DiversifyAI

Maintaining Diversity of Human Expression in the Age of AI

AI systems are rapidly converging toward statistical averages, limiting the richness and value of their outputs. Our research focuses on developing methods for diversifying AI outputs across creative and technical domains, ensuring that AI enhances rather than diminishes human autonomy and expression.

The Homogenization Problem

AI systems are producing increasingly uniform outputs, threatening diversity of expression and limiting value creation

249 of 250

In a recent study, 249 out of 250 poems generated by ChatGPT without special prompting were all identified by ChatGPT itself as being derived from the same poet (John Keats). This striking homogeneity illustrates how large language models tend to converge toward a "safe" statistical mean.

This homogenization extends beyond poetry to code, design, business ideas, marketing copy, brand slogans, and other domains where innovation requires deviation from the norm. It's not just a problem for artists—it's a critical issue for businesses, consumers, and society at large.

Why AI Outputs Become Homogeneous

Large language models (LLMs) like ChatGPT learn patterns from vast datasets, but they tend to optimize for the "average" response that's statistically most likely to be correct or well-received. This leads to several problems:

Output Homogenization

AI systems converge toward a narrow range of outputs that represent a statistical average rather than the full spectrum of possibilities

Deprivation of User Autonomy

AI tools can replace human decision-making rather than augmenting it, reducing agency and creative control

Economic Implications

Businesses and markets thrive on differentiation and innovation, which are threatened by homogeneous AI outputs

For Business

Learn how AI diversity research can improve competitive advantage, customer experience, and innovation in business contexts

Explore Business Applications

For Creatives

Discover how AI diversity research can enhance creative expression, artistic exploration, and cross-media collaboration

Explore Creative Applications

Get Involved

Join our mission to maintain creative and technical diversity in the age of AI

For Artists and Creatives

I invite fellow poets, narrators, musicians, and visual artists to contribute to this research by sharing your experiences and creative insights. Your perspective is invaluable as we develop techniques that preserve the full spectrum of creative expression.

Ways to collaborate:

  • Submit your identity statement to explore AI-generated interpretations
  • Share your artistic manifesto to develop alternative creative personae
  • Request a cross-medium homage to your creative work
  • Collaborate on research exploring creative applications of diverse AI
Submit Your Vision

For Researchers and Developers

We're always looking for collaborators interested in the technical and social aspects of AI diversity. Whether you're a machine learning researcher, social scientist, or software developer, there are many ways to contribute:

  • Collaborate on research papers and projects
  • Develop tools and techniques for measuring and improving AI diversity
  • Help build applications that demonstrate the value of diverse AI outputs
  • Propose new domains where diversity enhancement is needed
Contact About Research Collaboration

Join Our Research Community

Even if you're not sure how you want to get involved, we'd love to hear from you. Whether you're an artist concerned about AI's impact on creativity, a business leader interested in the competitive implications of AI homogenization, or simply someone interested in this important issue, please reach out.

Get in Touch

If you want to get involved but don't know in what capacity, click this link and send an email (optionally blank, or with whatever information you want).

About Halley Young

Performing research on AI diversity and human-AI collaboration

I'm a postdoctoral researcher at Microsoft Research in NYC, with a focus on diversifying the output of AI in both creative and technical domains. With a background in computer science and expertise in deep learning, generative models, and music AI, I approach artificial intelligence from a unique perspective: how can we ensure AI systems enhance human creativity and autonomy rather than diminish them?

Background and Expertise

My research addresses two related but distinct dangers in current AI development:

  1. Output homogenization: The tendency of AI systems to converge toward a narrow range of outputs that represent a statistical average rather than the full spectrum of human expression
  2. Deprivation of user autonomy: The risk that AI tools replace human decision-making rather than augmenting it

Education

  • PhD in Computer Science, University of Pennsylvania (2017-2023)
  • B.A. magna cum laude, Computer Science, New York University (2014-2017)

Experience

  • Microsoft Research, NYC - Postdoctoral Researcher (2023-Present)
  • Google Brain/Magenta - Research Intern (2021-2023)
  • Riffit - Software Engineer/Researcher (2019-2020)
  • University of Pennsylvania - Teaching Assistant (2020-2021)

Research Areas

Deep Learning Generative Models Steerable AI Music AI Formal Methods Human-AI Interaction AI Diversity Neurosymbolic Models

Contact

Get in touch to discuss research collaboration, creative partnerships, or learn about research findings

Email
Website
SoundCloud