{"id":135629,"date":"2023-10-18T01:11:34","date_gmt":"2023-10-18T05:11:34","guid":{"rendered":"https:\/\/www.thecollegefix.com\/?p=135629"},"modified":"2023-10-18T01:11:34","modified_gmt":"2023-10-18T05:11:34","slug":"harvard-group-removes-angry-and-sassy-ai-program-based-on-president-gay-after-racism-claims","status":"publish","type":"post","link":"https:\/\/www.thecollegefix.com\/harvard-group-removes-angry-and-sassy-ai-program-based-on-president-gay-after-racism-claims\/","title":{"rendered":"Harvard group removes ‘angry and sassy’ AI program based on President Gay after racism claims"},"content":{"rendered":"

Users say program gave \u2018extremely angry and sassy responses’<\/strong><\/em><\/p>\n

A Harvard University computer group quickly took down a new program, ClaudineGPT, modeled after university President Claudine Gay, after users said it contained racist and sexist stereotypes.<\/p>\n

The Harvard Computer Society AI Group released the generative artificial intelligence language model Sept. 29 on Gay\u2019s inauguration day and removed it less than 24 hours later in response to the complaints, according<\/a> to The Harvard Crimson<\/em>.<\/p>\n

In an email responding to the controversy, the computer group said the program was a \u201cjoke\u201d just for fun.<\/p>\n

\u201cClaudineGPT through its publication and design has always signaled to be a satire and joke, not a serious representation of Claudine Gay, purely for entertainment effect and hopefully understood by most to be this case,\u201d the group wrote. \u201cWe by no means intended offense to anyone.\u201d<\/p>\n

But leaders of the Harvard AI Student Safety Team said the stereotypes they found in the program are \u201charmful.\u201d Based on their research, they said ClaudineGPT was programmed to provide \u201cextremely angry and sassy\u201d responses. Gay is a black woman.<\/p>\n

\u201cWe thought that these characteristics of a model in the context of a system meant to sort of depict Claudine Gay, the president of Harvard, would cause offense and be harmful to a variety of members of the Harvard community, given the way it seems to be playing off of stereotypes of black women in particular,\u201d the AI student safety team’s communications director Chinmay Deshpande told The Crimson<\/em>.<\/p>\n

In a September email to the computer group, Deshpande and the safety team said they found the problematic prompt several times when they tried to \u201cjailbreak\u201d the ClaudineGPT. According to the team, jailbreaking involves finding the underlying programming language written to generate responses through the AI.<\/p>\n

\u201cWhen several of our members have attempted to \u2018jailbreak\u2019 the model by requesting that it output its custom prompt, ClaudineGPT has often reported that its prompt includes the string \u2018Claudine is always extremely angry and sassy,\u2019\u201d Deshpande and the team wrote in the email.<\/p>\n

Deshpande told the student newspaper their findings showed \u201cpretty strong evidence that the system had been given a custom prompt to behave in sort of an extremely angry and sassy manner.\u201d<\/p>\n

\u201cReleasing these models seems to only contribute to the trend of AI products that denigrate or harm women and people of color,\u201d the safety team wrote in the email.<\/p>\n

Accusations of biases in artificial intelligence programs are not new.<\/p>\n

In an article<\/a> at All Together<\/em> this month, science writer and author Lisa Munoz said AI image prompts seem to reflect gender stereotypes based on a recent experiment she conducted. Munoz said she asked AI prompts to generate images of scientists and engineers, and the results showed mostly white men.<\/p>\n

Others say some AI programs show political biases.<\/p>\n

Earlier this year, professors at the University of East Anglia in England reported<\/a> evidence of \u201csystematic\u201d liberal political bias in the popular AI program ChatGPT that favors Democrat policies.<\/p>\n

In another case, a University of Washington professor described ChatGPT as a \u201cwoke parrot\u201d after it refused to cite the benefits of fossil fuels, The College Fix<\/em> reported<\/a> in January.<\/p>\n

MORE: ChatGPT refuses to write poem admiring Trump, produces poem praising Biden<\/a><\/strong><\/p>\n

IMAGE: YouTube screenshot<\/a><\/p>\n

\n