https://www.youtube.com/watch?v=E3LtFrMAvQ4
Hook up the camera movement to a phone's gyroscope in real time, add a few post processing filter, 3D ambient audio and it'll be pretty close to what you describe.
That's the problem talking about stuff like this. These renderings are all amazing -- they're the best of the best. But we're talking about realism indistinguishable from a camcorder, which is simply impossible given our current technology. (After all, we haven't figured out any techniques to do it.) So it's not enough that it looks pretty good: if you set up side-by-side footage with a real-life scene, viewers would immediately identify the real footage.
Watch Cloverfield, there's a ton of CGI involved dumbed down to camcorder quality.
Every single comment moves the goal posts. It's not clear what your standard is. Would you mind specifying your entire criteria for realistic graphics in one place?
Why is it impossible? Not having an example doesn't mean it's impossible, it just means it hasn't been done yet. Can you back your claim that it's impossible with a proof? Why do you think zero techniques we have now will apply or be involved once a video that satisfies you is produced? Considering that can rendering realistic looking still photos, and realistic video in constrained situations, I find it puzzling to assume that none of it will carry forward.
Show a shuffled playlist of 20 videos, half of which are simulations. The simulations must be entirely synthesized; no mixing in footage of real life.
As long as the videos are non-trivial (~60sec and reasonably complex scenes), the viewers will correctly select all 10 of the simulated videos. If we were capable of generating simulated video identical to a camcorder, people would score no better than random guessing.
So not only have we failed to reach that goal, we are nowhere close -- it's effortless to identify all the fake and real videos. Especially nature videos, for example. When put side-by-side with real-life footage, our best simulation attempts are nowhere near the same ballpark. They're not even the same area code.
The idea that we really haven't accomplished this may seem offensive. My frustration is what led me to work on it. But unfortunately, we're nowhere close. I'm not sure traditional techniques will be enough to close the realism gap.
I've outlined a scientific test that anyone can conduct. The results may be discouraging, but they establish our capabilities. The goalpost is rigid.
To answer your other questions: Obviously everything is impossible until it's not. But I never meant to claim this was impossible, just impossible with our current techniques. I think physically-based rendering circa 2010-2020 will turn out to be a failure from the standpoint of replicating camcorder-grade video, but that's speculation. When you try to simulate physics in such great detail, you end up with hundreds of approximations. When these integrate together, you get modern Unreal Engine videos: they don't look real. They look pretty great! But not real.
Who cares about looking real? Well, like AGI, it's a goal worth striving for. Ever since we've been scribbling on cave walls we've tried to capture realism. Unlike AGI, there isn't much use for perfect realism; it might not affect the world at all. But I'm pointing at the summit; wouldn't it be a fun climb?
I think this is way too vague. You have already talked about camcorders and phones: handheld cameras. You are clearly assuming moving cameras and moving objects. Be more specific, what needs to be in these videos before you'd accept them?
> the viewers will correctly select all 10 of the simulated videos
Maybe, you're predicting the outcome of an experiment nobody has ever done. And I would predict that eventually some, and then later all videos could be mis-predicted.
If it's not possible, then you should be able to demonstrate it's not possible without doing the experiment.
> I've outlined a scientific test that anyone can conduct.
You've outlined a test that can prove that we got there at some point before the test was conducted. It cannot show that we're not there yet, and failure to get a positive result proves nothing. The test you outlined can fail and we can still have all the technology we need to create realism, both can be true at the same time.
If we can't fool viewers, we haven't achieved realism. And the only way to check is to put some nature videos side by side with our simulations and ask "Which one is real?"
https://youtu.be/Bey4XXJAqS8?t=1089 will kick the crap out of any simulated video. No contest.
I agree that it's not a happy thought, but it is objective truth. On the other hand, the moment we reach that goal, the world will notice. Or at least Hollywood.
I'm objecting to your claims that it's not possible. We can't check whether it's not possible. We can only verify that it has been done, after it's been done.
Nobody has done the experiment you've laid out. Not having examples that contradict you is not evidence that it can't be done. The criteria and claims you made are yours to defend.
I also counter that light transport, and specifically color multiplication, are not the problems getting in the way of realism.
It seems like you're intentionally discounting and ignoring progress in constrained realism. We have achieved some CG video that is indistinguishable from realty. Not hours of alligators in Thailand, or of people talking, but short videos of fruit bowls and architecture for sure. The dominos are falling, one by one.
I think you are making a classic fallacy of 'argument from ignorance'. https://en.wikipedia.org/wiki/Argument_from_ignorance
Your proposed experiment cannot tell us whether current techniques fall short.
The experiment can only demonstrate when we are there, after the fact. It cannot demonstrate that we won't get there on our current heading, and it can't even demonstrate that we're not already there.
This is simply false. That's the whole point of the test.
Read the article about argument from ignorance. For your test to work, you'd have to guarantee it had the best CG that's possible. You can't guarantee that. You can never prove we're not already there because no matter what CG you make for the test, someone else can always try harder, take more time, use better artists and bigger computers, spend more money. Abscence of evidence is not evidence of abscence.
Your logic is faulty, your argument is suffering from a fallacy. You can speculate that your idea of realism isn't possible yet, but no test can prove it.
You are completely misunderstanding this topic area. We already know how to compute ground truth references based on the real physics. I suggest starting with the book Physically Based Rendering.
The question now is, how close can we get a real time approximation to match unbiased path tracing? The OP's link is a pretty definitive step along this direction. But even if that whole research agenda fails, so long as we keep getting more computation at a cheaper price, even unbiased renderers will become real time.