs like 35 pounds. And they
use a lot smaller type than they used to, so you can fit more into the proceedings.
[flips through 1998]
Well, let's go through a few kind of changes between then and now.
Click for
text
Then. In 1974, that was before the invention of pixels, almost. Displays
were calligraphic, which means you steer the beam around on the screen to
trace out lines. Now, they're rastergraphic.
Click for
text
Then. Most of the graphics was done on mainframes or minis, which cost
in the $100,000s range. Then things worked down into workstation end of
things, which is in the $10,000s range. And now, you can do pretty major
things with just a desktop machine in $1,000-2,000 range. The change in
cost of these things of a factor of a couple hundred or so, over the years,
has obviously had implications in terms of accessibility. Now, a lot more
people can get at it than used to.
Click for
text
So, at the time, there weren't very many people around, because it took,
like, institutions to fund these displays that now we have on our desktop.
Now everybody is getting into the act.
Click for
text
Back then, everything was hard. It was hard just getting a line on the
screen. It was real hard getting the line recorded onto film or on video.
Now days, everything is easy. That might come as a surprise to all these
people who are trolling away, fudging pixels in these high-res special effects
movies. But, a lot of the stuff you can get now is off-the-shelf and you
don't have to reinvent everything in order to get any sort of thing
on the screen at all.
Click for
text
Then, since computers were slow, the emphasis was on clever algorithms
that made large scale decisions to skip things or to make simple pictures.
Nowadays, computers are fast enough so that you can get away a lot more
with brute force type algorithms.
Click for
text
Then, most of the applications of computer graphics were in the computer-aided
design and data analysis. And that was what computer systems were sort of
designed or aimed at. Nowadays, a lot of the applications are in special
effects, image processing, graphic arts, and so forth. And so programs are
designed to do that.
Click for
text
Then, most of the papers were concerned with the feasibility of something.
Is it at all possible to make a cube rotate on the screen? Is it all possible
to make something that looks vaguely like a human face? But, just getting
it to work at all was the major accomplishment. And having it be incredibly
good looking was a little beyond what we could do at the time. Nowadays,
people are focusing on more refinement and practicality of things, and our
standards are a lot higher. Just putting a face on the screen is not interesting
unless you can make it look like a really good face. Which is what's going
on nowadays.
Click for
text
Also, in going through the old conference proceedings, I noticed an interesting
pattern as well. Back then, most of the papers were application papers.
There were things like, you know, computer applications in cartography and
computer applications in architecture and civil engineering, biomedical
applications, data analysis, and social sciences and so forth. Nowadays,
most of the papers are on rendering, which is like how to make pictures
of faces, or cloth, or plants, or various geometric things in multi-resolution
surfaces, and what have you. So, the applications don't show up so much,
which is maybe something we ought to think about a bit. Although, I'm sure
that SIGGRAPH people are interested in applications papers, but probably
fewer people are sending them in.
Click for
text
Back then, the papers were simple enough so that the talk you gave at
the conference, you could give the whole paper and tell what you did. Nowadays,
the talk is more like a movie trailer version of the paper. There's no way
that you can get everything in a paper into a twenty-minute talk because
the things are a lot more complex and a lot more involved. Nowadays, a SIGGRAPH
paper, in fact, is a yearlong effort, a yearlong multimedia project. First
you have to write the paper. Then you have to do the video that goes with
the paper. Then you have to do the finished version of the paper. Then you
have to make the slide show version of it. Then you have to perform it in
public. And then you have to put it on your website. So, you know, life
is a little more complicated now than it was getting something published
in SIGGRAPH.
Well, what else am I going to say? I seem to have a few more minutes
here.
Click for
text
One other kind of venerable thing that people have done over the years
is the concept of unsolved problems. That was first done in 1966 by Ivan
Sutherland, and, eleven years later, Martin Newell and I wrote up a list
of unsolved problems. 1987, Paul Heckbert did one. 1991, there was an entire
SIGGRAPH panel on it. And so, 1998, I'll do my part again.
Click for
text
Let's just kind of review the old ones briefly. In 1966, these were what
Ivan thought were the unsolved problems. Basically, cheap machines - we've
got that now. Basic interaction techniques are the things that we use a
lot with mice and so forth. Coupling simulations to display - there are
a lot of things doing that. Most of these things have been addressed quite
a bit, although not everything is perfectly solved.
