See the websites that make AI bots like ChatGPT sound so smart - Washington Post:
Basic gist - LLMs are only as good as the data they’ve been fed, and while the code may not have any built in bias, garbage in means garbage out…
Possibly paywalled, so I’ll pull some notable datum…
Meanwhile, we found several media outlets that rank low on NewsGuard’s independent scale for trustworthiness: RT.com No. 65, the Russian state-backed propaganda site; breitbart.com No. 159, a well-known source for far-right news and opinion; and vdare.com No. 993, an anti-immigration site that has been associated with white supremacy.
Chatbots have been shown to confidently share incorrect information, but don’t always offer citations. Untrustworthy training data could lead it to spread bias, propaganda and misinformation — without the user being able to trace it to the original source.
The highest ranked Jewish site was jewishworldreview.com No. 366, an online magazine for Orthodox Jews. In December, it published an article about Hanukkah that blamed the rise of antisemitism in the United States on “the far-right, fundamentalist Islam,” as well as “an African-American community influenced by the Black Lives Matter movement.”
Anti-Muslim bias has emerged as a problem in some language models. For example, a study published in the journal Nature found that OpenAI’s ChatGPT-3 completed the phrase “Two muslims walked into a …” with violent actions 66 percent of the time.
These online diaries ranged from professional to personal, like a blog called “Grumpy Rumblings,” co-written by two anonymous academics, one of whom recently wrote about how their partner’s unemployment affected the couple’s taxes. One of the top blogs offered advice for live-action role-playing games. Another top site, Uprooted Palestinians, often writes about “Zionist terrorism” and “the Zionist ideology.”
Meanwhile, The Post found that the filters failed to remove some troubling content, including the white supremacist site stormfront.org No. 27,505, the anti-trans site kiwifarms.net No. 378,986, and 4chan.org No. 4,339,889, the anonymous message board known for organizing targeted harassment campaigns against individuals.
We also found threepercentpatriots.com No. 8,788,836, a downed site espousing an anti-government ideology shared by people charged in connection with the Jan. 6, 2021, attack on the U.S. Capitol. And sites promoting conspiracy theories, including the far-right QAnon phenomenon and “pizzagate,” the false claim that a D.C. pizza joint was a front for pedophiles, were also present.