You have 4 summaries left

The Ezra Klein Show

‘Artificial Intelligence?’ No, Collective Intelligence.

Fri May 24 2024
AI artmusic generationAI voice modelsdata controleconomic modelsAI-generated contentweird AI artcollaborative technology

Description

This podcast explores the intersection of AI technology and art, focusing on music generation and artistic expression. It delves into the work of artist Holly Herndon, who uses AI to create weirder and more personal art. The podcast also discusses the challenges and implications of using AI voice models in music creation, as well as the importance of human relationships and collaboration in AI art. It explores topics such as data control, ownership, and economic models in the digital age, and the impact of AI on artistic expression and media consumption. The podcast concludes with insights on creating weird and unique AI art and a discussion on a science fiction story and collaborative technology.

Insights

AI art can be weirder and more personal

AI art is often criticized for being generic and unoriginal, but artist Holly Herndon uses AI to create weirder and more personal art.

The importance of human input in AI models

AI and collective intelligence are intertwined in music creation, emphasizing the importance of human input in AI models.

Challenges of imbuing meaning in AI interactions

The artist reflects on the challenge of imbuing AI interactions with meaning and how it can impact the sustainability of those relationships.

Governance and control over personal data

The concept of governance and control over personal data is a significant concern in the digital age.

Diverse economic models for artistic practices

There is a need for diverse economic models that cater to different artistic practices, avoiding a one-size-fits-all approach.

AI-generated content and human creativity

Concerns about AI replacing artists are addressed, emphasizing the importance of human creativity and meaning-making in art despite the rise of infinite media.

Creating weird and unique AI art

AI can lead to a push towards sameness, but the guest artist focuses on creating weird and unique art by training models with specific data.

Collaborative technology and democracy

The podcast discusses a science fiction story about intelligent genetically modified spiders and a book called Plurality focused on collaborative technology and democracy.

Chapters

  1. AI Art and Holly Herndon
  2. AI Voice Models and Music Generation
  3. AI in Music and Human Relationships
  4. AI, Collective Intelligence, and Governance
  5. Data Control, Ownership, and Economic Models
  6. AI, Artistic Expression, and Media Consumption
  7. Weird and Unique AI Art
  8. Science Fiction Story and Collaborative Technology
Summary
Transcript

AI Art and Holly Herndon

00:04 - 07:57

  • AI art is often criticized for being generic and unoriginal, but artist Holly Herndon uses AI to create weirder and more personal art.
  • Herndon's project at the Whitney Biennial involves manipulating AI-generated images of herself to challenge and control the AI's perception of her.
  • Holly Herndon discusses her background in folk singing, techno music, and technology, highlighting the human aspects of techno music despite its synthetic sound.
  • Techno music is described as a coordination technology that facilitates communication and participation through its rhythmic structure.
  • The appeal of creating music that sounds artificial or machine-like may stem from a desire to explore humanity's place in a technologically mediated world.

AI Voice Models and Music Generation

07:38 - 16:17

  • The speaker found it liberating to digitally manipulate their voice using AI technology.
  • Spawn, an AI experiment, was trained on the voices of an ensemble and used for generating music.
  • The early version of Spawn had a low fidelity sound quality, which evolved over time with advancements in technology.
  • The value of using Spawn was in its unique timbre quality and the snapshot it provided of a specific period in machine learning research.
  • Training models directly informed the speaker's work and understanding of technology's impact on art.
  • A more modern voice model called Holly Plus was developed, trained on the speaker's voice, providing higher fidelity performance capabilities.
  • Holly Plus can autonomously perform songs based on written scores with text in phonemes, showcasing advancements in AI music generation.
  • The concept of creating models that mimic a person's voice raises questions about identity, collaboration, and scalability in artistic expression.

AI in Music and Human Relationships

15:57 - 23:22

  • The artist used machine learning to create a model of their voice, allowing them to perform in ways they couldn't alone acoustically.
  • The use of AI in music allowed the artist to explore different styles and songs they wouldn't have attempted themselves.
  • The artist reflects on the challenge of imbuing AI interactions with meaning and how it can impact the sustainability of those relationships.
  • The focus of the artist's work is not solely on high-tech AI advancements but also on showcasing human relationships and collaboration in creating art.
  • The concept of 'protocol art' is introduced, emphasizing collaboration between the artist and audience through shared models and creations.

AI, Collective Intelligence, and Governance

22:58 - 30:22

  • AI and collective intelligence are intertwined in music creation, emphasizing the importance of human input in AI models.
  • There is a debate between viewing AI as a societal output versus an individual technological feat, impacting economic compensation and ownership.
  • Spawning is a new concept similar to sampling but with distinct differences, allowing for performing based on trained information.
  • An organization was founded to address data manners in AI training, aiming for a middle ground between open use and strict IP lockdown.
  • Experiments like Holly Plus and choir recordings for data trust showcase innovative ways to involve communities in AI art while addressing governance issues.

Data Control, Ownership, and Economic Models

30:03 - 36:42

  • The concept of governance and control over personal data is a significant concern in the digital age.
  • Challenges arise from platforms using individuals' data without explicit consent, leading to concerns about lack of control over one's online presence.
  • Efforts are being made to establish universal opt-out standards and tools to give individuals more control over their data inclusion in models.
  • Proposals for collective ownership and compensation in the realm of art challenge traditional business models based on exclusive rights.
  • There is a need for diverse economic models that cater to different artistic practices, avoiding a one-size-fits-all approach.

AI, Artistic Expression, and Media Consumption

36:13 - 44:06

  • People should have the freedom to create unique subcultures and economic models without being constrained by a standardized approach.
  • The concept of all media being training data is explored in a forthcoming book, discussing how once something becomes machine-readable, it can be used for AI training.
  • Creating deliberate training data as artwork is being experimented with, such as composing songs to train an AI on English phonemes.
  • Concerns about AI replacing artists are addressed, emphasizing the importance of human creativity and meaning-making in art despite the rise of infinite media.
  • AI-generated content may lead to a surplus of generic material, potentially sparking a backlash that drives interest towards more unique and human-created works.
  • Personalized content creation based on individual taste profiles could shape future media consumption experiences but may risk stagnation if not balanced with exposure to diverse content.

Weird and Unique AI Art

43:40 - 50:57

  • AI can lead to a push towards sameness, but the guest artist focuses on creating weird and unique art by training models with specific data.
  • The guest artist trained an early voice model on their singing voice and fed it stems from a collaborator's music, resulting in a unique blend of sounds.
  • The guest artist values the weirdness and unpolished nature of AI-generated music over polished and slick productions, aiming to meet the technology where it was at that time.
  • To participate in the world of AI models, one can start by interacting with publicly available models, reading academic research papers, and getting hands-on experience.

Science Fiction Story and Collaborative Technology

50:31 - 51:48

  • The podcast discusses a science fiction story about intelligent genetically modified spiders and a book called Plurality focused on collaborative technology and democracy.
  • Plurality was led by Glenn Vile and Audrey Tang with contributions from various individuals, including the speaker.
  • The book was written in an open collaborative democratic manner, reflecting its theme.
  • The episode credits Annie Galvin as the producer, Michelle Harris for fact-checking, Jeff Gell as the senior engineer, and others involved in production.
1