About Sean Sean Byrnes has been the CEO & co-founder of many category-defining companies, raised over $100M in financing from leading venture firms, built teams of hundreds of people, and both bought and sold companies in his career. Sean is also an artist - check out his portfolio here. Sean has a deep understanding of AI gained as a founder and CEO of Outlier.ai, the world's first automated data analysis platform and Flurry, the largest analytics platform for mobile applications. Sean is also a partner at LucidFog, a venture investment firm that provides capital to early stage founders. Summary In our conversation, Sean discusses how art has always been created throughout time, generative AI and the creative process, why some artists have a visceral reaction to generative AI and the ethical aspects. Finally we conclude with a deeper question about the meaning of art. Art creation throughout time The analog era When we think of art creation we imagine a genius artist waving pencils and paint brushes in front of the canvas. However, this has never been reality. The way it actually worked, for example during the Renaissance is that Rembrandt had armies of assistants who would do background painting. Many assistants used mirrors for tracing. There were no magical mythical lone artists. Throughout time, artists have always used shortcuts, tools and borrowed from each other, stealing ideas. The digital era As technology evolved new tools became available. An early example is scanning art and editing it in photoshop. Everyone is familiar with the iPad, but in reality, even 15 years ago we had digital tools. Today most of art has been treated digitally, modern art programs have the ability to import 3D models - people can take pictures of a college campus for example and trace it to render a 3D model. All the tools are there, generative AI is the next stage of tool evolution with a constant march forward. Artists can now create more art faster to fulfill their vision. But what happens when computers can create art without humans in the loop? Right now in 2023 we’re at a point where we’re in the next step of tools. What generative AI means for the Creative Process Historically the creative process meant that an artist sits with a sketch pad to see if they can capture something interesting. Musical artists have 100s of bits and assemble them into songs. As artists look at tid bits and look to capture it. Can we now sit down with generative AI to do that? Before we’ve had to sketch things out ourselves, now we can use AI to come up with something in the beginning. * Generative AI doesn’t have a voice - for Sean art is meaningful if it triggers something - AI doesn’t do that. Rembrandt makes him feel like he’s there. Generative AI on the other hand doesn’t have a consistent voice, it’s conglomerating all the content and it gives an average voice of everyone. However, it gives artists a good way to start their explorations. * Exploration of ideas with generative AI. Whenever artists create something they typically run into problems where the piece is not coming together. Digital art allows creators to try a bunch of stuff - like an army of Renaissance assistants, we now have an army of AI assistant painters to help explore ideas. AI is invaluable for experimentation, it helps to unlock new things. The fidelity of work today has vastly improved when we compare it to what we had in the 80s vs today. This is all due to tools we’ve available now. Why artists are frustrated with Generative AI * Natural resistance to automation. The new generation of AI tools is making hard earned skills obsolete. Bagels today are made in bagel machines. It used to be the case that bagel making was a sought after artisant form. Bagels were expensive and very rare. Bagel machine wiped out artisan bagel making. This is true in manufacturing as well. Artists are afraid for their livelihoods. They invested time to learn and hone artistic skills yet today AI can generate similar creative works within minutes. This means that artist livelihoods are threatened. * Use of original art to train AI models Generative AI models are trained on corpuses of content which is copyrighted. By law technology can’t plagiarize but can take copyrighted content and can own the model. It’s not entirely clear how this works. Artists publish art and portfolio Somebody else can show up because ML models can create things. Stable diffusion generates artists' signatures where you can see it stealing. Will have to pay an artist to use it in an ML model. Artists have a real concern. * Art doesn’t solve an explicit problem - Creating art has always been a difficult endeavor with artists’ work subjected to opinions. Artists have to have a thick skin in order to deal with a constant barrage of opinions. AI Ethics * Attribution How can artists be compensated and acknowledged when generative AI models are trained with their work? * Copyright In a recent court case Kris Kashtanova, a comic book author received the first known copyright registration for AI art. It was initially revoked after the U.S. copyright office discovered her images were created with Midjourney. “We conclude that Ms. Kashtanova is the author of the Work’s text as well as the selection, coordination, and arrangement of the Work’s written and visual elements. That authorship is protected by copyright. However, as discussed below, the images in the Work that were generated by the Midjourney technology are not the product of human authorship. Because the current registration for the Work does not disclaim its Midjourney-generated content, we intend to cancel the original certificate issued to Ms. Kashtanova and issue a new one covering only the expressive material that she created. “ Zarya of the Dawn letter. A lot of the copyright issues around AI and art are fuzzy at the moment and will need to be ironed out. * Accountability If a self-driving car kills someone whose fault it is? We don’t have provisions for robots being responsible today. Art has the same problem, what if generative AI creates child pornography - whose fault is it - the company, people who put images in the training data or the person who generated them? Who is responsible for when these things go wrong? Is it ethical for AI to absorb images that are not copyrighted? Who has economic ownership and rights of the art that’s created? The blame and responsibility is a big question, people will hide behind the ambiguity until we figure out a way to take care of it, this is dangerous and scary. Final Words Sean left us with one last question - why do we create new art? Our libraries are full of books. We’ve created so much content? Why do we need more? This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit aistudios.substack.com
