Hey guys, I was looking at this
https://shapescale.com/Its a device you step on and scans you and makes a 3d model of your body. I believe at this moment we are at a modern industrial revolution where machines like that are being developed, and the best of them will stay and get improved over the years.
But I was wondering...
How far do you believe this can go? If we fuse this technology with augmented reality technology, we can create characters or every type of humanoid by scanning one person. Do you think this could expand further into creating everything fast? Like speedtree or something?
Replies
Few days ago I had an awesome idea for neural ai concept art tool. Just imagine if you feed a lot of photos of things to neural ai and it will automatically combine them in such patterns that it will look like photobashed concept art.
If anything you make has a defineable ruleset, this can start to be automated.
For a long time to come though we'll still need talented artists to make customized content.
It's not like game art jobs are highly at risk, stuff like self driving cars are much sooner on the automation timeline, and even that is still 10-15 years away from actually taking people's jobs.
I think machine learning will be used more, but still requires a fitness parameter. Maybe you could make a model then the AI will make hundreds of variations of the same model.
If we automated the shit out of the pipeline, we would get something low quality.
But IMHO it would be in the realm of 'good enough'.
- Scan
- Zremesh
- Automatic UVs
- Bake.
This would look crappy, but i think it would still be 'acceptable' by mass market standards.
The main thing about this pipeline, is that it would simply get better in quality once we get more power and better algorithms in place.
- Higher fidelity scans
- Better Unwrap algorhythms
- Better Remeshing algorhythms
- Reducuing issues with higher poly/texture budgets.
In 4-5 years we could very likely have games that rival some of the 'average' looking games today.
2nd thing is:
I think the worst looking console games today still look pretty good. I sift through the metacritic scores in the 30s for games released THIS year, and the graphics for most of these games aren't bad at all.
IMHO The gap between the worst looking game and the best looking game is a lot tighter than it's ever has been. The reason this is important is, I honestly think the proceduralization of a lot of games is going to overtake most "realistic" games sooner rather than later.
Also, the cost of automation is pretty much always less than the cost of employing people on a long enough timeline.
IMO, it has existed far too long in CG and with the photorealistic pipeline being longer than ever, it's just an unnecessary crutch. Unlike modeling, texturing or even lighting, you never see UV maps being paraded on screen in movies or video games. No one would know or tell the difference had a robot unwrapped a model perfectly versus an artist who had to spend an entire day doing planar or pelt unwraps, getting perfect texel density or laying out a uv island so it looks tidy.
That's all I want out of modern R&D.
http://ptex.us/overview.html
I dont understand it, but I guess wishful thinking and fear is a powerful driver.
its only a question of time, and I think there is gonna be a real fun sweet spot pretty soon, when you can automate all the boring things, but still get to do the fun stuff. you know, right before the computers get better at that as well :P
as for mesh creation I would say focus on being a good designer, photogrammetry is also going up in demand not only have I ran into companies wanting photogrammetry skills and when it is coming to photorealistic AAA games it is essential.
the future or even the present in growing AAA circles should have western artists (as tragic as it may seem) focus on optimizing and deisgn, because more often then not your going to be working in mesh cleanup for outsourced art.
Unless something radically changes this is largely going to be a continuing trend and the future of jobs in the industry.
Can they be as robust and flexible as you will ever need them?
That's a pretty tall order and it will cost a lot of time and resources to pull that off. That's the rub, if the cost of automation is higher than employing people, then not that many people see much business sense in pursuing it, unless your goal is to sell it to other people who in turn won't hire as many people. Still getting over that hurdle and getting people to adopt something new can be a daunting task.
Is it fully tested and iterated upon under a lot of different production settings? To the point that you won't need a team of humans to come in and save the day when it can't deliver? Are you willing to stake your game on it?
It's something most people who write scripts to automate mundane tasks take into account. Will the time I invest in writing this thing, will it pay for itself or will I spend more time writing it than the time it saves.
You can also waste a lot of time trying to future proof your life only to have something you didn't think of, that wasn't even around way back when you where planning, sink your your ship. It's really smart and plan ahead and think about the future, but don't let it paralyze you and keep you from taking a risk on something that could be really rewarding, even if it only lasts for 5-10-30-40 years.
That's my 2 cents, I've been in 3 industries now that have faced the unrelenting march of automation and it's brutally terrifying, until it smacks you in the face and you deal with it, then it's not so scary. You pick up and move on and try to find a way to use the skills you have to do something new.
And things like sci-fi mechs, or crazy ancient temples with bizarre wall paintings and statues, I'm not sure how any kind of real automating could be done. Without, of course, really cool inputs.
https://m.youtube.com/watch?v=7Pq-S557XQU
But yes. The idea is that even creativity can be automated, as they're still governed by quantifiable rules.
