How An AI Generated Photograph Won The Sony World Photography Award
Most people think that they can tell an AI image apart from a real one.
But above image might call that into question. It won first prize in the Sony World Photography Awards, one of the most prestigious photography competitions on Earth. And as you can guess by the context of this title it's not a real photo, but AI generated, and it fooled some of the keenest eyes out there.
The prompter of this image, Berlin-based Boris Eldagsen, has been a photographer since 1989 and jumped at the opportunity to explore the creative possibilities of AI
How many of you knew or suspected that this was AI generated? Something about this doesn't feel right, does it? AI images and photography should not compete with each other in an award like this. They are different entities. AI is not photography. Therefore, I will not accept the award -- Boris Eldagsen
This highlights a glimpse into the future where AI creations become indistinguishable from reality. The world has just changed and we're going to be coming to grips with what that means for a while. The field of the creative arts has been hit the hardest by AI, and I'm sure you've all seen AI images all over the internet.
This rapid change has seen fierce resistance from artists around the world. And who can blame them? Professional artists devote everything to their craft, and now AI can do it in seconds. And to add insult to injury, it's been trained on some of their work, so there's a lot of questions here.
What happens next, and how can an AI steal from humans anyway? Is it actually stealing? And if so, don't all humans take inspiration from others. But when a machine starts taking ideas from humans, it feels different. So today, we'll look at how AI can make such artwork, and why the process is stirring legal controversies surrounding copyright laws, and we'll also take a quick look at some other insane things going on in the world of AI.
Let's look at some things that have been going on with AI in the field of creative art. First up, Runway machine learning version 2 has dropped, and people are having fun with it, creating whatever cool videos they can imagine. It's at its early stages, but Runway seem to be proving to be the leader in generative AI video production.
Google wants to have a slice of this play, and they're going full steam ahead with their generate of video AI, Dreamix , it can create videos from a reference video or photo, and the results are pretty good. Just a few days after 谷歌 , 英伟达 throws their hat in the ring, announcing their AI video generator, which also looks pretty decent.
Meanwhile, two computer scientists managed to fool a bunch of Redditors ( Reddit, Inc. ) into giving them money. They used Stable Diffusion to generate fake images of an attractive female and convinced the man to give some money if they wanted to see more pictures. I'd imagine that this was one of the easiest scams to pull off, and unfortunately, there's going to be plenty more like it.
An AI version of Drake and an AI version of The Weeknd collaborated in a fake song called Heart on My Sleep, and it was doing the rounds on social media and Spotify gaining hundreds of thousands of streams on Spotify alone. Universal Music Group caught wind of this and pulled it down from streaming services. In a statement, they stated that they wanted to ban the technology and called it a fraud. But the thing is, it's happening everywhere with big artists like Kanye West, Jay Z and Ariana Grande, people are creating backing tracks and putting AI-generated vocals on top of it.
Music labels may be angry now, but once they figure out how to do it themselves, you can just bet that they'll be making money from AI-generated music of artists who have passed away.
ChatGPT4
In more general AI news, science gets a boost. It was found that GPT-4 answers US medical exam licensing questions correctly more than 90% of the time, better than some licensed doctors. Dr. Isaac Kohane (Chair of the Department of Biomedical Informatics, Harvard Medical School) ran a test. He gave GPT-4 a real-life case that he treated several years earlier. He gave the machine a few details about the case and it was able to diagnose a 1 in 100,000 condition. It would reason just like he did. And on top of this, multiple papers have now come out stating that GPT-4 can improve its own output when asked to reflect on its performance. But all of that is already outdated.
AutoGPT
What's taking the world by storm now is auto GPT and similar systems. These are AI agents that can go out into the online world and do tasks for you. You just say the task that you want and that's it.
I'll actually tell you how I built it. So there's no individual lines of coding here. The AI's I'm using here are GPT-3, GPT-3.5, GPT-4 and Whisper. When it needs to speak, it'll use ElevenLabs . The AI agents don't wait for you to ask more questions or wait for your response for the next step. They just do it automatically by prompting themselves until the objective is met.
These AI agents can do much more than just order pizza. People have even asked them to do market research and come back with a report.
And just to be clear about the new world we're living in, the heavily used BabyAGI tool was created by a single guy called Yohei with no coding experience in just three days. How insane is that? He had the idea to create this crazy app, so he just asked ChatGPT to write the code for him.
In other AI news, Samsung calls Google's stock price to drop by 55 billion because Samsung hinted that they were going to replace Google's search as the default to being AI and being search on their Samsung phones. Whatever happens, it seems that investors are worried about Google losing some market share.
Google would hit back with the announcement of Project Magi, an AI-powered component to Google's search.
