Scenes with Simon #2 - AI Lawsuits, Mirrors, and Free Homes in Japan
AI Lawsuits, Mirrors, and Free Homes in Japan
Copyright + IP Law. On the one hand, it’s a confusing mess that, like tax law, can send you into the abyss of despair, confusion, and frustration. On the other hand, it’s one of the most interesting parts of the law that leave you feeling like you’ve gotten lost and then stumbled into a cozy bar in a back alley in Tokyo. Some days you want to abolish all of it in a fit of rage, but on other days, you know that once you step in, there’s a world of interesting implications for how we think about creativity, our livelihoods, and the future.
This week there’s been two stories touching at details of IP Law: new lawsuits against AI Companies + DND’s continued backtracking against changes to their licenses (“Open Gaming License”) that governs how players use original IP + create derivative content.
Artists + Getty Images Sue AI Art Companies
Filed Under: ip, ai, art
In the case of the artists filing a class-action lawsuit against Stability AI, DeviantArt, and Midjourney, the lawsuit had been made public. A contentious issue with this suit surrounds how AI art generators actually work. It’s not merely a “collage tool”. It’s more complicated than that. Tech enthusiasts replied with a detailed response, describing where the lawsuit is wrong (both in terms of copyright law and technical details). This is a great learning resource to understand how this works and where people differ. The primary issue and concern is a belief around copyright violations.
While the Getty Images suit is not public yet, it’s “angle of attack” seems like it might be likelier to stick: that of terms of service violations. The EULA + Site Terms of Getty Images both point to being against the ToS to scrape for the purpose of building training models.
No Machine Learning, AI, or Biometric Technology Use. Unless explicitly authorized in a Getty Images invoice, sales order confirmation or license agreement, you may not use content (including any caption information, keywords or other metadata associated with content) for any machine learning and/or artificial intelligence purposes, or for any technologies designed or intended for the identification of natural persons. Additionally, Getty Images does not represent or warrant that consent has been obtained for such uses with respect to model-released content.
That being said, I’m not sure how defensible this is, because the US recently confirmed that data mining (web scraping) like this is legal + the UK is also planning to change laws to be friendlier to public data mining.
What’s interesting, however, in the Getty Images case is what’s not being said.
When asked what remedies against Getty Images would be seeking from Stability AI, Peters said the company was not interested in financial damages or stopping the development of AI art tools, but in creating a new legal status quo (presumably one with favorable licensing terms for Getty Images).
When you look at what Shutterstock did with OpenAI, it looks like Getty is trying to position itself in a similar vein with Stability. Shutterstock is building their own walled garden. You aren’t allowed to scrape and use the images for training your own model, but Shutterstock allows you to use an AI generator that they built with OpenAI. If your images were used to train the model, you will be paid out a proportional value when someone licenses the resulting output from Shutterstock.
For data deals, a portion of the contract value is paid to all contributors involved, distributed proportionally by the number of assets and metadata a contributor has included in the deal. For generative licensing, a portion of every customer license for generative content is paid out to the contributors whose content was used to train that particular model (eg. OpenAI or LG AI Research). Generative licensing also distributes earnings proportionally by the number of assets and metadata a contributor had included in the original data deal.
This makes more sense in terms of how AI models work. The latent, trained, space don’t point to specific images being used, but rather, the model as a whole when it generates new content. It’s going to be interesting to see if this works, especially when stock image websites already have a large corpus of license-restricted infrastructure already built.
Ultimately, I sympathise with some artists. It undeniably represents a threat to the livelihood of a subset of them. A great, earlier example comes from how translation technology affected translators. The “middle-class” of translators lost their jobs. Now you have lots of okay, free translations that are good enough, surviving online and in the real world. You still need expert translators if you want to verify and ensure it’s 100% correct (for those that can afford it). Some artists who live on commissions will lose their livelihoods if they don’t adapt. There are ways to adapt, such as adopting the technology itself to scale your own output. If one becomes renowned and manage to build a brand around one’s own style, it becomes another tool. For example, what if fans are allowed to generate works in your style and they pay you nominal fee to have it “signed by” the artist? It’s a small example and this will undoubtedly start to change, rapidly.
While I think that these lawsuits are necessary for us to understand where the boundaries lie, I don’t necessarily think that fighting in courts is the best place to fight for one’s livelihood. We live in a time where change happens faster than the court process itself. By the time these are done, there might be new AI technology + companies out there. I never expected that we’d, for example, see a truly open source AI generator that could easily run on off-the-shelf laptops so soon. Now, we do. The cat’s out of the bag, and I think it’s better to advocate for a system we would love to see: a global class of creatives thriving in tandem with tools that make it easier for people to imagine and express themselves. It starts with things like Shutterstock paying for training data alongside more flexible ways to consent (despite what the law says).
And yes, it’s not all amazing. I know that Nick Cave hates ChatGPT because the outputs sometimes feel dead and devoid of meaning: mimicry without feeling. I agree with some of Nick's sentiments. There *is* always more meaning to a creation than merely its output. A story matters in its context. And with content already being abundant, we gravitate to media that allows us to connect to others through it. That being said, imitation isn't all bad, and learning from an imitator can be useful. AI can be a mirror: an effortless reflection that also allows us to introspect. Sometimes we need our creations reflected back to us: from art to AIs.
DND Backtracks Their New License
Filed Under: dnd, ip
After sending around new contracts to be signed with a new license, the backlash led to Wizard of the Coast backtracking a lot of the egregious changes.
I enjoyed Cory Doctorow’s take. I didn’t realise that the original OGL actually took away certain rights (to parts that are usually non-copyrightable to start with). Some of these parts have actually now been re-licensed under creative commons 4.0 to reduce ambiguity.