Stories, music, art, world creation will all be automatable (very likely sooner rather than later).
its a bit scary
Right now we have tools to procedurally create and assemble geometry, textures, and animation (rag doll physics, and the stuff being done in Overgrowth). Basically as long as there are very clear rules that the software can understand, it will be one of the first things to be automated.
Next, we have artificial intelligence doing some crazy things like delighting textures, creating full 3d models (nothing close to what a human can do, but starting to get there), and animation based on actors in a physical world (picking up boxes, walking and jumping, etc). I think this level of automation in production is coming in the future, but I am not sure if it will be in our lifetime or not.
Lastly I think eventually the whole process will be automated. Any sort of sameness can be trained away, and I wouldn't be surprised if given time, the computers will produce better art and games than we ever can. But at that point, I don't think there will be any jobs left in the world. I'm betting this will start once we all retire or later.
So, being afraid does make sense, but I don't think you'll be out of a job because of your robotic coworkers/ overlords any time soon.
https://www.engadget.com/2017/09/26/microsoft-new-coding-language-is-made-for-quantum-computers/
What the shitty computers did for the analog world in the 40's-50's, we will see a similar leap forward with quantum computing, we are in the stone age of computing.
Lets take a theoretical leap forward. What if we where able to take a 4k static image and calculate all of the possible pixel color combinations to make an image. Out of the nonsense you would also get every picture ever taken, every movie ever made, a complete history of anything that ever was or could have been. Just like how it's been an impossibility to calculate pi we might have a machine that churns on something similar but is image based.
Extrapolate that to voxels and let it churn. Imagine giving it a narrower scope to focus on.
Quaduped Mechs with an 80's theme, add options for shoulder mounted rockets and jump jets.
Remove superfluous detailing, but add access hatches.
Here are 2,000 mechs to choose from.
Give me 300 rounded versions and 300 sharp angular versions of these two.
Apply mechanics and physics to figure out which 100 would function the best.
Here are 100 mechs to choose from.
What if future creative professionals are just really good at narrowing the scope and getting it to spit out a narrow band of options to choose from?
Maybe they iterate on those options, until those iterations can be quantified and reproduced.
Maybe that is what thumb-nailing looks like in the future.
It might not happen this decade or in the next 50 years but that's the road we're on.
capitalism will ultimately fail once everything is automated, maybe we will end up in a unified world, everything is peaceful, everyone is happy
but i fear this will end in total chaos and everything goes to shit...
"Where else can you wake up 2000 workers at 1am and have them stonewash 50,000 pairs of jeans with only 5min notice?" It's not a mega factory running on automation, its in a country with questionable human rights. "Why are we rubbing rocks on pants? Because it's a fashion thing now. Don't ask questions just take your rock and rub it on the jeans."
Maybe some day it will be automated and it will handle all of the weird case scenarios they can ever think of. Acid washed, rainbow tie die, no problem. Then along comes tinfoil shirts and that factory is useless. Up pops a new one with 2000 workers and later someone automates that one, but the cost of creating automation is a high hurdle.
Is it possible to automate any process, yes of course you can. But is it cost effective to go through the process of automating it? That is a much more complex question, that unpacks a lot more questions that need to be answered.
saying the things you say either proves wishful thinking or un-awareness of how much improvement AI has been doing in the last few years.
I mean, our jobs will probably be among the last once replaced, but replaced we will be never the less. Id really encourage doing some AI research, its really interesting and at the same time scary what its achieving lately.
I for one embrace our robot overlords.
two hundred years ago 90% of the population was in farming on a capitalist basis. did mechanized agriculture cause capitalism to fail? no. those who owned the capital were able to leverage it into greater output. those who sold their labor found more profitable industries in which to do so. many of those industries were not even fathomable before this. (we make 3d art for video games on computers. explain that to an ancient)
technology does two things: lower production cost, extend production chain. make more things, make new types of things. books used to cost a fortune, now they cost a fraction of the value of one hour of work. cars used to not exist, today you can be a professional driver, tomorrow a robot will drive you around. none of this does damage to capitalism.
the key thing to remember in these technophobic times is that scarcity is not going to disappear because of some project google puts out next friday. nothing about automation will obviate labor-as-labor, or capital-as-capital. people will have different kinds of occupations, the standard of living will continue to improve, but the idea of fully automated luxury communism is just a fantasy. resources are scarce and human wants are unlimited
its a really interesting video on AI and how its gonna change everything in the future.
actually, everyone should watch it, its a good video, but especially if you dont believe that AI will take over everything.
and to clarify, personally, I still think us humans will "work" and you know, DO things.
but we will do things for fun rather then the need to.
https://www.youtube.com/watch?v=7Pq-S557XQU
I think Amsterdam's post sums up my opinion pretty well.
