So, what, you’re training the AI on successful films to identify the common elements, then taking a genetic approach to refine the results? Again, that seems to dramatically underestimate the importance of structure and, well, craft. Having two people walking along a beach holding hands can be done well or badly. Even done the same, it can work or not work depending on what preceded it. And that’s before you get to exogenous factors like star appeal. Studios have been trying to come up with a formula for success for over 100 years and, especially in comedy, they’ve failed spectacularly.
Basically, absent the animation automation part, I’m not sure exactly what the problem being solved here is, assuming it’s actually solvable. Writers aren’t that expensive.
I assumed the animation was going to be stop motion photography, not CGI, because the ad-hoc amateurism of real life brick films is half the charm. Maybe you can crowdsource the animation, put the brief out to every scene and hope for the best?
The lego CGI movie was a big hit, but they had to go to incredible detail to make those bricks interact and look like they would IRL, which was deceptively expensive and difficult, with a massive crew. You can do cheap simple CG bricks, but it’ll look too machined, charmless and sad compared to RL, and to what is being pumped out by the established gaming and film corporations. And then, unless your AI is surprisingly nimble, you risk alienating an audience when they realize you’re not even putting in effort to support your dialogue with crafted performance… no one wants to have an audio book read to them by Dr. Sbaitso, even if the book is genius.
Obviously, if you’ve created something mind blowingly game changing, then all bets are off (insert scepticism here), but the entertainment landscape is littered with producers who were charmed by technology shortcuts to find out there is no silver bullet to crafted material, unless you make the dialogue and story so good we don’t care about the shitty background art (SouthPark, Dinosaur Comics, XKCD, etc). And/or if that silver bullet cost a fortune to make.
Perhaps I’m not explaining it. Anyway, an IP lawyer has asked me to stop discussing details in public. You’ll either read about it in the next year or not, I suppose. Thanks everyone. You may resume your cynicism about innovation being possible elsewhere.
Virtually all AI used in video games is fairly trivial. Generally, it’s fairly simple heuristic rules, and hard-coded scripted sequences of actions. Generally, what makes AI look “realistic” in most games tends to boil down to the human mind’s willingness to anthropomorphize observed actions in others. It stems from our empathic ability which plays such a key role in human interactions. When you see another person doing stuff, it triggers some of the same pathways in your brain as if you were doing it yourself, which makes you feel like you are doing that action, which then triggers many of the pathways associated with the cognitive motivations which would lead to that action.
When we see characters in a video game performing actions that LOOK like actions a human would take, we have a natural inclination to feel as though there is some sort of underlying intelligence driving those actions. But in most video games, there is nothing of the sort. It often boils down to something as simple as, “If X occurs, play animation Y.” It’s the artistry in the animation which ends up “selling” the realism.
For some of what you’re describing, you’re going much deeper into what I would call “real” AI, which involves more complex machine learning. What’s more, you’re kind of talking about learning things which involve a high degree of subjectivity. Defining what makes a “good” movie is a seemingly impossible problem to really solve, given that any two humans will give you a different answer if you ask them what the best movie is. Computers are good at concrete, quantitative measurements. They aren’t so good at subjective qualitative measurements.
Not to be arrogant but perhaps the entite field of A.I. research is going about it from the wrong direction. They are trying to get the computer to have meaningful conversations about love or something. My idea may be radically from the other direction. I don’t care how smart the AI is and am not educated in that field enough to program that. All I care about are end results and making money. Thus the quick and easy approach that works.
Sure, this is kind of in line with what I mentioned above with the types of AI you see in commercial games. However, I suspect that even the simplified description you have here is still going to involve some tricky wickets. That is, I’m not sure there is a particularly “quick and easy” approach to what you are looking to do. There are relatively quicker and easier approaches, but even those may involve some seriously complex issues that would need to be resolved.
I’m not being cynical about anything, just curious. And, in the absence of detail, skeptical. All you’ve really said is that you have (or may have?) an algorithm that makes writers and animators obsolete. I mean, I can totally understand not wanting to divulge all the details before you’ve locked it down, but you were the one who posted the thread, so you can’t blame us for asking questions about it.
Probably the main thing that I’ve learned working for tech companies is that ideas are relatively easy and cheap, while execution is difficult and expensive. I think that’s where (aside from just being curious) a lot of these questions come from.
That being said, best of luck! I hope you can come up with something useful and profitable.
