PS9 conference at GDC2004!!

[Quaz51]

PS9
Speaker(s): Dominic Mallinson, Richard Marks
Time/Date: Thursday (March 25, 2004) 4:15pm — 5:15pm
Track: Game Design
Format: 60-minute Sponsored Session
Level: Intermediate - Basic familiarity or some experience.

Description: Science or fiction? Join us as we repeatedly ask the question, "What if...?" and then seriously consider plausible ramifications. What if game machines had massive network bandwidth between one another? What if network latency was microseconds and not milliseconds (faster than light!)? What if compute power was not a limit (petaflop performance)? What if AI systems start comfortably passing the Turing test? What happens when we can no longer tell that a character is rendered? What if we all had true 3d displays? What if we have ubiquitous networked displays (on cornflake packets, clothes, walls...)? What if we have ubiquitous networked 3d cameras? What if games could do real-time motion capture in the home? What if a game machine could understand natural speech? What if a game machine knew if you were happy, sad, or bored? What if telepresence becomes commonplace (drive a real race car from your living room, visit real locations with real-time 3d sensing)? What if the cyberspace of science fiction was a reality?

We will also provide technical opinions on the likelihood of when/if each question may be relevant, and for some questions we will show research results that represent initial steps.

Idea Takeaway: An awareness that technological breakthroughs and hardware advancements will continue to re-invent this industry, thereby keeping it healthy and fresh for years to come. Or at least a few good ideas for some science fiction novels.

Intended Audience and Prerequisites: All game developers interested in speculation as to how computer entertainment might evolve in the years to come.

http://www.cmpevents.com/GDx/a.asp?option=C&V=11&SessID=2585

 

[Evil_Cloud]

That's in fact THE most important question in the industry... Where the fuck are we heading to!

 

[Wunderchu]

ooo , cool

 

[McFly]

What if network latency was microseconds and not milliseconds (faster than light!)?

Oh my ... they should take a physic lesson first when they talk about something like that.

Fredi

 

[Guden Oden]

What a nutty topic... Don't the people behind this conference have better things to do than stand on a stage waffling speculative BS for an hour?

 

[pcostabel]

Agree. Clearly there weren't enough good speakers, so they conjured up this bs to fill an empty slot...

 

[MrWibble]

Yes... it's completely beyond me as to why anyone would want to hear the head of Sony R&D in the US and the guy that came up with the idea for eyetoy discuss what the future might bring...

Lets all just ignore them and discuss exactly the same topic amongst ourselves, like usual.

 

[nobie]

The absurd thing is that petaflop performance and truely photo-realistic 3d are pretty far off. Even the pre-rendered CG you see in movies is not indistinguishable from reality.

 

[Vince]

The absurd thing is that petaflop performance and truely photo-realistic 3d are pretty far off

You know, before the board change, I was basically the only person here stating that a TFLOP/IC was even remotely possible. This entire board basically dismissed it entirely out of the realm of possibility - yet, 3 or 4 years later you'll be hardpressed to find a person to say it isn't possible.

Once again, you're opening your mouth and not looking forward. Nobody stated that it had to be a logic construct akin to anything you see today - which isn't a problem as there are many alternative forms of computation, but one thing's for certain.. it will happen and it will happen [relatively] quickly compared to past advances.

PS. Mr.Wibble is the man, as usual.

 

[Guden Oden]

Yes... it's completely beyond me as to why anyone would want to hear the head of Sony R&D in the US and the guy that came up with the idea for eyetoy discuss what the future might bring...

Dude, I think most of us are pretty good at imagining what the future MIGHT bring, including faster than light networking and whatnot. What the future WILL bring, now that's another matter entirely, but I don't see that mentioned in the announcement quoted in the first post.

How unexpected! .

 

[ERP]

To be fair these types of talks are normally the most interesting at CGDC, although the general rule if you want to stay awake is pick the speakers, not the subjects.

 

[MrWibble]

Dude, I think most of us are pretty good at imagining what the future MIGHT bring, including faster than light networking and whatnot. What the future WILL bring, now that's another matter entirely, but I don't see that mentioned in the announcement quoted in the first post.

How unexpected! .

I like to try and imagine a future where we all just get along - but that's just science fiction.

Anyway... from the original abstract:

"We will also provide technical opinions on the likelihood of when/if each question may be relevant, and for some questions we will show research results that represent initial steps."

