Abstract

For the past few years, we have witnessed eye-opening generation results from AI foundation models such as GPT-3, and DALL-E2. These models have set up great infrastructures for new types of creative generation across various modalities such as language (e.g. story generation), images (e.g. text-to-image generation, fashion design), and audio (e.g. lyrics-to-music generation). Researchers in these fields encounter many similar challenges such as how to use AI to help professional creators, how to evaluate creativity for an AI system, how to boost the creativity of AI, how to avoid negative social impact, and so on. There have been various workshops that focus on some aspects of AI generation. This workshop aims to bridge researchers and practitioners from NLP, computer vision, music, ML, and other computational fields to create the 1st workshop on “Creative AI across Modalities”.

Invited Speakers

Snigdha Chaturvedi
(UNC)

Chris Donahue
(Google Magenta)

Andrew Owens
(UMich)

Niki Kittur
(CMU)

Mark Riedl
(Georgia Tech)

Diyi Yang
(Stanford)

Aaron Hertzman
(Adobe Research)

Accepted Papers

Schedule

For the in-person participants, please go to Room 146B for joining all the talks and the poster session.
08:50 am - 09:00 am (EST) Introduction and Opening Remarks
09:00 am - 09:45 am (EST) Invited Talk Andrew Owens
(UMich)		 Title: Cross-modal Synthesis with Sight, Sound, and Touch
09:45 am - 10:30 am (EST) Invited Talk Mark Riedl
(Georgia Tech)		 Title: Computers, Creativity, and Lovelace
Abstract: In this talk we examine the what attributes we should expect in human-level creative systems, and the mechanisms by which we might achieve them. I provide examples from the domain of automated story generation. I conclude the talk with some informal analysis of recent progress toward AI systems that express creativity.
10:30 am - 10:40 am (EST) Break
10:40 am - 11:25 am (EST) Invited Talk Chris Donahue
(Google)		 Title: Frontiers in Controllable Music Generation
Abstract: For music generation and creative generation more broadly, control is key to unlocking human expression. In this talk, I will discuss the recent improvements in and remaining obstacles to building controllable music generation systems that unlock exciting new expressive capabilities for musicians and non-musicians alike. Additionally, I will discuss control considerations that are more specific to music and argue that text is useful but not sufficient for expressive musical control. As a case study, I will discuss SingSong, a recent system from the MusicLM project at Google which learns to translate vocal performances into instrumental accompaniments, thereby allowing anyone to create rich music featuring their own voice.
11:25 am - 12:10 pm (EST) Invited Talk Niki Kittur
(CMU)		 Title: Scaling Analogical Innovation
12:10 pm - 01:30 pm (EST) Lunch
01:30 pm - 02:50 pm (EST) Poster Session (virtual + in person)
02:50 pm - 3:35 pm (EST) Invited Talk Aaron Hertzman
(Adobe)		 Title: Can Computers Create Art?
Abstract: Can AI algorithms make art, and be considered artists? Within the past decade, the growth of new neural network algorithms has enabled exciting new artforms with considerable public interest, including DeepDream, GANs, VAEs, and diffusion models like DALL-E and Imagen. These tools raise recurring questions about their status as creators and their effect on the arts. In this talk, I will discuss how these developments parallel the development of previous artistic technologies, like oil paint, photography, and traditional computer graphics. I argue that art is a social phenomenon, and discuss possible—but very unlikely—scenarios for when these algorithms could someday be considered artists.
03:35 pm - 04:20 pm (EST) Invited Talk Snigdha Chaturvedi
(UNC)		 Title: Modeling People in Automatic Story Generation
Abstract: Automatic story generation is the task of designing NLP systems that, given a prompt, can produce the text of a story. Most methods for this problem focus on modeling events and their coherence. However, an alternate perspective to story generation can be from the viewpoint of people described in the story. In this talk, I focus on one aspect of modeling people in story generation -- modeling their social relationships. I describe our story generation approach to incorporate a desired social network demonstrating relationships between various people to be mentioned in the story. We propose a model that uses latent variables to incorporate social relationships. Apart from generating coherent stories that reflect the desired social network, the latent variable-based design results in an explainable generation process.
04:20 pm - 04:30 pm (EST) Break
04:30 pm - 05:15 pm (EST) Invited Talk Diyi Yang
(Stanford)		 Title: Improving Everyday Interaction through Human-Centered Text Generation
Abstract: As natural language generation has gained popularity and produced extensive industrial applications, there has been an increasing focus on enabling the use of natural language in human-like interactions. How can we improve such everyday interactions and build language generation systems that are more aware of human factors? In this talk, we take a closer look at human-centric language generation and present two recent works that promote positive language use and summarize daily conversations. Specifically, the first part examines positive reframing by neutralizing a negative point of view and generating a more positive perspective without contradicting the original meaning. The second part demonstrates how more structures of conversations can be utilized to generate better summaries for everyday conversation.
05:15 pm - 05:25 pm (EST) Closing Remarks

Call for Papers

Authors are invited to send the following relevant work, either archival or non-archival, in the AAAI-23 proceedings format:
  • Long paper: Submission of original work up to seven pages for contents and one page for references.
  • Short paper: Submission of work in progress with preliminary results, and position papers, up to four pages for contents and one page for references.
    • Topics including but not limited to:
    • Creative language generation: stories, poetry, figurative languages.
    • Generative model and algorithms for image/audio, and multi-modal/video generation.
    • Theory and analysis for creativity (e.g., humor understanding)
    • Detecting and quantifying creativity
    • Using AI to improve human creativity (e.g., HCI+ML studies to accelerate scientific novelty)
    • Data and resources for creative generation
    • Applications of creative AI generation, such as automatic video dubbing
    • Novel evaluation for creative AI generated outputs
    • Social, cultural, and ethical considerations of creative AI generations, such as racial/gender bias, trustworthiness
    A submission should take the form of a AAAI long or short paper in PDF format using the AAAI style. We will accept submissions of (1) papers that have not been previously published or accepted for publication in substantially similar form; (2) papers that have been published or accepted for publication in recent venues including journal, conference, workshop, and arXiv; and (3) research proposals for future work with a focus on well-defined concepts and ideas. All submissions will be reviewed with double blind policy.


    Open Review submissions website: https://openreview.net/group?id=AAAI.org/2023/Workshop/creativeAI


    Key Dates:

    • Submission deadline: Nov. 18, 2022 (11:59 p.m. Anywhere on Earth)
    • Notification to authors: Dec. 20, 2022 (11:59 p.m. Anywhere on Earth)
    • Workshop date: Feb. 13, 2023