The rapid progress of artificial intelligence has generated a lot of buzz, with some predicting that it will usher in an idyllic utopia and others warning that it will bring about the end of mankind. But speculation about where AI technology might go in the future, while important, can also drown out important discussions about what to do with currently available AI technology.
One such technology is generative AI, which can create content such as text, images, audio, and video. Popular generative AIs, such as the chatbot ChatGPT, generate conversational text based on training data obtained from the internet.
Today, a group of 14 researchers from a number of organizations, including MIT, published a commentary article. chemistry This will help set the stage for a broader discussion of the direct impact of generative AI on creative work and society. His MIT co-authors on this paper include his Ziv Epstein SM ’19, PhD ’23, postdoc at the Media Lab. Matt Groh SM ’19, PhD ’23; and PhD students Rob Mahari ’17 and Hope Schroeder.
Massachusetts Institute of Technology News We spoke to Epstein, lead author of the paper.
question: Why did you write this paper?
answer: Generative AI tools are doing things that were even thought impossible a few years ago. This raises many fundamental questions about the creative process and human role in creative production. Will the work be automated? How do we keep the human side of creativity alive with all these new technologies?
The complexity of black-box AI systems can make it difficult for researchers and the public to understand what is going on under the hood and how these tools affect society. I have. Many discussions of AI anthropomorphize technology, implicitly suggesting that these systems exhibit human-like intent, agency, or self-awareness. Even the term “artificial intelligence” reinforces these beliefs. ChatGPT uses first person pronouns and we say the AI is “hallucinating”. These agent-like roles we give to AI can discredit the creators doing the work underlying the system’s output, and can discredit developers and decision-makers if the system does harm. may divert responsibility.
We help think about the interdisciplinary links and areas of research needed to address the immediate dangers to humans posed by the deployment of these tools, such as disinformation, job displacement, and changes in the legal system. To that end, we are trying to build a coalition that transcends academic circles. and culture.
question: What do you see as the gaps in research on generative AI and art today?
answer: The way we talk about AI is broken in many ways. We understand how our perception of the generative process influences our attitudes toward the output and the author, and we want to be truly transparent about the generative process, in a way that avoids some of the misleading interpretations. You have to design interfaces and systems. How do we talk about AI and how do these narratives cut along the lines of power? I have a theme. Legal aspects of ownership and credit. labor; and its impact on the media ecosystem. For each, highlight the big open question.
Regarding aesthetics and culture, we examine how past art technology influences how we think about AI. For example, when photography was invented, some painters said it was “the end of art.” But instead it became a medium of its own, eventually liberating painting from realism and giving rise to the Impressionism and modern art movements. We say that generative AI is a medium with unique affordances. At the same time, the essence of art evolves. How will artists and creators express their intentions and style through this new medium?
The issue of ownership and credit is tricky, requiring copyright law that benefits creators, users, and society at large. Under current copyright law, these systems may not be able to distribute rights appropriately to artists as they train their style. What does copying mean when it comes to training data? It’s a legal question, but it’s also a technical one. We’re trying to understand if and when these systems are copying.
When it comes to labor economics and creative work, the idea is that these generative AI systems can accelerate the creative process in many ways, but eliminate the ideation process starting from a blank slate. In some cases, starting with a blank page can yield good results. We don’t know how AI will affect creativity, but we need a better understanding of how AI affects different stages of the creative process. We need to think carefully about how we use these tools to complement rather than replace people’s work.
In terms of the impact of generative AI on the media ecosystem, the risk of AI-generated misinformation must be considered, as synthetic media can be generated at scale. We need to protect the media ecosystem from the possibility of massive fraud on the one hand and people losing trust in authentic media on the other.
question: How and by whom do you hope this paper will be received?
answer: Conversations about AI are very fragmented and frustrating. It’s been hard to think deeply about these ideas because technology is advancing so fast. To ensure beneficial use of these technologies, we need to build a common language and begin to understand where to focus our attention. We hope this document is a step in that direction. We are about to start a conversation that will help build a roadmap for understanding this rapidly changing landscape.
Artists are often pioneers of new technologies. They’ve been playing around with this technology long before there was commercial use for it. They study how it works and struggle with its ethics. AI art has been around for over a decade, and artists have been addressing issues we face as a society for just as long. I think it’s important to raise the voices of artists and other creative workers whose work is influenced by these tools. Art is a way of expressing our humanity. It is a core emotional part of human life. As such, we believe it is at the heart of a broader question about the impact of AI on society, and hopefully we can build on that debate.
