AI Art: How artists are using and confronting machine learning | HOW TO SEE LIKE A MACHINE

The Museum of Modern Art
15 Mar 202314:56

TLDRThe transcript from 'How to See Like a Machine' discusses the intersection of art and artificial intelligence (AI). It highlights how artists are not only using AI as a creative tool but also challenging its underlying systems and societal implications. The conversation covers the generative turn in art, where traditional methods are being rapidly transformed by AI's ability to generate new content. It delves into the biases inherent in AI training data and the potential for AI to offer a multidimensional imagination that transcends human categorizations. Artists like Trevor Paglen and Refik Anadol are showcased for their work in exposing the cultural and political nuances within AI systems and for creating innovative AI-driven art installations. The discussion underscores the importance of critical engagement with AI, both in understanding its capabilities and in considering its broader impact on society and the world.

Takeaways

  • 🤖 Artists are increasingly using AI as a tool to not only create art but also to challenge and provoke thought about AI's role in society.
  • 🧐 There is a lack of public understanding about AI, despite its widespread use in daily life, and artists aim to bridge this knowledge gap.
  • 🏭 Artists are exploring high-level existential questions through AI, such as the concepts of free will and the nature of perception.
  • 🎨 By using AI, artists are able to make technology perform in unexpected ways, demonstrating a creative subversion of its intended use.
  • 📈 AI breakthroughs like DALLE-2 and Stable Diffusion are enabling new forms of interaction, but some artists find more inspiration in alternative uses of AI.
  • 🧠 In unsupervised learning, AI models create their own classifications and maps, which can lead to a more speculative and imaginative form of art.
  • 🌌 Refik Anadol's work at MoMA involves using the museum's archive data to create a dynamic, ever-changing presentation that visualizes potential AI dreams.
  • 🔍 The biases inherent in AI systems are being highlighted by artists like Trevor Paglen, who show the real-world implications of these biases.
  • 🌐 AI's multi-dimensional imagination can blend past, present, and future, offering a different perspective on reality that is not confined by human categorizations.
  • ⚖️ There is a concern that AI, when deployed by large corporations, could lead to further consolidation of wealth and power, potentially exacerbating social inequities.
  • 🌱 Artists bring a unique perspective to the conversation about AI, informed by a deep historical understanding of the nature and role of images in society.

Q & A

  • How do artists use AI differently from the general public?

    -Artists use AI not just as a tool, but also as a means to explore and challenge the technology. They intervene at a high level, contemplating existential questions about free will and perception, and often attempt to make AI do something it wasn't originally designed to do.

  • What is the difference between supervised and unsupervised learning in AI?

    -Supervised learning involves humans tagging information to guide the AI, while unsupervised learning allows the AI model to tag and categorize data on its own, which can lead to more speculative and imaginative outcomes.

  • How does Refik Anadol's work at MoMA differ from conventional AI applications?

    -Anadol's work at MoMA diverges from conventional AI by not simply mimicking reality or following labeled data. Instead, it explores the potential for AI to dream and speculate, creating a new kind of multi-dimensional imagination that blends past, present, and future.

  • What is the 'latent space' that Refik Anadol refers to?

    -The 'latent space' is a concept in AI where the algorithm processes and represents data points in a multi-dimensional space. Anadol uses this concept to navigate through the AI's internal representations to reconstruct potential AI dreams.

  • How do AI systems reflect biases and cultural influences?

    -AI systems are trained on datasets that are inherently biased, reflecting the cultures and values of the society from which they originate. These biases can have real-world implications and are often exposed through the work of artists and researchers who study and critique AI systems.

  • What is the 'generative turn' that Kate Crawford refers to?

    -The 'generative turn' is a term used by Kate Crawford to describe a pivotal moment where traditional understandings of creative fields are rapidly changing due to the advent of AI and its ability to generate new content.

  • What is the concern regarding the deployment of AI tools by large corporations?