It seems pretty clear to me that the intention is to back up the wild flights of fancy with actual fact - now wouldn't that make a nice change?

 

[Deadmeat]

...

To do next generation graphics you need a massive jump in memory capacity.

256 MB is not enough.

 

[Fafalada]

It seems pretty clear to me that the intention is to back up the wild flights of fancy with actual fact - now wouldn't that make a nice change?

But that sounds like so much hard work.
Isn't it much more convenient to just talk about how we will use Directx666 with Pixel and Vertex shaders 42.0 and how Geforce-XXX with Fuel injection and Turbo chargers will run them in the future?

 

[nobie]

The absurd thing is that petaflop performance and truely photo-realistic 3d are pretty far off

You know, before the board change, I was basically the only person here stating that a TFLOP/IC was even remotely possible. This entire board basically dismissed it entirely out of the realm of possibility - yet, 3 or 4 years later you'll be hardpressed to find a person to say it isn't possible.

*golf clap*

You seem to think of yourself as a gloriously intelligent person, but your attitude portrays you as more infantile than intelligent. Aesop said "the smaller the mind the greater the conceit."

Once again, you're opening your mouth and not looking forward. Nobody stated that it had to be a logic construct akin to anything you see today - which isn't a problem as there are many alternative forms of computation, but one thing's for certain.. it will happen and it will happen [relatively] quickly compared to past advances.

I hardly think I'm going out on a limb by saying petaflop performance and photorealistic 3d are pretty far off. PS3 is pretty far off, and it "only" has one teraflop performance. When will their be something 1000x faster? When will a computer generate images indistinguishable from reality? Pretty far off again.

 

[Vince]

You seem to think of yourself as a gloriously intelligent person, but your attitude portrays you as more infantile than intelligent. Aesop said "the smaller the mind the greater the conceit."

Hardly, but what happened happened. I can't change what people argued or stated, nor can I what I said. Nor was it a self-reinforcing remark, but rather to use the case as precedent for the current thinking being used.

How many people, in total, believed a TFLOP/IC was possible in 2001? Hell, most people here told me I was wrong when I said it would be <100nm and told me I was wrong when I said Cell was for PS3 and not a Network Router - so don't pull this crap with me. I know that's been my position, V3 sort of came around in that time IIRC as did a few others; but the opposition was huge. Hell, even someone who truely is an "intelligent person" - Mfa - said they'd never reach it and wouldn't pass 200GFLOP/IC.

I do think this is a case of you (a) misreading why I stated that and (b) forgetting what was said.

I hardly think I'm going out on a limb by saying petaflop performance and photorealistic 3d are pretty far off. PS3 is pretty far off, and it "only" has one teraflop performance. When will their be something 1000x faster? When will a computer generate images indistinguishable from reality? Pretty far off again.

Well, see, I hardly consider PS3 as "pretty far off." Actually, I consider that insane as I understand "pretty far off" as over a decade. In 1993, the world computing record was held by dual systems at Los Alamos and by the NSA peaking at 131GFLOPs. Within 15 years of 1993, IBM should be at 1,000,000GFLOP/sec using constant IC counts. Assuming the same relative time from supercomputer to commodity IC is upheld by any number of promising avenues of computation.. well, you can do the math.

"Indistinguishable" is a such a plastic term, do you have to solve every nondeterministic or chaotic occurance? I know of people who mistook FF:TSW for real when flipping threw the channels. Or all those people fooled in Gladiator - but you wouldn't consider these valid since they're obviously people inferior to you.

And at the end of the day, what's more absurd than having nothing better to do than critisize a GDC2004 talk for having a catchy abstract/title as to draw people who are up for a good discussion on the future than taking part in the same discussion on a console forum.

 

[nobie]

You seem to think of yourself as a gloriously intelligent person, but your attitude portrays you as more infantile than intelligent. Aesop said "the smaller the mind the greater the conceit."

Hardly, but what happened happened. I can't change what people argued or stated, nor can I what I said. Nor was it a self-reinforcing remark, but rather to use the case as precedent for the current thinking being used.

