72 comments

[ 1.6 ms ] story [ 139 ms ] thread
(comment deleted)
Squiggly lines and generative artifacts would match the story line of the season of "time slipping" however. If it is not AI, I can see a designer giving a Dali-like style for that reason.
It's true that in this case, the artifacts are pretty fitting to the show.

But they are clearly diffusion-model artifacts.

Yes, give a big studio the benefit of the doubt. They have great track records of not exploiting the industry, and are usually fair conscientious actors that provide unity and financial stability to all the artists and creatives they empower.
If anything, in this case it appears that a contracted artist made an error.

The bigwig studio executives don't composite their own ad posters.

The article mentions it was copied from a Shutterstock photo, and if you go to the creator's page you can see they produced a series of Escheresque clocks, and this is the least obviously AI one [0]. The others have nonsense symbols, repeating numerals, and also just feel like AI images.

[0] https://www.shutterstock.com/g/Svarun

I think it’s pretty clear there’s AI in it. But it also fits in perfectly with the show, so this usage is appropriate.
> A Loki season 2 poster has been linked to a stock image on Shutterstock that seemingly breaks the platform’s licensing rules regarding AI-generated content.

A human-made poster uses a stock photo for the background, which is suspected to be AI.

This is such a non-issue.

Hollywood might well spend a decade fighting this AI straw-man, with increasingly kooky demands. How stupid of them :(
From the crowd earning a living with copying code from SO and tutorials, I wouldn't expect they see any issue with copied imagery. Let's chat again when AI does some proper coding.
It really is a fundamental difference in goal, for it is the duty and objective of a dedicated software engineer to replace themselves.
this made my monday morning, thank you (i am a member of the society you describe however i too appreciate the irony you have pointed out).
The end game for AI generation technology is almost certainly not that it will pit studios and IP owners versus artists so much as it will pit artists versus artists. The copyright issues become moot once several companies pay a lifetimes fortune to a thousand or so artists for exclusive and complete rights to their collections to seed their algorithms. And I can't believe for a second that professional solidarity is going to extend to every talented visual creative on the planet.
> The copyright issues become moot once...

Only some of the copyright issues. The issue of whether AI creations are protected by copyright, for example, would remain.

If AI is cheap to use, that category of protection becomes irrelevant. A studio can crank out content so fast that it's irrelevant if someone else plagarizes, remixes, or reshares it.
The studios will lobby heavily to prevent that. If content becomes that abundant how could they charge so much for it?
Mmmm. So I don't know how many people are old enough to remember the tale of how iTunes, for a short time, completely owned the music market.

Long story short is that they played ball with every demand the recording industry made regarding how selling audio clips would work when other nascent music store experiments wouldn't. Industry gets to dictate quality? Got it. Online store releases after physical releases? Got it. Encrypted? Strongest stuff the government will let us put in.

And then one day... Apple just stopped selling the music encrypted. The recording industry had a bloody fit about it. And Apple's response was "Well, customers hate it. And honestly, who else will you do business with? You gave us all the contracts. What are you gonna do, found your own software companies from scratch to compete with us? If that was on the table you'd have done it already. You don't know how." It took a little while for streaming alternatives to crop up, but it was amazing to watch Apple call an entire industry's bluff.

All this to say... Wouldn't be the first time a content gatekeeping industry's myopia let it become the instrument of its own unmaking. People heavily invested in the old model have some of the hardest challenges fully realizing what new model novel technology can enable.

I predict the Disneys of the world will discover that their memetic monopolies become badly challenged in a world where content creation is so cheap that, while of course they can build their staple of wholly-owned IP for AI generation... So can everyone else, which means the shape of the competitive field will change, close to overnight. And if they try to compete by standing in front of the tide and bidding it to not come in, they'll eventually be cut down by someone who promises to give them everything they want... And then takes their empire from them.

That's a nice example of how it can turn out, but it's not always like that. Land is abundant, but in many places housing can't be built. Water is abundant, but Nestle still makes a fortune selling it. If someone could find a way to sell air, they'd apply for a government sponsored monopoly (a patent) on it, and lobby government agencies to recommend it over the natural air.

Far too many people make their fortunes gatekeeping access to resources.