Click for
text
The things that Martin and I talked about in 1977 were mostly rendering
problems rather than interaction problems. Just making more complex scenes,
fuzzy objects - this is back, of course, when we could make cubes, and that
was about it. Teapots were the state of the art. Fuzzy objects - hmm - I
guess we got that covered now. Transparency, refraction, and so forth.
Click for
text
Paul Heckbert had another set of rendering things that he published in
1987. What's interesting about this list is he's again getting into the
area not so much of how to do something at all but how to do something practically
and fast, and, you know, in a production sort of environment rather than
just showing feasibility only.
Click for
text
And then the SIGGRAPH panel, by 1991, the thing had gotten complex enough
that it took an entire panel to come up with some unsolved problems, and
these were the ones that individuals in the panel found important.
So, what am I going to do in 1998?
Click for
text
Well, in thinking about this, first of all, the question comes up "what
does solution mean?" Is a solution to a problem mean being able
to do it at all, in 8 or 9 hours long rendering and so forth, or does the
solution mean doing it in a practical sense and doing it quickly enough
so that people can use it other than as an existence proof. There are some
problems in either of those categories - some things that nothing has been
touched and others that people need to make practical in order to make usable.
Click for
text
Before getting into those, I want to talk a little bit about some problems
which nobody put in their list and that have been addressed, but I won't
say have been completely solved. For example, non-photorealistic rendering
was something that nobody really predicted a need for in any of these things.
Likewise, image based rendering. And, something which I call the "tyranny
of the framebuffer," which is the recognition of the fact that having
just one piece of memory where the entire image is stored is, actually,
rather inconvenient in today's systems which have multiple windows moving
around and you have to rerender things that get uncovered and so forth.
What's really a lot nicer is if you have, kind of, almost a return to the
mechanism of the calligraphic systems where you have a list of things that
you are going to draw. All you, as the programmer, need to do is to manipulate
that list and then the hardware dynamically composites all the items in
that list, together, into one image on the screen. This is one of the components
of the Talisman system that Jim Kajiya and his colleagues put together a
couple of years ago. Talisman includes a lot of things, but I think that's
one of the most interesting aspects of it, and I think it will effect how
displays are built in the future.
But, these are problems that nobody had in their list and, maybe, it
means that nobody thought they were worth solving, but, any event, let's
march forward here. What are my ten unsolved problems in computer graphics?
Click for
text
Problem number one - finding out something that hasn't been done yet.
[laughter] There's thousands of people out there, all beavering
away, figuring out how to draw this, how to draw that, how to interact with
that and so forth, and it's really becoming difficult to figure out something
that somebody hasn't already invented somewhere else.
Click for
text
Problem number two - this is kind of related to the first one: Finding
out if somebody has done it yet or not. That is, keeping track of all the
things that have been done. It used to be that SIGGRAPH was the only place
that would publish computer graphics papers, and so all you had to do was
read the SIGGRAPH Conference Proceedings and you knew you were up to date.
But nowadays there's lots of other journals, and it takes more and more
effort to make sure that you know what's happening. On the other side of
the fence, disseminating the discoveries that you make yourself is also
a problem, figuring out the best way to expose as many people as possible,
so they don't go through and be tempted to reinvent it, although it's sometimes
a lot more fun to reinvent things.
Click for
text
Number three - systems integration. That is putting all the techniques
into one system. We've got a lot of people doing cloth. We've got a lot
of people doing faces. We've got a lot of people doing dance motions and
interaction and speech and what have you. Putting that all into one system
is still something that we've got a lot of inroads into, but something that's
always a tricky business.
Click for
text
[4] Simplicity. Making this so that's not so complicated that you can't
figure out how to use it. Now, simplicity is a tricky concept. I'm not sure,
in fact, that simplicity is possible in anything that is complex enough
to do something useful. If you look at systems like the human brain or the
telephone system or any of these things that have been built up over the
years, they have layers and layers of different components in them, and
they still work together pretty well. And so, simple systems might not be
achievable, but they're still a goal that you should try to make it as simple
as possible.