How many people, in total, believed a TFLOP/IC was possible in 2001? Hell, most people here told me I was wrong when I said it would be <100nm and told me I was wrong when I said Cell was for PS3 and not a Network Router - so don't pull this crap with me. I know that's been my position, V3 sort of came around in that time IIRC as did a few others; but the opposition was huge. Hell, even someone who truely is an "intelligent person" - Mfa - said they'd never reach it and wouldn't pass 200GFLOP/IC.

I'm not familiar with the discussions you're referencing. If you made a successful prediction, congradulations.

I do think this is a case of you (a) misreading why I stated that and (b) forgetting what was said.

The content of what you said is of no consequence, I was only commenting on your puerile condescending attitude. To make a serious argument you first have to be taken seriously.

I hardly think I'm going out on a limb by saying petaflop performance and photorealistic 3d are pretty far off. PS3 is pretty far off, and it "only" has one teraflop performance. When will their be something 1000x faster? When will a computer generate images indistinguishable from reality? Pretty far off again.

Well, see, I hardly consider PS3 as "pretty far off." Actually, I consider that insane as I understand "pretty far off" as over a decade.

I think in this instance you are the one who misread what I stated. I did not say "over a decade," in the context of this discussion I don't think that's a reasonable interpretation of what I said. If you misunderstood, I'm sorry, but I don't think that was my fault.

In 1993, the world computing record was held by dual systems at Los Alamos and by the NSA peaking at 131GFLOPs. Within 15 years of 1993, IBM should be at 1,000,000GFLOP/sec using constant IC counts. Assuming the same relative time from supercomputer to commodity IC is upheld by any number of promising avenues of computation.. well, you can do the math.

Well their is no "commodity IC" capable of 131 gigaflops, even after eleven years. So making the assumption in your example, a petaflop processor in consumer devices is clearly over a decade away. I'm not making that argument, but I do think your logic is flawed.

"Indistinguishable" is a such a plastic term, do you have to solve every nondeterministic or chaotic occurance? I know of people who mistook FF:TSW for real when flipping threw the channels. Or all those people fooled in Gladiator - but you wouldn't consider these valid since they're obviously people inferior to you.

If I can distinguish CG in a movie from reality, they are obviously not indistinguishable.

And at the end of the day, what's more absurd than having nothing better to do than critisize a GDC2004 talk for having a catchy abstract/title as to draw people who are up for a good discussion on the future than taking part in the same discussion on a console forum.

I am not taking part in such a discussion.

 

[Vince]

I think in this instance you are the one who misread what I stated. I did not say "over a decade," in the context of this discussion I don't think that's a reasonable interpretation of what I said. If you misunderstood, I'm sorry, but I don't think that was my fault.

What? I think it's pretty clear that the "over a decade" comment was mine. Unless your interpretation of this comment:

• Actually, I consider that insane as I understand "pretty far off" as over a decade.

Can somehow turn that into one in which I'm attributing it to you. *shakes head*

Well their is no "commodity IC" capable of 131 gigaflops, even after eleven years. So making the assumption in your example, a petaflop processor in consumer devices is clearly over a decade away. I'm not making that argument, but I do think your logic is flawed.

First of all, I stated "15 years" - not "11 years." Lets stick to what was said (a recurrent problem here it seems) and go from there in explaining this discussion.

You originally stated that a few topics at discussion in the SCE talk @ GDC were "absurd" in that "petaflop performance and truly photo-realistic 3d are pretty far off".

I then questioned your use of the term "pretty far off" - since I took this to mean "over a decade" - which would make sense as the name of the GDC event is PS9 (That’s "Nine" - as in 6 more than 3). I then further reinforced the validity of the non-absurdity of this by stating that if you look at the trends in supercomputing scaling (at a constant IC count) that they've actually hit your "1000X" number within 15 years. I further posited that there is a relatively stable period from which a given performance level is seen in supercomputing clusters and makes the transition to consumer ICs - which would coincide with with the context of the debate and what a "normal" individual would consider "pretty far off" at conference entitled PS9, and most likely much sooner than most expect.

I also made a parallel to the previous belief of people and PS3's expected computation power, but since you admit to having no clue what was stated then or the mentality of most people, you just used that against me. So, whatever.

You responded with a "*golf clap*" and stated that even "PS3 is pretty far away" which is just so amazingly obtuse in light of the conference name being PS9 - but who actually thinks rationally when you can just attack my "infantile" attitude?