200 years ago, there wasnt even 1 billion people on this planet, food was so hard to make it needed far more people to feed far less people than 100 years ago than now. I get your point, people are fine now, despite wars and revolutions also caused by some of those changes, there is plenty of work now. But now we have more than 7times the population than 200 years ago and 50% of jobs in the west are at stake. Not ours in the entertainmentfield for a bit, but still a lot.
And it doesnt only hit us here, Foxconn (the guys producing our iphones and other gadgets) replaced 60k until now needed workers by an automated factory.
Capitalism is based on people working, to earn money to buy stuff other people made or offer. And great that you brought it up, its also based on growth which is, as you said, limited. This stuff will cause some backlash.
Now if you get your hair cut by robots, your books written by ai, your taxes dkne by ai, your cabs driven by ai etc etc, this chain breaks. companies make money by supplying these robot/ai services, but how do people pay if they dont have jobs? Not everyone can work in the entertainment industry and model for games.
Given that a lot of our work will get automated as well. Heck right now you can watch how things that used to be jobs in our field, as those things took time, crumble away. Given that there might only be need for artdirectors to feed our ai replacements, how do the grunts replaced by ai earn the money to spend in the system? Right now the exploding need for more content saves us for a bit, but it also drives tech artists to replace many of us with tools
of course its a fantasy, a pretty grim one, despite the shiny look, but one of the possible outcomes the cool and crazy and scary ai research could lead to.
And I believe same thing will happen in other industries. With self driving cars we will need specialists who can program these cars and maintain them and so on.
while the more automation we have , the more indie and small studios popping out.
sure some job will disappear in near future, but also automation will bring new job, giving people chance on creating great thing in low cost.
nobody want to go back to the time where they have to create their own programming language in order to make simple IOS games.
some of the stuff that I like :
this program will color your sketch
https://github.com/lllyasviel/style2paints
this much more advanced, where art direction can be created by sentence of words, and let A.I . process the photorealistic result.
https://www.youtube.com/watch?v=rAbhypxs1qQ
your arguing in the old way of thinking, and its simply not working any longer. there is a difference between changing an existing system and completely breaking it.
right now we have a stable echo system that is 100% dependent on the part that work = money. robots dont need money, and thus the entire system falls apart.
I see where you're coming from @Amsterdam Hilton Hotel , but like others have answered, machines seem to break the chain by not even needing maintenance. I guess only time will tell.
Citizen pay/basic income might be one solution. when you have say 80-90% of the population not working, socialism in some form might be the only solution.
the real question should not be: how will we get work in the future? its what do we do when nobody is working in the future?
yep, that will definitely help as well. Im not sure if it will decrease dramatically, but it will definitely be smaller.
here is an interesting video about it btw
https://www.youtube.com/watch?v=QsBT5EQt348
With the way things are, currently, people are able to have children and live a normal life, because they do a job. Even if it is not paying them well enough, it is still manageable to survive.
In a future where automation takes over those jobs, it can easily reach numbers like 70% of the worldwide jobs being done by robots. Robots do not require salary, they do not get tired, and even if their cost is HUGE, it might be more benefitial than humans in the long run. If you put this variable into the futuristic "stabilised" population growth, it will have a huge impact on the decline of human species, because if people barely have money or no money at all to survive, how are they going to make it while having children?
Terminator style machines, an't happening for a long, long, long, long time if at all. Whether you believe in the Bible or not; Jesus hasn't even come back yet, yes he is suppose too.
Anyhow Automation like Terminators or what you saw in the video is well; hogwash.
When Jesus comes, he'll be bringing automation while taking a piggy back ride on a T-800.
I mean no one actually believe half life 3 will ever be released right?
https://github.com/Unity-Technologies/ml-agents
https://developer.nvidia.com/deep-learning
http://pyml.sourceforge.net/
You talk of "50% jobs in the west" being at stake. But jobs aren't static at all. The jobs we have are only a secondary dynamic, arising from the material conditions that happen to exist at the moment. The resources of the present constrain the jobs that people do in the present. Nobody made game art in 782 AD, nobody fishes in Nevada.
Conditions are always changing. I say it again: 200 years ago, 90% of us laborers were involved in farming. Well, that changed. Turns out 90% of the jobs in the west were at stake! Surely capitalism should have collapsed - by your own reckoning, if those people couldn't work those jobs anymore, they would have had no money to buy goods and capitalism should have ended.
But capitalism didn't end. It responded how it always does. Capitalists invested their capital differently. Laborers remained employed - in different occupations, whose production cycle was more appropriate to the ends and means that existed at the time.
Cheer up, friends. Technological improvement can be scary, but it's inevitable. The economy isn't frozen in equilibrium, and it never has been. Embrace change. Position yourself to profit from it.