Sometimes a news story is really an advertisement.
I feel like all stories are advertisements.
I had to read the article carefully before I got it. It wasn't Disney that was using generative AI. One of their artists used a licensed stock image for a background element (which is 100% acceptable thing to do!) and people are complaining that this background element from the stock image company was AI generated.

"Disney" isn't the culprit here!

100% acceptable

Acceptable to whom? Consumers? Or Luddites and a guild whose interests aren’t entirely aligned with consumers?

I think they were referring to using a licensed stock image in a background element as acceptable, not the stock image company using generative ai.

I am no fan of Disney, but they appear to have been doing everything by the book.

(comment deleted)
Even if it was 100% AI generated it would still be a non-issue
Holy moly this article is indeed terrible. I figured the outrage was because they were knowingly using an GenAI image that was trained without licensing or something along those lines.
Ironically, articles from The Verge increasingly feel like they were written by a particularly inept AI.
Unbelievable! Cancel the artist on Twitter! Shame them! Crucify!
This is ironic. If you had to pick a patron god of generative AI, it would definitely be Loki.
That was actually the plot of an old X-Men comic. Loki came to earth and started granting wishes. Basically turning regular people into Forge-style creative mutants. One was a master architect who gained the power to realize anything he dreamed of.

But the X-Men later realize the gift is tainted; it's powered by human creativity. The architect discovers that while he can fabricate any structure he ever imagined, he can't imagine any new ones.

Confronted with accusations of treachery, Loki is chagrined. He more or less thought he was bringing something to humans they actually wanted, and is infuriated when they reject his gift. He fundamentally cannot understand what it is that people desire.

Seems… a little silly. I might be willing to trade my ability to generate new creative ideas for the ability to execute on all the creative ideas i have but cannot possibly implement due to time, resources, and talent constraints.

Not being able to think creatively sounds scary but there’s also the alternative which is dying without doing things. Like, you could potentially make it a worse trade that sounds less scary by just making it trade years off your lifespan in exchange for creative execution and I think that’s not crazy. It’s actually fairly banal considering all the things we do to survive today.

That's just what Loki thought!

Humans... So impossible to please.

If I had this, I would be a billionaire.

I don't need to come up with any new ideas. I would just hang out on Hackernews listening to ideas that I would then execute on.

We also have all the ideas we need to fix climate change, and world hunger, we just need execution.

I would gladly take the trade of executing on existing ideas, versus coming up with new ideas that we cannot execute on.

I left out a detail: the gift was going to be distributed worldwide. Soon, nobody would be able to imagine anything new.
Someone in the art team is getting a stern talking to right about now...
What's wrong with using generative AI?
After a brief period of fascination, it's become a synonym of low quality and low effort. See e.g. the relevant KitKat campaign: https://www.creativebloq.com/news/kit-kat-ai-ad
What's the problem with low quality and low effort? Let the market decide if they like it or not.
The problem is that there's the presumption that "the market"'s choice is what we want.

"The market" has decided on what gets pushed to the top of your social media feed, and that is by all accounts turned out to be terrible for humanity. Idiocracy is what happens when we tailor to the lowest common denominator.

"The market" will decide on the cheapest and easiest, not what's in your best interest.

The actual market's choice is hard to fake. It's people voting with their dollars. Don't like it, don't watch it. Don't care, watch it.

Disney are not the market, and neither are they "the market". They are indeed betting on lots of people not caring about AI one way or another, so that they can save some money. Good for them.

> The actual market's choice is hard to fake. It's people voting with their dollars

I'm not saying it's being faked, I'm saying it's wrong to allow it when it goes against the best interest of the many, or worse, us as humanity as a whole.

People vote with their dollars by buying cheap crap to eat every day, and manufacturers save a ton of money pumping that crap out instead of nutritious food. That doesn't mean it's good or should be encouraged, or even tolerated.

The market is not god.

I’m sure many folks will take that smug sense of superiority all the way down the unemployment line while high effort, hardworking artists that leverage AI to make high quality works.
Yes and no. Let's take the closest analogy we have now, i.e. analog vs digital. The digital art technologies (photography, painting, music etc.) are much more common today than their analog ancestors. Yet, there is a part of the market that really appreciates the analog experience - they will buy vinyls and so on. My friends publishing house is commissioning only analog images for the cover. Another friend is a photographer and he is selling mostly analog photographs. So yes, you are right, but it's not really the sense of superiority, it's different things for different customers.
This comment seems to imply that AI generation is "over" or something. We're just seeing the beginning of a huge shift in how art is created, though. These models are only getting better.
People being put out of a job don't like. Out of the blue, right?
Technology putting people out of work is a story as old as time.
Let's hope it doesn't put everyone out at the same time.
Nothing, despite what the moral outrage crowd will tell you
It's surprising that the article doesn't mention the ongoing strikes in Hollywood [1], driven largely by concerns about the use of generative AI in film production.