You seem to think of yourself as a gloriously intelligent person, but your attitude portrays you as more infantile than intelligent. Aesop said "the smaller the mind the greater the conceit."

So, I'm the "gloriously intelligent" one who’s really "infantile" but, yet, you're incapable of presenting an actual objective counterargument other than stating things like:

• but I do think your logic is flawed.

  
  

  If I can distinguish CG in a movie from reality, they are obviously not indistinguishable.

Whoa... I'm blown away. I really am. So, I suppose you're beyond actually responding with some objectivity and tangible position which we can examine. I guess you're just above that in stature, eh?

Aesop, whom you seem fond of quoting, also stated that:

• "It is easy to be brave from a safe distance"

Perhaps you should think about it.

PS. If you're not taking part in any such discussion (like this), are you bipolar?

 

[Tagrineth]

Jesus christ, guys. Grow up.

Vince, he said eleven years not as a target date, but as the actual year count from 1993 to now (2004 - 1993 = 11). He's also using PS3 as a point of reference for 1TFLOP, not PFLOP.

You're both needlessly mincing words and obfuscating your arguments, making both incredibly hard to follow... which results in more semantics bickering, and so on... most of what I read in Vince's last post above mine is a more or less complete mis-interpretation of nobie's post.

Vince... if IBM or whatever had a 131GFLOP multiprocessor system in 1993, and we still don't have a single IC capable of that today (hell... I remember EE's 6.2GFLOPs being impressive, and that's three processors)... what makes you think we'll have a 1PFLOP single IC in eleven years - which is more than a decade, or what you would consider pretty far off yourself?

And besides anything else, Cell isn't really a single processor now is it? AND on top of that, we have yet to see concrete proof that a single IC Cell in "PS3" will indeed provide 1TFLOP. It's actually looking pretty likely at this point, but there is in fact no proof.

 

[Vince]

Jesus christ, guys. Grow up.

Thanks Mom.

]Vince... if IBM or whatever had a 131GFLOP multiprocessor system in 1993, and we still don't have a single IC capable of that today (hell... I remember EE's 6.2GFLOPs being impressive, and that's three processors)... what makes you think we'll have a 1PFLOP single IC in eleven years - which is more than a decade, or what you would consider pretty far off yourself?

And besides anything else, Cell isn't really a single processor now is it?

Tag, several things. First of all, you keep confusing "processors" with ICs. I really don't care how many "cores" or "processors" there are per die, nor does it really matter. It's still a single IC on a single die or component which is what matters. If you can get N preformance from using 1 massive uniprocessor or 500 redunctionist cores or 1billion Chinamen on speed in individual rooms, honey-I-shrunk-the-kids style, is insignificant in the debate.

Second, I've stated that there's been a more or less consistant rise in processing power in supercomputing applications. This has been trailed to a large degree by consumer computing's increase. Do the back of the envelope math and you'll see a roughly ~15-20 year lag from supercomputing to consumer adoption. Which, is around 4 console cycles and would definitly fall into the over the horizon range the PS9 discussion was targetting.

If I'm not mistaken, I'm usually pretty good at remembering things I've seen or read, you read Ray Kurzweil's book (?) and while he's a contoversial figure in his own right and would make a good discussion some time - he did predict such a level of preformance by 2019 for a consumer PC with a constant $1K price IIRC. Which would trail any dedicated IC designed for a console by ATI or SCE by abit. So, how again is discussing this topic, under these pretenses, "absurd"?

That is "absurd" in any regards other than the same tired arguments made against the discussion in 2000/2001 about PS3 being over a TFLOP?

Vince, he said eleven years not as a target date, but as the actual year count from 1993 to now (2004 - 1993 = 11). He's also using PS3 as a point of reference for 1TFLOP, not PFLOP.

Right, but I stated 15 years. He then countered that in 11years (thanks for the math ) we haven't see such a rise. Now, this is debatable in it's own right as I'm sure the guys from ATI could make a strong case that their high-end ICs are closing in really fast - but what's beyond debate is that what he's doing is akin to a guy predicting he'll own their first house in 5 years and then someone countering "But, it's been 3 years and you don't own anything yet". Yeah, well, no shit sherlock - he made that time prediction for a reason and behind it is the work and money leading upto those ends. Just as there is active work by many parties in getting to that level by 2005/2006.