    -The concern is that AI tools, when deployed by large corporations, could lead to a massive consolidation of wealth and political power, potentially resulting in an increasingly inequitable society.

  • How does Trevor Paglen's work 'Behold these Glorious Times!' challenge AI systems?

    -Paglen's work challenges AI systems by exposing the training data sets they use, revealing the inherent biases and the real-world implications of these biases. It also questions the oversimplification of complex human experiences and identities into single labels.

  • What is the significance of Marcel Duchamp's approach to art in the context of AI and technology?

    -Duchamp's approach, which involved redefining the role of the artist and the nature of art in response to industrial production, is significant because it mirrors the current challenge artists face with AI and technology, forcing a reevaluation of creativity and authorship.

  • How does the evolution of human and machine interaction reflect in the design over time?

    -The evolution is reflected in designs like the OCR-A font, which was initially created for machine readability, and later, machines were programmed to make concepts as readable as possible by humans, indicating a symbiotic growth process.

  • What is the potential alternative use of AI tools suggested by the speakers?

    -The speakers suggest that AI tools could be used in ways they were not originally designed for, such as making them inefficient or working against their intended purpose, to challenge the expectations and norms associated with these tools.

  • What future challenges does Refik Anadol foresee for AI algorithms?

    -Anadol foresees challenges related to creativity, questioning who will define reality, and the emergence of a collective consciousness through AI algorithms that utilize collective memories to create shared dreams.

Outlines

00:00

🤖 AI and Artistic Intervention

The first paragraph discusses the pervasive yet poorly understood role of AI in daily life. It highlights the dual approach of artists towards AI: using it as a tool and aiming to increase public understanding of AI. The conversation touches on the passive acceptance of technology and the desire of artists to engage with it at a deeper level, raising existential questions about free will and perception. Artists are portrayed as innovators who can repurpose existing tools, including technology, to challenge norms and explore new possibilities. The discussion also covers recent AI advancements and the shift from supervised to unsupervised learning, exemplified by the 'Unsupervised' exhibition at MoMA, which uses AI to reimagine and speculate on new forms of art beyond human categorization.

05:02

🎨 The Generative Turn and AI's Cultural Implications

The second paragraph delves into the cultural and political implications embedded within AI systems. It emphasizes the generative turn, a pivotal moment where traditional methods of creation are rapidly evolving due to AI. The speakers challenge the perception of AI as an objective scientific tool, revealing inherent biases from the training data. The discussion includes a collaboration between Kate Crawford and Trevor Paglen, who examined the datasets used to train AI, highlighting the skewed nature of AI's worldview. Trevor Paglen's artwork, “Behold these Glorious Times!”, is mentioned as a way to expose the biases in AI training sets and the real-world consequences of these biases. The paragraph concludes with a critique of the oversimplification of complex realities into singular labels by AI systems and the importance of considering the broader context and impact of these technologies.

10:02

🔍 The Evolution of Human-Machine Symbiosis

The third paragraph explores the historical relationship between humans and machines, noting the ongoing fascination and fear that has been a subject of artistic and design exploration. It references the early 20th-century artistic response to industrialization and the subsequent shift in the definition of art and the role of the artist, as exemplified by Marcel Duchamp. The evolution of human and machine interaction is traced from the 1934 Machine Art exhibition at the Museum of Modern Art to modern font design aimed at machine readability. The paragraph also addresses concerns about the potential for AI to exacerbate wealth and political power disparities, emphasizing the role of capitalism rather than technology itself. The life cycle of an AI system, from the extraction of rare earth minerals to the end of device life, is discussed to illustrate the full planetary cost. The speakers express interest in unconventional uses of AI tools, suggesting a reimagining of their purpose beyond efficiency and work, and hint at future questions regarding creativity, reality, and the definition of consciousness in the age of AI.

Mindmap

Keywords

💡AI

Artificial Intelligence (AI) refers to the simulation of human intelligence in machines that are programmed to think like humans and mimic their actions. In the video, AI is a central theme as artists use it as a tool to create art and provoke thought about free will, perception, and the existential questions of machine versus human will. It is also discussed in the context of how AI systems are trained and the biases that can be embedded within them.