Click for
text
Number five - pixel arithmetic theory. This is something that, actually,
I've been spending a lot of time playing around with recently, and that
has to do with the idea of "what is a pixel?" and "what sort
of arithmetic do you do on it?" About 1984, I guess it was, Porter
and Duff came up with their paper on alpha blending and the concept of alpha
pre-multiplied into the pixels, and still a lot of people doing computer
systems haven't gotten that concept. Pre-multiplied alpha is a good thing,
virtually always, but the fact that there are systems that don't use it
indicate that we maybe don't understand the problem completely, that there's
more going on than just that.
There is such a thing as an alpha that's stored in each pixel that indicates
the shape of the entire object and there is a global alpha that is applied
to the entire picture to make it, like, fade-in and fade-out, things like
that. Alpha assumes that the edges of the object are completely uncorrelated
- maybe there should be some additions to correct for correlated edges instead
of uncorrelated edges. And when you start dealing with alpha blending of
things stacking on top of each other, each of them are partially transparent,
you think of them as, maybe, colored cellophane or something like that.
Then you start getting to spectral matching and how do we combine the alpha
concept with the idea of light going through transparent, colored glass.
And combining this with models of light reflection where, every time light
of a particular color bounces off something, it gets changed to another
color. All of these things indicate to me that we don't completely understand
this process and all aspects of it to form a kind of "unified field
theory of pixel arithmetic" and so this is something that I think is
worth pursuing.
Click for
text
[6] Kind of on a related idea, there's compatibility of older systems.
Two, in particular, I'm keeping in mind. One is TV pixels in the upcoming,
uh, let me call it convergence of TV and computers. We are going to start
running into this problem more and more, and that is that TV pixels are
typically gamma-corrected and computer graphics pixels are typically not.
And doing calculations on these things, a lot of people just kind of sweep
that difference under the rug and do the arithmetic on gamma-corrected pixels
as though they were not, giving the wrong answer. Not very wrong, but sort
of wrong. Also, TV pixels have a different range of quantization - instead
of going from 0 to 255, they go from 16 to 235, they have a narrower range.
Making sure that we can do arithmetic on TV images the same way we can do
arithmetic on computer graphics images requires all these conversions back
and forth which is either slow or, if you don't do the conversion correctly,
it's inaccurate, and you've got this balancing act that you're doing. How
do we do that?
The other legacy compatibility issue is 3D APIs. There are a bunch of
them around. Some of them are very widespread and popular, but they were
invented many years ago, and we've thought of new ways of doing things since
then. And how do we figure out how we can put new capabilities into the
3D APIs that we're doing in the future while not completely leaving all
the old ones in the dust. Legacy compatibility is, for all these issues,
something that we're going to have to deal with more and more as we have
more and more legacy stuff.
Click for
text
Number seven - arithmetic sloppiness. There is an interesting phenomenon
that I'm discovering as I learn what the recent developments in real-time
3D hardware are going on, as well as real-time software rendering systems.
And that is, in the old days, when we did computer images, we knew it was
going to take 45 minutes, so we did the arithmetic right. Nowadays, these
things are almost real-time, but if you take this little shortcut or this
little approximation, you can make it real-time, but it's not as accurate.
And, so, there's this jump of quality in order to get to real-time, since
computers aren't quite fast enough to do it correctly in real-time. So a
lot of the systems make pretty astonishing, crude approximations to what
the actual calculation ought to be for blending pixels together or for doing
the geometric arithmetic and so forth. I find this kind of irritating that
people are making these things, but they do it for a reason - because it
makes it faster, and the trade off between speed and quality is something
that everybody has a different position where their application fits.
Problems with dealing with quantized pixels, when you have either eight
or even as few as five bits per pixel, doing the arithmetic properly is
important, but maybe not that important. Texture filtering: a lot of that
is being shortcut a lot in real-time systems. Lighting models... there's
something that always bothered me about lighting models. Bui Tuong Phong
is a great guy and he did wonderful work and so forth. But he introduced
this concept of the cosine power. That is, the light reflection for specular
reflection was a function of how far your viewing ray was away from a reflected
ray. And in order to sculpt this function into roughly the shape he wanted,
he just took the cosine of that angle and raised to some power. You use
a larger power, you get sharper highlights - a little smaller power, you
get fuzzier highlights. The thing is, this has no physical basis whatsoever,
but all graphics systems today, now, do highlights this way, and the idea
of the cosine power, what's the cosine power of your surface? Surfaces don't
have cosine power. This was a crude approximation at the time. There are
better ways of calculating things like this. I'd like to see cosine power
retired and better approximations being done.