This quote from Cory resonates in context with the AI Art debate:
The copyrightpilling of the world sets people up for all kinds of scams, because copyright just doesn’t work like that. This wholly erroneous view of copyright grooms normies to be suckers for every sharp grifter who comes along promising that everything imaginable is property-in-waiting (remember SpiceDAO?):
https://onezero.medium.com/crypto-copyright-bdf24f48bf99
If one argues for more stringent copyright protection (like relinquishing fair use in the USA), one makes a Faustian bargain, which ultimately means that mega corps like Disney win: they can afford to sue for infringement and we lose access to fair use derivative work like fan art or fan fiction.
Filed Under: twitter, api
Twitter intentionally shut off API access to apps that recreate the Twitter app experience. The absolute end of an era. One forgets what it felt like back in 2007-2010 when this was all new and interesting. My love for what the web could do stems from the Twitter API, creating mashups, and dreaming about the ability to compose new experiences. It’s what ultimately led me to blockchain technologies: a substrate for composability.
It’s sad. But, hopefully we realise again also what the web should be: an open set of protocols rather than walled gardens.
Filed Under: acting, creativity, roundtable
An amazing roundtable from an accomplished set of actors. It almost felt like a therapy session as all of them spoke to the ups and downs of the creative process: the high and the lows. It was also great to just see a bunch of men be vulnerable, sincere, and emotionally honest with each other. One line stood out to me, a metaphor I'd never heard before. Jeremy Pope talks about a conversation with his therapist in which they describe that life is like your heartbeat. There's lows and highs. If you wish things to just be nice, to have less highs and less lows, you'll have a flatline. So, remember that as you ride the waves of the creative process (and of life). Inspiring roundtable.
Who could write protocol fiction for speculative infrastructure?
Filed Under: fiction, speculative, protocols
An interesting proposal to broaden fiction towards documenting speculative infrastructure and protocols. Definitely resonates with me. In "Hope Runners of Gridlock", there's a citizen assembly where the society discusses raising the taxes on a novel self-assessed Georgian tax system (Harberger tax/COST) on the city's gridlock. I've contemplated actually writing an addendum where you could read the entire report containing expert and citizen input. Fiction that's about speculative infrastructure and protocols sounds interesting and fun. I’d actually add Kim Stanley Robinson’s “Ministry for the Future” as speculative protocol fiction. (ht Venkatesh).
JPG Discourse: Conceptual NFTs
Filed Under: nft, jpg, tcr, canon, curation
JPG's new experiment in curating NFTs using a modified token-curated registry has led to some really interesting discussions. Canon 1 was curating on-chain NFT works. Canon 2 is attempting to curate conceptual NFTs. The discourse has been fascinating.
China's population declines for 1st time in decades
Filed Under: china, fertility, population
I can't make up my mind what nation states will do in the face of population decline? It's probably the most "rorschach test" of all political beliefs: what you see is probably wrong with society is also the reason why you would think that people aren't having kids. I have a hunch that we might see countries adopt more lenient immigration reforms towards ancestry or extended family (alongside skilled immigration)? I always look to Japan, being ahead of the curve of other countries that will follow, to see what choices a country will make. Speaking of:
Why Japan is Giving Away 8 Million Free Houses
Filed Under: japan, urbanism, zoning
Japan's demographics alongside its unique zoning and housing laws means that many homes in Japan are becoming available for very cheap (with caveats of course). A good overview though on a fundamental part of Japanese urbanism: nationalised zoning rules + a history of property as closer to a consumer good, not as an investment.
This image is a great example of how too much, is indeed too much.
Filed Under: nft, art
This is an interest NFT art experiment. Play games of Connect Four on-chain. The winner wins the final image as the end-game state. Cute, smart, and creating interesting art along the way. Reminds me of Clovers and the end-game state of Clovers being collectible.
The TikTok Family Guy Pipeline
Filed Under: tiktok
Look. I'm not one of those people that are all like: "TikTok is brain mush content". There's amazing creativity and great content on there. But, you have to be aware of how you use it. The algorithm is your assistant! If you don't know what's going on, you WILL go down a rabbithole content to pure brain mush. The most notable ones are those where there's multiple pieces of content happening at once on the screen (like a TV show, split with a weird, fun game). These are the worst kind of brain mush content, because it really feels like you are tuning into a wavelength rather than actual content. Background noise to medicate the doomscroll. (ht Allan)
Rhea Myers: Conceptualism Rehashed
Filed Under: art, blockchain, conceptualism, nft
Great article from Kevin Buist detailing the work of Rhea Myers. I still have a vivid recollection of discovering "Is Art" in 2014/2015, and realising that there was so much that was going to change and become possible with this technology. I'm happy to see her recognized for the pioneer that she is.
She does have a book out now detailing her work. I ordered a copy, you should too!
Arlington's Traffic Going Down
Filed Under: urbanism, traffic
I despise how intensely car-centric the US is. A lot of cities are changing for the better, though. In Arlington, traffic has gone down while population growth is up. It has been done. It can be done. And more of it should be done.
Casino Versus Japan - It’s Very Sunny
Filed Under: music, chill
I’m back on a “Casino Versus Japan” kick recently. Such a warm, ambient, and ethereal tone to the music. I leave you with this tune and this great quote.
In response to a tweet about an art director having to filter through applicants using AI Art, Christopher Doehling responded in reference to a key conundrum of Blade Runner: the plight of what is and who is human.
I’ve seen things you people have already imagined. Attack ships on fire on the trending page of Artstation. Fanart glistening on the DeviantArt main feed. All these memories… will be lost… when they pull the plug after the lawsuit.
What do you see in the mirror?
See you all next week!
Enjoy the sunsets.
Simon