💡Supervised Learning

Supervised learning is a type of machine learning where an algorithm is trained on labeled data. The script mentions this as a conventional mode where humans tag information, allowing AI to learn and make predictions. For example, if shown many pictures of a pencil, an AI can learn to identify a pencil.

💡Unsupervised Learning

Unsupervised learning is a type of machine learning where the algorithm identifies patterns in data without prior labels or categories. It is depicted in the video as a method that allows machines to create their own classifications and 'dream' or speculate, leading to a more imaginative and less human-centric view of data, as seen in the 'Unsupervised' exhibition at MoMA.

💡Machine Learning Model

A machine learning model is a system that uses statistical methods to analyze and learn from data. In the context of the video, the model is used to create art by interpreting and transforming data from MoMA's collection, building a complex classification system that serves as a basis for generating new, speculative artwork.

💡Black Box

The term 'black box' in the video refers to the opacity of AI systems, where the inner workings and decision-making processes are not transparent or understandable to humans. This concept is central to discussions about the trustworthiness and ethical implications of AI.

💡AI Dreams

In the video, 'AI dreams' symbolize the potential for AI to generate new realities and ideas that do not exist in the human world. These are the outputs of AI when it operates in unsupervised learning mode, creating a multi-dimensional imagination that blends past, present, and future, and challenges human biases and categorizations.

💡Generative Turn

The 'generative turn' mentioned by Kate Crawford refers to a pivotal moment in the history of technology and art where the way we understand and create in fields like illustration, film, and publishing is rapidly changing due to advancements in AI and machine learning.

💡Algorithmic Bias

Algorithmic bias refers to the inherent prejudice in AI systems that can arise from the data on which they are trained. The video discusses how these biases can have real-world implications and how artists like Trevor Paglen are exposing these biases through their work.

💡Training Data Sets

Training data sets are the collections of data used to instruct AI systems. The video script discusses how these data sets are not neutral but are skewed from the start, which influences how AI perceives and interacts with the world.

💡Rare Earth Minerals

Rare earth minerals are essential materials used in the manufacturing of electronic devices, including those that power AI systems. In the video, they are mentioned as part of the life cycle of an AI system, highlighting the environmental costs and the need for a broader understanding of the resources that go into creating AI technology.

💡Collective Consciousness

The concept of collective consciousness in the video is tied to the idea that AI algorithms could help in creating a shared human experience by using collective memories to generate dreams and potentially a shared sense of awareness. This speaks to the potential of AI to influence and shape human culture and perception.

Highlights

AI is increasingly integrated into daily life, yet there's a lack of understanding about it.

Artists are using AI not just as a tool, but to increase public understanding of AI.

Artists intervene in AI processes to explore existential questions about free will and perception.

Artists repurpose existing tools in the world to create unexpected outcomes.

Unsupervised learning allows AI to create without human-imposed labels, leading to unique insights.

AI's multi-dimensional imagination transcends human biases and categories.

AI systems are not objective; they are skewed by the data they are trained on.

Trevor Paglen's work reveals the biases in AI training data and their real-world implications.

AI's value system differs from human values, offering a unique perspective.

The generative turn is a pivotal moment in how creative industries will operate.

AI's full life cycle, from creation to disposal, has a significant planetary cost.

Artists bring a unique perspective to the conversation about AI and its impact on society.

AI has the potential to consolidate wealth and political power, leading to an inequitable society.

Artists are exploring how to use AI in ways it was not designed for, challenging its intended use.

AI algorithms may redefine human creativity and our understanding of reality.

Artists have historically responded to technological advancements, redefining their roles and the nature of art.

The evolution of human and machine interaction is a continuous process of adaptation and growth.

AI can help solve complex problems, but its deployment by corporations raises concerns about wealth and power consolidation.