Anyway, what this ultimately boils down to is the fact that we need better
criteria about how accurate we need to be. A lot of these sloppy calculations
are done, and, you know, purists will say, "ick, that's awful."
But somebody who does it says, "well, look at the picture - it doesn't
look all that much worse than doing it right." Then you look at it
and you say, "well, they do look pretty close." So what we need
is better criteria for how accurate we need to be to be able to properly
place ourselves on the performance and quality curve here.
Click
for text
[8] Antialiasing. Antialiasing of textures is still, uh, we have various
approximations to that and we can do better. In particular, one of the things
that we want to do is to be able to make text that remains readable as we
take a frame of text and flying it around in three dimensions with perspective
and so forth. Being able to do the filtering properly to make it readable.
Antialiasing is a problem that will always be with us. I might actually
state that, well, when you're talking about problems and predicting things,
sometimes it's maybe better to say what never will be done, because that
guarantees that somebody is going to come along and do it. For example,
at the turn of the century there was somebody in the patent office, I guess,
who said all possible existing inventions had already been invented - there's
nothing more to invent. And, of course, they're quoted all the time. There's
a quotation from Bill Gates that said 640KB of memory is good enough for
anybody. He's right, actually. [laughter]
So I think may I can become more famous by saying what can't be done,
because a hundred years from now, people will say, "Jim Blinn said
this couldn't be done, and, of course, we've done it." So the things
I want to be done, I'll say can't be done.
"Nobody will ever figure out how to do antialiasing." [laughter]
Click
for text
[9] No list of unsolved problems is complete without some challenge to
the community for modeling, and rendering, and animation of something. It
used to be just "how to do computer graphics on something," but
we need to say:
- modeling is figuring out the shape of it
- rendering is how to make a picture of it
- animation is figuring out how it moves with time
So here is my challenge to the computer graphics community. [puts
up "spaghetti"] Now, people have been doing cloth for quite
a while, which is basically two dimensional shape and they're figuring out
to make it crumple as it falls and the intersection and so forth. Spaghetti
must be a lot easier problem because it's just one dimensional instead of
two-dimensional. So, I want you to be able to figure out to drop a piece
of spaghetti onto the plate and how it squiggles up and model the sauce
on there for the frictional coefficients and so forth. [laughter] This probably has applications beyond spaghetti. One can imagine, you
know, dropping ropes into piles on the floor, or string, what have you.
It could even lead to models of protein folding and so forth.
So that's my challenge to the computer graphics community.
And the final unsolved problem...
Click
for text
[laughter / applause]
[10] Finding a use for it. This is actually a real problem in the business
side of things because so far we have games and simulations that are driving
the hardware business on the low-end of things, but that's not enough users
to justify the cost of making these things. Even though the graphics chips
for doing real-time 3D are getting down into the twenty-dollar range, the
companies that are making them are going out of business. So we need figure
out some way that business people can get some benefit from this, communications
can get some benefit from it, visualization and so forth for real-time 3D
that will make lots more people want to buy these things.
Consider user interfaces. One of the things that I've seen that occurs
to me that I don't like is the way system state is represented visually
on the screen. Settings you have in your program or settings you have configuring
your computer, you click around, there is the window, the window points
to the other window, and there's this list and menu so forth. It seems like
a better way of doing it would be some sort of three dimensional shape,
kind of like a car engine, where you can see the components in there, and
you can pick this one, open it up, and see the thing inside there. Some
sort of 3D user interface to represent system state or represent the program
settings I think would be a great thing and would make it a lot easier to
be able to find the way to change a particular setting you want. So this
is perhaps a fertile ground for ways of milking benefit from the 3D thing.
So, here's Dr. Blinn's ten unsolved problems as of 1998.
Click
for text
And so, well, none of these are going to be solved, you know.
Click for
text
Now, no keynote address is complete without some discussions of the future.
