Forum - View topic

No, You Can't Make Anime with AI

I agree, and yes I do think this model will be useful in speeding up animators work (provided they touch over it). I can see Toei probably using something similar to churn out One Piece episodes quickly.

But it will never replace 2D animation entirely. It can't.

it's like. a screwdriver, if you are a carpenter and want to make a living room using only a screw driver, yes, you are screwed. but that does not mean the screw driver is useless or has no place while creating a living room.

Very skeptical of the people suggesting this will improve speed by generating a base for humans to "touch up." People consistently make the same claims about MTL (much of which now relies on ML models as well), but the reality has been that "editing" or "proofreading" a machine translation is often more arduous and time-consuming than making one from scratch, for usually less than half the pay and zero credit (the real reason businesses push for it). I've heard artists make similar observations after experimenting with incorporating ML-generated images into their workflow.

There is one way to fix its largest issues, but that would literally be a case of hiring animators to review it and make adjustments. And at that point, why not just hire them to begin with?

Let me rephrase the title: You Can't Make Anime with AI using today's technology

You probably can in about 15 years.

X

Let me rephrase the title: You Can't Make Anime with AI using today's technology

You probably can in about 15 years.

Say what you want, this rock paper scissors thing was awesome.
stop having a whinge

rikku45 wrote:

Say what you want, this rock paper scissors thing was awesome. stop having a whinge

I think a lot of people, except the ones that are very anti-AI Art for one reason or another, think it's cool. Heck, I think even the people who are anti-AI Art think it's cool, and that's why it provoked a stronger emotional response.

Regardless, the whole discussion only happens because of Corridor crew's video being titled "Did we changed animation forever?" Some are more heated about it than others, because it's the internet.

But it's a fair question to ask if, when, for what, and how will AI be incorporated into anime, as well as what are the ethical, moral, and economical implications of it. Especially since the industry is, quite frankly, broken in so many ways that AI to shore up some of its issues are inevitable.

“It passes a simulated bar exam with a score around the top 10% of test takers,” OpenAI added. “In contrast, GPT-3.5’s score was around the bottom 10%.”

The figures show that GPT-4 achieved a score of 163 in the 88th percentile on the LSAT exam — the test college students need to pass in the United States to be admitted into law school.

GPT-4 also scored 298 out of 400 in the Uniform Bar Exam — a test undertaken by recently graduated law students permitting them to practice as a lawyer in any U.S. jurisdiction.

The old version of ChatGPT struggled in this test, finishing in the bottom 10% with a score of 213 out of 400.

As for the SAT Evidence-Based Reading & Writing and SAT Math exams taken by U.S. high school students to measure their college readiness, GPT-4 scored in the 93rd and 89th percentile, respectively.

GPT-4 excelled in the “hard” sciences too, posting well above average percentile scores in AP Biology (85-100%), Chemistry (71-88%) and Physics 2 (66-84%).

OpenAI said GPT-4 and GPT-3.5 took these tests from the 2022-2023 practice exams, and that “no specific training” was taken by the language processing tools:

“We did no specific training for these exams. A minority of the problems in the exams were seen by the model during training, but we believe the results to be representative.”

...
So, ChatGPT took these exams cold. That means 90% of us are now dumber than a machine

enurtsol wrote:

... So, ChatGPT took these exams cold. That means 90% of us are now dumber than a machine

I like how all this AI research is just a thinly veiled tutorial on how to cheat on your homework, the only "practical use" for chatbot AIs.

However, though ChatGPT-powered Bing can accept random questions in different ways and formats, the software giant noted some specific ways to effectively extract game details from the bot. Microsoft added that through follow-up questions, Bing could deliver more detailed insights into the previously provided information. Here are some of the query samples Microsoft said gamers could use as guides:

"What is the best Overwatch character for me?"
“What loadout should I use in Warzone 2.0 multiplayer?”
“Give me a recap of Ni No Kuni: Wrath of the White Witch up to chapter 4.”
“How do I prepare a meal based on the ramen recipes found in Yakuza: Like a Dragon?”
“Where can I find that one quest where there’s that guy who did that one thing with the fire sword that one time?” (Your follow-up responses will be crucial on something like this!)

What's most impressive about all of this is how vague you can sometimes be with your initial question. We've all had times where we attempt to look something up but find ourselves at a loss for the exact wording, have a hazy memory of the thing we want to learn about, or we find the thing that we want information on is hard to describe. Having an intelligent search engine that can ask the right follow-up questions in order to steer us to a better overall question will be a major step forward.