Meanwhile, Snapchat users get an AI bot (MyAI) and they're facing a massive backlash. And Elon Musk announces his own AI project, TruthGPT. And we could just keep going with the news, but there's literally too much happening and not enough time for this article. So let's get back onto the topic at hand, how AI images are coming into conflict with human artists and the law.
Fun fact: Did you know that the Midjourney team only consists of 11 people?
Just a few years ago, before the advent of capable, generative AI, even the world's largest companies with all their resources would struggle to produce AI images with such quality and coherence. And today, nobody even thinks about it. We're all used to it. It's almost like this has always been possible.
From Midjourney to Stable Diffusion and Dream Up, any user can now create stunning and surreal images from text prompts in under a minute. It's amazing when you think about it.
But the thing is, the real magic is happening in the background. So how does it work? Well without going into too much detail, these AI systems use a method called diffusion. This is essentially where an AI learns to create images by starting with a picture and destroying it with noise. It then goes through a series of steps to reverse the noise and reconstruct a new image. Understanding how all of this works is beyond the scope of this episode, but what is pertinent to this discussion is the training data used to achieve this. So we need to look at how these systems are trained.
Most AI image generators get their training data from an unexpected place, a small German non-profit, this company has a data set called "Large Scale Artificial Intelligence Open Network" or LAION, which consists of over 5 billion indexed images.
Unfortunately, this includes the works of many unsuspecting artists and also copyrighted images such as those from Getty Images. Some see this as a democratisation of art and a source of inspiration. Others see it as a threat to the originality and value of human creativity. Midjourney claims it does not copy or infringe on any existing images, but rather synthesises new ones from scratch based on text input. However, when you look deeper, things get more complicated and not everyone agrees with Midjourney's claim.
Critics also point out that these AI systems are not truly creative or artistic but rather rely on existing data to produce derivative and superficial outputs.
What Is Creativity In The First Place?
Is it simply the human mind remembering past inputs and mashing them together in a new way? Don't humans borrow from each other all the time?
Steve Jobs once quoted Picasso stating that quote, good artists copy and great artists steal.
Even further back, in the 19th century, the poet, W.H. Daffenport Adams wrote, great poets imitate and improve, or a small one steal and spoil.
Although these quotes are the opposite way around, they both indicate something interesting, taking the ideas of others in art is a practice as old as time. But what if creativity is more than that, the sparks that seemingly come from nowhere?
Instead of the situation of a musician sitting at a café, some innocuous sound or smell triggers a distant memory in their mind and it inspires them to write a masterpiece, a biological and uniquely human process.
But it's seeming like regardless, AI is mimicking creativity. Despite the controversies, AI art has been featured in various exhibitions, competition, magazines, and of course social media platforms. To give you an example, the following cover of the magazine Cosmopolitan Magazine was created by DALL-E 2.
An AI artwork called Portrait of Edmond de Belamy won first prize in a prestigious art competition in France, beating out hundreds of human-made artwork. The painting was later sold for more than $400,000, becoming the first AI-generated artwork to be auctioned by a major art house.
The image below, titled Space Opera Theatre, when translated from French to English, won first place in the Colorado State Fair Fine Arts Competition last year, and the category was digital arts/digitally manipulated photography.
All of these examples just underscore the sheer scale and somewhat wide proliferation of AI art.
And because it's accessible to anyone, Instagram , TikTok and Twitter are flooded with them. It's all happened so quickly, and remember, the technology is in its infancy and it's the worst it would ever be. In light of all this, artists are left devastated. Making the process of creating art easy and for everyone has devalued the price that clients are willing to pay for it. Why would a client pay thousands for a custom work of art if they can get the same thing essentially for free? If needed, it can be touched up for a fraction of the cost.
领英推荐
If you're not an already established artist, it's going to be a really tough time right now, and I do feel for the artists out there. Obviously, things around AI art have been developing very quickly. It's on a level. It's of course very heartbreaking. And then also seeing on a collective level for the community of artists out there who are suffering, it's also just so heartbreaking.
Definitely sad to see younger artists or any artists out there who feel like they have to give up drawing because they think it's not going to be a viable career.
And I don't know what the future holds, but remember why you draw and why you started drawing because I would guess that most of the time you started drawing because you enjoy it. And I don't think you should stop. Okay. For some artists, it gets even worse. What happens when their art is used within an AI's training data set? Kelly McKernan had to find out the hard way.
Legalities & Lawsuits
A rising star in the art world, McKernan is known for their distinctive style. Their paintings combine fantasy and sci-fi elements with bright colors. On Discord's Midjourney platform, their name appeared in over 12,000 public requests, resulting in images that resembled McKernan's work, but with different themes and subjects. If you remember my early AI videos, this is called a style transfer. But for McKernan, the use of their name crossed a line.
They've now joined a class action lawsuit in January of 2023 alongside two other artists, Sarah Anderson and Carla Ortiz. They filed the lawsuit against Midjourney, stable diffusion, and dream up for infringing on their artistic identities.
