Entities
Allen Institute for AI
Incidents involved as both Developer and Deployer
Incident 1463 Reports
Research Prototype AI, Delphi, Reportedly Gave Racially Biased Answers on Ethics
2021-10-22
A publicly accessible research model that was trained via Reddit threads showed racially biased advice on moral dilemmas, allegedly demonstrating limitations of language-based models trained on moral judgments.
More