I would say we won't see that many TFLOPS for 15-20 years.

A decade? I'm going for 20 years at least. A decade ago what was the best we had? The Radeon 9800 XT with it's 40 GFLOPS or so? And now AMD's best (single GPU) is the 7970 with 4300 GFLOPS. That's like 100x in a decade. Really far from the 1000x or so to get into the 5000 TFLOPS range. I tried to find the FLOPS performance of the ATI 3D Rage to compare it against the Radeon 9800 XT, too see the improvement in that decade, but I failed.

The point is: Computational power is growing at a slower rate now. It's gonna take A LOT of time to achieve the computational power Sweeney mentions... BUT I think that "reality" can be achieved with a far less than that. We just need a developer that's willing to invest it's time on that.

I don't know...I see a lot of today's gaming sucking a big one due to this attempted "realism". Realism is cool and all but 90% it's at the cost of gameplay.

I'd much rather see 5000 TFLOPS invested in making amazing AI or coming up with crazy gameplay and set pieces than visual realism. My life is real enough, I need an escape from realism when I'm gaming.

Aproximately over 80% of what your brain processes comes in through your eyes.  So it would seem logical that games place an emphasis on visuals. They don't have to be realistic to be delicious.  Just look at Limbo or Braid.  They both have stunning visuals.

Can you give exampls of good and bad AI.  I know some bad ones, but sometimes it looks more like programs thought 'good enough' or did as best they could with the talent and time crunch.

What do you see as the best AI?

the more forward we move the cheaper the assets used to make games.

nowadays even people with no tools can make games with freeware software and they're "Full HD , high end graphic....etc etc"

the idea of development costs rising due to technology rising is backwards thinking. 

While i dont know enough about how the brain processes information to agree or disagree with your first statement, I will say that for myself and gaming in general, gameplay engagement has much higher value than stunning graphics. Braid was awesome because of the gameplay mechanics (and music) to me and while graphics were nice too, I wouldnt care if they werent. Fez, Super meat boy didnt have such a visually pleasing art style and were both awesome in my books.

On the other hand, when graphics take over and game strives for a cinematic approach with no gameplay to fall back on, i tend to hate it. When Kratos goes off on his cinematic tantrums in god of war, and my involvement is reduced to tapping a button every now and then, i become bored out of my mind, no matter how awesome those cinematics are.

As far as AI, any game that involves a gamer in to beating an enemy with an engaging and challenging gameplay, regardless of difficulty, I would say has good AI. I find that most games these days try to do too much and dont have a focused enough gameplay and gameplay flow in order to hide  how bad the core gameplay and enemy AI really is. Something like Resident Evil 6...its just a bloody mess gameplay wise. And why is it a mess? Most likely because they spend 80% of the time making sure it looks awesome...

As far as good AI, something like Geometry Wars 2 has awesome AI as do most Mario and Halo games... any game that throws a bunch of pieces on to a blender and pushes the on button and lets the gamer figure out and choose one of many ways to beat the challenge at hand is awesome in my books. On the other hand, The scripted gameplay AI which was invented to provide a more cinematic experience is not cool at all...

Dont get me wrong, Id love to see a game that has both, visual and gameplay engagement, and they do exist...but are becoming a rare breed.

How does the PFLOP/TFLOP conversion work again?

trying to figure out how much this compares to the Titan.

Tim Sweeney, one of the few guys that actually don't say shit unless they know what they're talking about.

Probably be another 3-5 years before a 500 man team with sizable render farms (Let's say WETA Digital) are able to do real photo-realism, a lot is still semi-faked for obvious time and cost constraint purposes. I don't particularly need photo-realism in my games though.

Enough is enough, give me 2014/2015 level graphics at a stable framerate, and start focusing on massive worlds and exciting new gameplay after that, I want my planet to space+vice-versa flying combat in Battlefront 3, and other fun stuff like that.

Never the less, fun topic to discuss.

I can't get upper index to work, so:

Meanwhile, games tend to get even more boring with ZERO innovation that truly brings the industry forward.

I hope someone creates something interesting in about 2 or 3 years.