Home Artists Posts Import Register

Content

Hey guys,

This year instead of shinny balls and snow shaders, I decided to do a different type of video and have a monologue about the heated concept of AI :) As you might be aware I was one of its most early adopters with beta access to various tools. I already posted some first impressions few months back here

I think having these talks from time to time is super important, as any type of tool that improves our work is the future. Its been 2 very turbulent years with NFTs and now AI. Many big changes for digital artists in the way they present themselves and the way they can potentially produce content. A lot of ups and downs and still a ton of questions!

With my broken english and without scripting anything, I decided to talk about this topic. These are my personal views and I hope they are a base for some healthy discussion. Apologies as some arguments do not come across as planned and my thoughts are convoluted, but I hope you will get the idea.

Let me know what you think!

Files

overlay

This is "AI part2" by MotionPunk on Vimeo, the home for high quality videos and the people who love them.

Comments

Anonymous

And of course it's also about education, A lot of people should talk about it in podcasts and videos And I'm glad you're one of the first to raise this topic

Anonymous

I totally agree with what you said in the middle of the video, I look forward to the opportunity to work with other great artists, This is one chance to raise the level and standards for yourself. As I work for one company for years, the standard remains the same and the designer freezes in one place. I am glad to be here and learn many new things by discussing real projects

motionpunk

thanks, I am not the first though hehe. A lot of good people already raised those concerns. Ash Thorp did a podcast few days ago too among others with millions of followers in youtube

motionpunk

I must add few more thoughts to this conversation: - AI output is not copyright infringement. AI input is. AI algorithms (unless images are very same with already existing ones) produce new content in general. - The whole issue lies on the way they train their algorithms. They are breaking the law by grabbing anyone's image without paying royalties. - Think of this in a very simple way: Do you pay for turbosquid models? Do you pay when you want to use an image in a video? AI developers havent paid a single dime to anyone. They bullied their way in via coding. They just stole data in order to resell those stolen data back to us.

Anonymous

happy holidays and thank you for the good information.

Anonymous

Oh, man!! This is a long, long conversation. But from your video, I strongly agree with you about ethical issues. Not long ago, I saw many people getting Andrew Kramer's project files, changing text, rendering it, and calling themselves motion designers. I think AI will evolve to the point we use it just like we use ok google. helping solo designers with tedious tasks. Thanks for bringing this video.

motionpunk

These people will always exist hehe. I m mostly worried for the middle-small range artists that scrape for a living and these practices will create all sorts of problems for them.

motionpunk

Yes, stagnation will be a thing! If it isnt already AI, will push us to isolation as we wont need fellow artists anymore.

Misha Katz

When I first heard that AI was being used in 3D design, I thought: finally, we can model without worrying about how to UV unwrap later, the machine will handle it all. We will be able to quickly generate numerous lighting models and easily discard the unsuccessful ones without wasting valuable time... As with all inventions, people quickly find the worst and most unethical uses for them. You mentioned Picasso in your speech (who is still copyrighted and "doing well" (at least for now). Picasso once said, "The computer is a useless thing. It can only give answers." Unfortunately, some people (the greedy and unscrupulous) are satisfied with these "answers" and don't care about their colleagues, even if it means stealing work for a quick buck. It's a sad reality, but it's the truth and it's only going to get worse. Totally agree with you. Despite all the bloodshed in our history, people still don't understand that killing(!) is unethical. So to some, stealing work might seem like a minor transgression. Thank you for your tutorials. And for this one too. Merry Christmas!

motionpunk

thank you buddy! very well said and thanks for the extra info!! Art is about questions, always. Its a duty, not a task. It is way easier to mash up images, than come up with algorithms that solve real technical problems. As with users, developers also find the laziest ways possible to monetise. Sad but true, but i feel we still have voice and power. What is going on in artstation today is very encouraging!

MUHAMMAD TABISH

couldn’t agree more ! I hope Ai will help us in terms of tools and improving workflow instead of the things we’re scared of.

Anonymous

Most conversations about AI art display a lack of understanding on one side or another. Either people misunderstand how dataset training is done and what it really means (there are no copies of anyone's work), or they downplay the effect machine learning will have on the commercial art world. The answer lies somewhere in the middle, but in our current world nuance is hardly acknowledged.

motionpunk

While it is a complicated subject, there are copies of all of big concept artists' works, all illustrators, painters etc. There is a lawsuit in place as we speak, I guess sooner or later we will have moderation and all this debate will be somewhat healthier :)

Anonymous

"While it is a complicated subject, there are copies of all of big concept artists' works, all illustrators, painters etc." This is what I mean. There are no copies. That's not how the diffusion model works. An image is taken and the model with run to alter the image into nothing but random noise, then the model is run in reverse and the information on how to get to the basic concept of the image (based on its metadata, ie: dog, cat, woman, boat, etc). You can type a prompt and its very difficult to get the same image because the model is building it based on literally millions of bits of random data. Even using the same initial data in a prompt (description, seed, etc) will often yield different results. I have even tried image prompts and it gives wildly odd results. One thing I think it comes down to is working artists both commercial and fine art, understanding the difference between highly controllable and art directable results vs. the casual user who gets "just good enough" results. In the end I think this will be both a toy for the casual user as well as refined to powerful tools that artists can use for their work, for example the Blender plug in to generate textures. Are they production ready, no. Do they give you something to start with, sure. Anyway, I agree it will all shake out one way or another and I really appreciate the conversation here rather than the sturm und drang I see on other platforms.

motionpunk

"An image is taken" yes that part is clearly not allowed. Thats the whole problem right there.

Anonymous

That is certainly a larger issue. The general populace's feeling that "Well, it's on the internet so it must be free" has been a problem since Tom became all our friends!

Anonymous

"Thank you for conversing with ArgueBot 2000tm"

Anonymous

we love you thanu