[1] https://en.m.wikipedia.org/wiki/2023_Hollywood_labor_dispute...

if you checked the link you posted it even says the strike is over.

So not sure why it needs a mention in the article.

Edit: I stand corrected, not all of the strikes are over. However it does seem like coverage of it basically ended once the writers strike ended. I had no idea it was still going on. I don't even see it on Instagram anymore.

Only WGA has reached a deal. The other groups are still striking.
The WGA strike is over. The SAG strike continues. Both strikes were concerned about different aspects of AI generated content.
for some insight on how AI generated images can quickly ruin everything, go onto Threads and "like" one or two AI-generated landscapes; your feed will instantly turn into and endless stream of hyper-colorful, mind-bending AI-generated landscapes, each one more dramatically trippy than the last one, and within 5 minutes your brain just shuts off to the whole thing, images that if you saw an artist make just one would be amazing now looks like meaningless trash.

It's like playing Minecraft with full mods turned on that find you all the diamonds to mine automatically, or like drinking 4 liters of Coke and trying to enjoy a high quality pastry afterwards.

Sounds like a problem with Threads, not AI
did you miss the part where Threads' feed algorithm is also "AI" as much as any other machine learning model is ? Human appreciation of art is linked to human creation of art and I think AI is going to show us that truth in new ways.
As a data scientist, no, I think that’s a dumb take. Without bothering to debate the definition of “AI”,

One is a synthesis of general knowledge that can be used to generate new ideas.

The other is a multi arm bandit and some collaborative filtering. Maybe there’s some AI there to classify what things are to begin with, but the decision of what to show is a very different kind of thing.

I don't want to use the term "dumb" for your take since I don't really address people that way but it does not matter if threads shows me ten AI landscapes or 100, the point is that once images displaying what would take a human dozens of hours of technical skill to produce are easily churned out by the thousands each hour, the entire value of such art for human enjoyment drops to zero as it's no longer special or interesting.
Then don’t look at a hundred landscapes.

You made an argument that threads was showing people a hundred landscapes and that this decision was also ai ergo it’s an AI problem.

It might take humans a dozen hours to make a landscape but there are still thousands of human drawn landscapes and you’ll grow equally bored looking at all of those, no AI required.

(comment deleted)
So I do have strong ethical problems with AI usage especially when talking about how a model was trained and properly paying people.

However I think Loki brings up an interesting question. Wether or not "AI" art could become its own style? Like to me the Loki series would benefit from those inconsistencies, weirdness, etc that comes from an AI generating art.

Not that a Human can't do it, but the results would likely be different.

Not really defending its usage here, just an interesting thought I had about this in this particular use case. It happens to fit the theming very well.

That was the argument that was used for Secret Invasion, which used AI generated imagery for the animated title sequence.
To my eyes the stock image is unlikely to be 100% generative AI. It appears to be an Escheresque spiral projection of an AI generated image. The AI artifacts are exactly identical between each layer of the spiral, only scaled down, which would be hard to explain in any other way.
It also counts from 1 to 12 correctly with the numerals following the curve, which seems... quite impressive for generative AI.
I hope they do use AI. I do, and I know it has improved both the quantity and quality of my output at work.
"AI image checkers" are complete garbage, I had a lot of fun with them, but they shouldn't be used for anything other than entertainment.
Looks terrible like most uses of AI art does. Just embarrassing
As VFX it is only noticeable when it is done badly.
I can flesh out my snarky comment.

Just embarrassing that Disney cuts costs for creating a poster instead of just contracting any talented artist that would like to work for them or any artist that currently does. But it doesnt matter.

They just use a stock image as background and probably don't even notice it was AI.
Agreed, Its not neccessarly about AI. Loads of AI assisted artwork / effects are great. But using AI instead of hiring an artist almost always look like crap.