Research by scientists at the U.K.'s University of East Anglia suggests OpenAI's ChatGPT has a liberal slant.
The researchers asked the chatbot to answer questions on political convictions as it assumed liberal supporters in the U.S., the U.K., and Brazil might answer them, then asked it to answer the same questions without prompting.
The outcomes indicated "significant and systematic political bias toward the Democrats in the U.S., [leftist president] Lula in Brazil, and the Labor Party in the U.K.," the researchers wrote.
These findings add to a growing body of evidence showing chatbots are rife with assumptions, beliefs, and stereotypes that were embedded in their training data.
From The Washington Post
View Full Article - May Require Paid Subscription
Abstracts Copyright © 2023 SmithBucklin, Washington, D.C., USA
No entries found