- No Longer a Nincompoop
- AI Fear Mongering at an All Time High
AI Fear Mongering at an All Time High
The Legality of AI Photoshop + US & EU Discuss Regulation While Japan Goes All-In on AI
Welcome to edition #17 of the “No Longer a Nincompoop with Nofil” newsletter.
I missed last week 😔. I’ve been busy sorting myself out (having a little existential mid-life crisis) but I’m back now 😤.
To make up for it, especially to the paying subs, I’ll be covering all the madness that happened over the last two weeks across multiple newsletters this week.
If you’re a paying sub, I’m also offering a free 30-minute consult to help you understand how AI can transform your business. You can reply to this email or email me at [email protected]
Here’s the tea ☕
Generative Fill & The Law 📜
Regulatory Lines are Being Drawn 🖌️
The Noise is Getting Louder
Time magazine just released their new cover and well, it’s not a good look for AI journalism. I get it. The largest companies and leaders have been calling on AI regulation for a while now, citing the catastrophic dangers that lie ahead. It’s easy to take this and start spelling doom for humanity and civilisation as we know it, but cooler heads must prevail.
At this point in time, there is no AI model that is even remotely close to doing anything dangerous or catastrophic. These models are not sentient, they are not conscious and they do not have the ability or the means to do anything remotely dangerous to humans. Can they be wielded for the wrong reasons? Absolutely. Can they do this willingly without any guidance or assistance? No.
It’s easy to get caught up in the hype, but it would be more beneficial for everyone if we thought of how we can use this incredible technology to better humanity - climate change, education, waste, medicine - there is so much potential for AI to bring about positive change, but I guess that won’t sell headlines and clicks.
Adobe isn’t Going Anywhere
Adobe released the best AI capability in Photoshop so far called “Generative Fill”. It essentially takes an image and is able to edit and/or expand it in a way that seems incredibly natural. Just look at some of these examples. The white box is the original image and everything outside is AI generated.
Source - Nirvana’s Nevermind album cover
Adele’s 21 cover
This all seems really fun, and it is. Even someone like me with no Photoshop experience can go in and edit an image or create an entirely new one with this feature. There are thousands of examples out there already. But there are a few pressing issues with this that need to be addressed.
Who Owns These Images?
Generative-filled art isn’t exactly something entirely new, but neither is it entirely fabricated either. So does it fall under copyright? Are artists and photographers protected from this new technology? Unfortunately, it’s unlikely. Technology is moving too quickly and governments risk halting innovation if they take a decisive stance on AI use. Image generators like Midjourney and now editors like Generate Fill are completely changing the landscape for photography, art and everything in between. It’s only a matter of time before photo-realistic video generation becomes widely available as well.
We already know that an image produced with AI cannot be copyrighted. So can an artist stop someone from using their work to create new AI generated work? There are legal grounds for this, yes. But one only needs to call upon fair use and suddenly we have an incredibly complicated and messy situation where the law can favour both parties at the same time. Read more about the legalities here.
Left: Goldsmith vs Right: Andy Warhol Foundation
Recently the US Supreme Court ruled in favour of Lynn Goldsmith, a photographer who took the photograph on the left of Prince. The Andy Warhol Foundation then created an inspired artwork and cited fair use and believe it or not, the court sided with Goldsmith. It’s a monumental victory for photographers, but I don’t see how these rulings will stand when artists start taking hundreds of people to court over AI generations inspired by their artwork. The world of regulation is going to be a nightmare in the coming years and the battle lines are already being drawn.
Regulation - US, EU & Japan
Leaders from the US and EU are meeting in Sweden as part of the Trade and Technology Council this week. Naturally one of the main talking points is AI regulation. Both the US and EU have very different stances on this.
The US owns the vast majority of AI research and innovation. Imposing strict regulations on AI research and training would undoubtedly stifle innovation and growth in an industry that is developing the most important technology in human history. There are concerns in the US that if regulation is imposed, countries like China and Russia will catch up and eventually surpass them. In my opinion, AI will never be appropriately regulated in the US. There is simply too much upside for companies, too much money to be made.
The EU is on a completely different wavelength. As I’ve spoken about previously, the EU’s stance on AI is strict. One aspect of their proposed law is for developers to document and track copyrighted material used in training data sets for AI models. They’ve also proposed licensing new models, with unlicensed models facing fines of up to $20 million.
The reality of the situation is this - if the EU imposes these laws, there will be zero innovation coming out of the EU. Companies will change their headquarters and developers will be extremely limited in the work they do. Non-profit organisations like LAION are pleading for the EU to not make it a technology backwater as these proposed regulations will also have a big impact on open-source innovation, the only thing stopping a complete duopoly of AI models, research and innovation. The US is stalling as the technology matures, the EU should probably do the same.
You know who doesn’t care about all of this copyright and regulation nonsense? Japan. They’ve become the first country to outright say that copyright does not apply to AI training. What does this mean? Companies will be free to use any and all data as part of their training, regardless of its purpose (non-profit, commercial, licensed etc).
Real or Midjourney?
One of the biggest and best image generators like Midjourney has already stated that they couldn’t care less about copyright because it simply can’t apply to these gigantic models. I wouldn’t be surprised to see companies opening offices or even completely shifting their headquarters to Japan. The problem is once even a single country does something like this, any other country that doesn’t follow suit is falling behind.
Japan has cast their hand. They don’t care about copyright and invite companies to train and build AI models in their country. This ties in with their push for using AI across all industries, including government. It’s no wonder OpenAI is opening up offices in Japan.
As always, Thanks for reading ❤️
Written by a human named Nofil