When’ll A.I. make next cinema hit ?

It was a year since DeepMind’s hot new shit called AlphaGo beaten the top Go player, known-as Ke Jie.
Just 3 years ago, most of AI researchers around the world guessed it would take another 15 years to compensate with any Go professional player.
So how it happened ? Well, the A.I. played 19x19 version which has 1,74*10¹⁷² legal moves, while in observable™ universe (read as universe we know shit about) there are 10⁸⁰ photons (and photons are smallest known-yet particles in universe).
The thing is, AlphaGo created its own intuition that drastically reduced number of possible moves it must imagine. Think ‘bout it yourself, when you’re drawing your artistic piece in MS Paint on daily basis, do you analyze each pixel seperately ? Of course not, you think about the big picture first, and then come to smaller & smaller details. And the more you’ve seen, the more [generally] richier your imagination is.
So, to make better comparisons, let’s transcribe our math to 2^x instead of 10^x.
1,74*10¹⁷² = 2⁵⁷² ±
With chess it’s even better:
10¹²⁰ = 2³⁹⁸ ±
What about audio tracks ?
VSauce made a shortcut and computed number of 2²¹¹ ⁰⁰⁰ ⁰⁰⁰ possible binary 5-minute tracks, so that does 2⁴² ²⁰⁰ ⁰⁰⁰ possible minute tracks.
But binary is somewhat comprimed. I’m not sure how to compute it using just Hz and amplitude, so let’s make a little experiment.
Audacity allows me maximum Hz of 384000 and amplitude of 0,000001–1 so 9⁷ possible amplitudes (4782969)
For that we would need to compute (384000*4782969) + (383999*4782969) + (383998*4782969)… finally * 60 to match one minute.
That is tad to hard even for me. Let’s do someone else do the computation.
One of songs i downloaded called C(o)unt Batula have 5,33MB and is 5:52 minutes long. That makes 5.33/(5+((52*(60/100))/60)) = 0.96558MB per minute.
1MB is 1,000,000 bytes = 8,000,000 bits
0.96558MB = 7724640 bits, so 2⁷⁷²⁴⁶⁴⁰ possible combinations.
What about visuals ? Well the question i posted on Quora some months ago luckily got a response. If you stumbled upon color pallete in Gimp or any other imaging software, you’ve seen that there are red, green & blue with 0–256 possible values. 256³ = 2²⁴ ex. So, for a 720p 30fps video we have 2^(24*(1280*720)*(60*30)) possible combinations for a minute, which makes 2³⁹⁸¹³¹²⁰⁰⁰⁰ possible videos you can with these preferences make.
But who watches 720p anymore. Doesn’t everybody watch 360° 3D 4K 60FPS videos ? Do you really think i would compute number this large for you tho ?
If yes, you’re on right address. 360° is practically a cube, cube has 6 sides, 3D consist of two layers. That makes 2^(24*(3840* 2160)*(60*60)*(6*2)) which makes revolting 2⁸⁵⁹⁹⁶³³⁹²⁰⁰⁰⁰ possible minute combinations.
8599633920000 = 10^ 12.9 ± so let’s account 10¹³.
Titan X GPU have 12 teraflops, so 1.2*10¹³ flops. If we somehow turn each of these FLOPS into a qubit, we could simulate every last movie combination with the preferences shown above. But movies nowadays last around 90 minutes, not just a minute. That would do 2⁷⁷³⁹⁶⁷⁰⁵²⁸⁰⁰⁰⁰⁰ so would need 10^ 14.88 ± FLOPS being turned into a quantum machine. Today’s 500th fastest “supercomputer” has 430.5 teraflops, that would do it and still it’ll be 2.8 ± times faster than needed.
Now you’re probably expecting some kind of end, some kind of veil retraction.
Nah, not today.
Edit: Okay, apparently there are 8 possible notes within 6 octaves, American album songs have minimum of 2,5 minutes and at average maximum 10 instruments. That makes (8*6),((1+2+4+8)*(2.5*60)*(10)) possible musical combinations, so 48²²⁵⁰⁰. Now this number is too big to get converted to 2^x system, at least by my regular tools.
Edit 2: Based on this article there ± 10⁵⁰ atoms on the Earth, so if you would want to do your own Earth-sized planet consisting of apparently 7 the most-widely used chemical elements, you’ll have 7¹³³⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰⁰ possible combinations to explore your full creative potential.