So, Dr. Blinn predicts...
In the future, computer graphics will be... faster. [laughter / applause]
In the future, computer graphics will be... cheaper.
In the future, computer graphics will... take more memory.
... it will be more realistic,...
... and, it will be seamlessly integrated with whatever it wants to integrate
with,...
... and, of course, it will run on NT. [laughter / applause]
Now, there are those of you who might think that these predictions are
not daring enough. So, let me be a little more daring.
Click for
text
I think that in the future... this is something that people have asked
me for many years, and I used to think it was really dumb - "When are
synthetic actors going to be indistinguishable from real actors?"
Actually, I think we're getting there. I think we're not quite there
yet, but I would say by the year 2000, we'll be there...
... in April...
... 23rd...
... at 3:00pm. [laughter / applause]
What other things are going to happen in the future?
Click
for text
Better display screen technology. They've been promising us flat panel
wall screen displays since 1956 or something like that. But, you know, they're
never going to do that. We're never going to have cheap flat
panel wall screen displays.
Sprite based displays. This is the Talisman concept of having a list
of images that are dynamically composited, so that the programmer just deals
with the manipulation of a list and the hardware deals with compositing
it together into one screen. Once you have that, it's possible to distribute
videos directly in this format. This is kind of, part of the concept of
MPEG-4. Rather than distribute video as final frames, you get much better
compression when you distribute the layers individually, compress them individually.
We're going to have holographic projectors... and direct brain input...
and, well, you might get the sense that I find it hard to take future predictions
all that seriously some how, but [continuing on] ...
Click for
text
We're going to replace the Hollywood backlots with complete synthetic
things. Theme parks are going to be completely virtual.
2003, we won't need to come to SIGGRAPH anymore. [laughter]
And, pretty soon, anything that's not on the web won't exist. I do research
on the web now, and, you know, if I can't find it on the web, I say, well,
I guess it's not around, and then I realize, oh, there are actually libraries...
[laughter] ... but... a few years, it won't be anymore.
Click for
text
So, 25 years. What does this all mean?
We've created a new industry. We've taken something that was an academic
curiosity back in 1974 and we've turned it into something that's so pervasive
and so common that it shows up in practically every aspect of imaging. From
film and movies to print medium... in fact it's sometimes cheaper to use
computer graphics animation for things than live action. We saw a commercial
on TV, once, where there's a moving van moving down the street and we realized
- that's a computer simulation. I guess it was just cheaper to do it that
way than to rent a real moving van.
We've also changed, I think, how technical conferences are run. SIGGRAPH
has brought in the artistic community. We have art shows as well as film
shows, and we have dance performances and what not. This is very unusual
for a technical organization. The American Society of Physics probably doesn't
do this. [laughter] But, I think we've changed the perception of
the technical community as to what sorts of interesting things can go on
at conferences, and I think that SIGGRAPH has been instrumental in making
life more fun in the technical business.
We basically changed the world. We've made manipulating pictures, which
is one of the most powerful communication medium we've known, easy and accessible
and available to everyone.
And, as Jim Kajiya says, we now have new medium. Back in my generation
we grew up not being able to imagine what the world would be like without
television. Nowadays, we have a generation of kids growing up not being
able to imagine a world without computer graphics.
And, sort of to underscore that, I want to show a little project that
I've been working on... on video. This is a video of my little boy watching
his favorite movie. [cues video of his son watching Toy Story]
It's a good thing it's such a good movie, let me tell you, because we see
it two or three times a day. [laughter] And what I'm looking forward
to is, in maybe three or four years, when he's old enough to understand
the concept, I'll be able to say to him, "see all that little bumpy
stuff on Mr. Potato's head? Your daddy figured out how to do that."
Thank you.
[ laughter / applause]
Volunteer transcription by John
M. Fujii, SIGGRAPH 98
Copyright © 1998, ACM SIGGRAPH 98
- How happy is the little stone
- That rambles in the road alone,
- And doesn't care about careers,
- And exigencies never fears;
- Whose coat of elemental brown
- A passing universe put on;
- And independent as the sun,
- Associates or glows alone,
- Fulfilling absolute decree
- In casual simplicity.
-
- Emily Dickinson, 1881
Dark green sweater
Light green sweater