Sarah Andersen, Kelly McKernan and Karla Ortiz?sued Stability AI, Midjourney, and DeviantArt for committing mass copyright infringement by using the artists' work in generative AI systems.
The legal battles against AI art generators gain more headlines when Getty Images , one of the world's largest stock photo agencies, sued Stable Diffusion for allegedly infringing on its copyrights.
Basically, they're saying that Stable Diffusion was selling and licensing AI-generated images that were derived from their collection, but without permission or attribution.
Legal analysts suggest that the Getty Images lawsuit has a stronger case against AI art generators than the artist-led class action lawsuit, but predicting the outcome is still difficult.
Aaron J. Moss, a copyright lawyer at Greenberg-Glusker and publisher of the Copyright Lately blog, praised Getty's complaint on Twitter .
He stated that it was better focused than previous lawsuits. Moss believes that the case will address the issue of ingesting copyrighted images to training data, and it seems like this will lead to a potentially fascinating fair-use legal battle.
Here's an interesting question.
Who Owns AI-made Works?
If a robot makes images, does that mean anyone can use them, even for commercial purposes?
The debate surrounding AI-generated works and copyright protection is very complex. On the one hand, there is an argument that AI lacks the creativity that's characteristic of human beings, and thus does not meet the originally stated requirement for copyright protection. However, others say that AI-generated works do show some degree of originality and creativity. It all depends on how the model is trained and the degree of human involvement in the creation of any artwork.
Let's take an example.
In late 2022, an AI-generated comic book named Zarya of the Dawn by Kris Kashtanova was initially granted copyright protection. The copyright protection was later revoked because the author didn't disclose that the images were created with the help of Midjourney, and the US copyright office said something very interesting about the ruling. They decided that only human-created works can receive protection. The images themselves are not protected by copyright, but the text and the arrangement of the images are still protected, and this is because that the creative work of a human author. Interestingly, the standards for determining originality and creativity differ between different jurisdictions.
In Chinese courts, the rulings have gone either way for granting copyrights on AI-generated art depending on how much human involvement there was.
In Europe, copyright protection for AI art is still up for debate.
The UK and Ireland have specific provisions for computer-generated works, while France and Germany require human authorship.
There's no clear consensus or establish law on how to handle AI-generated works, and whether we like it or not, AI-generated works will become increasingly pervasive as time progresses, so this is all just the beginning. A solution will have to be reached.
One possible way to approach this question is to look at how art evolved over history and how new technologies have influenced its development.
For example, when photography was invented in the 19th century, many painters felt threatened by its ability to capture reality more accurately and quickly than they could. They all feared that they'll be put out of business.
There were also privacy concerns, as people could now take pictures of others without their consent or knowledge. As photography became more widespread, there were legal battles over who owned the rights to photographs and whether they could be reproduced without permission.
One important legal precedent was the case of Burrow-Giles Lithographic Co. v. Sarony. The case established that photographs were eligible for copyright protection just like other works of art. This case helped establish photography as a legitimate form of artistic expression and pave the way for copyright protection for photos.
Another example is Photoshop.
After its launch in 1990, it quickly became a popular tool for artists, designers, photographers and advertisers. For the first time, the general user could crop, resize, filter, layer and blend their images. Photoshop was initially met with resistance and suspicion by some artists. They feared it would undermine the integrity and authenticity of their work. They argued Photoshop would enable fraud and deception as well as diminish the skills and talent required for photography, art and even painting.
But we all know what happened.
Photoshop became widely adopted and embraced by many photographers and artists. They saw it as a powerful and versatile tool for enhancing their work.
These are some historical examples of how these new technologies can both challenge and enrich art and creativity.
That does show that art is not a fixed or objective concept, but a dynamic and subjective one that continues to evolve with time and culture.
So in the future, will AI art be just another tool?
In terms of deception, will societies learn to adapt to AI images and look deeper into their origins as a reflex? Only time will tell, but the nagging thing for me that makes me think that AI art may not just be another tool is that this is the first technology that isn't static, it can actually learn.
What Remains To Be Seen?
So this new world of creative AI and in this case, specifically AI generated artwork is fascinating and complex. As with anything AI, there's both immense opportunities and challenges. And in this case, it's both for artists and society. Maybe it's time to rethink our assumptions about art. Firstly, what is art and secondly, who decides what art is? Taking a look at these pictures, maybe it's clear that they're AI generated. But the thing is, the idea behind them is powerful. Powerful people in the world that have untold riches, being in places of poverty. The visual aspect wasn't done by a human, but the thought is very human.
It's definitely an interesting discussion, so let me know what you think in the comments section below.
THANK YOU FOR READING
LIKE, SHARE & FOLLOW FOR MORE