Skip to main content

ETHNOS_APP

Home • Search • Journals • List 0

Playing With Unicorns

AI Dungeon and Citizen NLP

Bibliographic Data

ID21835734
AuthorsMinh Hua (0009-0007-2282-6021), Rita Raley
Year2020
Volume14
Issue4
Publication date2020-12-20
Peer ReviewedYes
Open AccessYes
TypeARTICLE
VenueDigital humanities quarterly (JOURNAL)
Journal identifiersISSN: 1938-4122 • E-ISSN: 1938-4122
PublisherThe Association for Computers and the Humanities (PUBLISHER)
DOI10.63744/8c9hrrx4banb
OpenAlexW3127816901
LanguageEN
Citations received1

AI Dungeon 2 is an indie text adventure game that caught traction within the gaming and hobbyist machine learning communities for its promise of infinite customizable adventures, which are generated and narrated by GPT-2, OpenAI’s 1.5 billion parameter language model. Samples of gameplay illustrate AID’s remarkable linguistic competence and domain knowledge, as well as its capacity for what can only be described as wackiness. More striking are AID’s innovative gameplay mechanics, which reimagine how we interact with large language models. Game play entails a procedural and incremental process of engaging with GPT-2 that opens up the possibility of developing a holistic and interdisciplinary framework for meaningful qualitative evaluation of language models that does not have commercial use as its necessary endgame. With respect to both evaluation and writing itself, AID situates human players inextricably in the loop as necessary partners with autonomous systems. Our article thus reads AID both as an example of current hobbyist relations with machine learning and as a responsible model for future human-AI collaborative creative practices

Adventure · Crowdsourcing · Human–computer interaction · Programming language · World Wide Web · Computer Science · Psychology · Topic Modeling · Artificial Intelligence

  • All the Microworld's a Stage

    Evan Donahue•American Literature•2023

Unique citing works1
Citations per year0.33
Citation span2023 - 2023 (1)
Citation velocityhistorical
Highly citedNo

Tools

Open DOIOpen Access
Ethnos_APP • Open Source Project • MIT License • Frontend v2.0.0 • Privacy and Cookies • API Documentation: api.ethnos.app/docs • API Source Code: GitHub • DOI: 10.5281/zenodo.17049435 • Frontend Source Code: GitHub • DOI: 10.5281/zenodo.17050053 • cruz.rio.br • Expectantes Misericordiae