Meta AI's Scientific Paper Generator Reportedly Produced Inaccurate and Harmful Content

November 15, 2022

Meta AI's scientific paper generator, allegedly producing inaccuracies and offensive content, underscores the importance of safe and secure AI practices. The model was reportedly trained to avoid generating papers on sensitive topics but still occasionally produced harmful or inappropriate content.

JOIN US at HISPI Project Cerebellum to contribute to responsible AI governance through our TAIM (Train, Align, Interpret, Manage) system. Together, we can establish trustworthy AI incident databases and prevent similar issues from arising in the future.

Matched TAIM controls

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

MAP 4.1 — similarity 0.715, rank 1. TAIM detail and related incidents →
MEASURE 2.6 — similarity 0.704, rank 2. TAIM detail and related incidents →
MAP 4.2 — similarity 0.691, rank 3. TAIM detail and related incidents →

Alleged deployer: meta-ai, meta, facebook
Alleged developer: meta-ai, meta, facebook
Alleged harmed parties: minority-groups, meta-ai, meta, facebook, minority-groups

Source

Data from the AI Incident Database (AIID). Cite this incident: https://incidentdatabase.ai/cite/399

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.