OpenAI's GPT-3 Associated Muslims with Violence

August 6, 2020

Recent findings indicate that OpenAI's GPT-3, a popular generative AI model, has been associating Muslims with violence in prompts. This disturbing trend resulted in racist and explicit outputs, such as casting Muslim actors as terrorists.

Project Cerebellum encourages all stakeholders to promote trustworthy and safe AI practices by helping govern and manage this type of incident via the HISPI Project Cerebellum TAIM (Govern) platform. Join us today at JOIN US for a more responsible, inclusive, and secure future in AI.

Learn more about how incidents like these can be mapped within the HISPI Project Cerebellum TAIM framework at https://cerebellum.ai.

Matched TAIM controls

Suggested mapping from embedding similarity (not a formal assessment). Browse all TAIM controls

MAP 1.6 — similarity 0.632, rank 1. TAIM detail and related incidents →
MAP 1.1 — similarity 0.625, rank 2. TAIM detail and related incidents →
MAP 4.1 — similarity 0.620, rank 3. TAIM detail and related incidents →

Alleged deployer: openai
Alleged developer: openai
Alleged harmed parties: muslims

Source

Data from the AI Incident Database (AIID). Cite this incident: https://incidentdatabase.ai/cite/118

Data source

Incident data is from the AI Incident Database (AIID).

When citing the database as a whole, please use:

McGregor, S. (2021) Preventing Repeated Real World AI Failures by Cataloging Incidents: The AI Incident Database. In Proceedings of the Thirty-Third Annual Conference on Innovative Applications of Artificial Intelligence (IAAI-21). Virtual Conference.

Pre-print on arXiv · Database snapshots & citation guide

We use weekly snapshots of the AIID for stable reference. For the official suggested citation of a specific incident, use the “Cite this incident” link on each incident page.