No, no you misread me! Sorry. This is science, right? It should be studied and critizied and picked apart and dismissed if not useful. I’m just not qualified to do that so I’m going to have some guys at UCLA look at it.
You should really go and enjoy the last ~8 pages or so of the Star Citizen thread, because DSmart has been going–as the metaphorical kids say–hard in themofuckin’paint on this tirade for a good while now.
P.S. - ElGuapo, I think everyone may be trying to politely tip-toe around the fact that over the course of the three pages of this thread, you seemed to transform from hopeful entrepreneur exploring the benefits of being a Youtuber on the side to typical-LA-producer-type-with-end-stage-sun-madness ;-).
I think, like you said, a good bit of that seems to stem from just not using the right words/terms to convey the concepts you’ve got rockin’ around that ol’ melon of yours, but there were a couple of points where my eyebrows went skyward. Like the others have said, best of luck with it–no matter how it pans out or what form it takes, your initial Daily-Show-meets-Rick-and-Morty-meets-stop-motion-animation pitch sold me hard, so I can’t wait for you to show me what you got.
Now it sounds like you found an artifact buried in a smoking meteor crater! But yes, if you’re moving home and dropping your old life for it, I’m sure it’s cool. Good luck!
Weirdly, my phone came with an app that makes aesthetically pleasing (to me) slideshows each week out of the photos I take with its camera. Whether those are the optimaly aesthetic slideshows I don’t know. No idea on the algorithm.
At a guess, aesthetically: vivid colors, high dynamic range, good focus. Other ways to choose: weight pictures out of groups that are taken in quick succession. (Either you’re taking a few pictures of the same subject trying to get a good one, or you’re somewhere you’re aiming to document extensively, which is either pretty or meaningful.) Do similarity comparisons so you don’t get the same one twice. (Or at least not twice in a row.) Maybe find some way to characterize exciting angles, but that gets a lot more into computer vision than any of the above.
Google has something like that on android phones, that pops up as “auto-awesome”. They put together neat little “scrapbooks” for trips where it tracks that you went somewhere, took pictures, and bundles them together, and has a little map-travel visualization. It’s neat.
I assume a lot of the image selection is just face detection, and noting when you have a lot of pictures in close proximity. If you take a bunch of pictures that are very similar (i.e. exposures of the same scene / group), it’ll even group them into an animated gif for you. I took a bunch of pictures of my daughter on a rocking horse, and it animated them, the result of which was very cute.
It also does some image detection stuff to apply effects. I took a picture of a Christmas tree, and it applied the “sparkle” effect, presumably because it detected a lot of pinpoints of light (and probably would have applied the same effect to appropriate outdoor nighttime shots or starfields)
I used to be an exec assistant in the music industry, on the labor side. We used corestaff to grab random office dorks to filter through (maybe 30 people, until we found a decent person to replace me)
BUT - my boss intentionally didn’t want anyone who was looking for industry work or gave a shit about famous people etc… (we had access to everybody’s everything, SSN, phone address, so getting a non fan, non musician was critical)
Also, congrats! if you are ever in Vegas drop me a line. Or if you ever need sculpting, molding, casting or miniature sets.
Yep,about a year and a half now, lemme know if you come up! Beats the shit out of SC and so much cheaper than when we were in L.A. Ill be in pasadena in nov, for designer con, but probably just for the con, i wont have a car on me.
This reminds me. One of the UK gaming podcasts (Crate & Crowbar? Daft Souls?) recently had a throwaway comment about a paper on using machine learning for animation from what the podcaster thought was UCL. Does that ring any bells?
Anyway, the AI idea is cool as hell and I’ll revisit, but for now it’s a distraction of my main goal, the animated series. I’m concentrating on getting that off the ground now. We’re concentrating on old school Ray Harryhausen stop motion animation, which a computer can’t do.
Met with some animators and actors this weekend in Burbank, which is where I guess a lot of the studios are. Super energizing and fun conversations. I have to learn all about this SAG thing. We even discussed the impending videogames voice over strike. I even met the guy who created that weird scary creature with eyes for fingers in Pan’s Labyrinth. I didn’t explorer much after (should have but I don’t know LA at all and it was like 104F out) but did see a water tower with the Warner Bros. logo on it during the drive out. Hollywood!
In other news, I’ve convinced my wife that if we win the contract today to provide the U.S. Army with a giant laser, I can use all the profits to fund getting some episodes of the series done. Yay! Now I just have to organize all of these talented people together.
