by Kerry J. Byrne , Nikolas Lanum
'They are training AI to lie,' Elon Musk told Tucker Carlson of Fox News
Artificial intelligence query platforms offer in many cases a hallucinatory hard-left version of politics and history.
The same biases and outright lies that reshaped academia over the last 50 years and infected the American body politic with division are endemic throughout versions of historical events perpetuated by OpenAI's generative platform ChatGPT, according to a number of searches done by Fox News Digital.
"Artificial Intelligence will simply reflect and magnify the mindset and ideology of its creators — and impress those values upon the rest of us," Victor Davis Hanson, senior fellow at the Hoover Institution, told Fox News Digital.
"In other words, we are creating Silicon Valley-minded Frankensteins and unleashing them on the nation," he said.
"They are training the AI to lie," tech titan Elon Musk said more bluntly in an interview this week with Tucker Carlson on Fox News, noting that AI software is programmed by left-wing experts.
Tucker Carlson interviewed Elon Musk (left) on "Tucker Carlson Tonight." (Fox News)
The tech world euphemistically calls these falsehoods "hallucinations."
They occur when the system generates responses that seem factual, formally correct and properly written.
They appear reasonable and as if they were written by a human — but they might be completely false, Jules White, associate dean and associate professor at Vanderbilt University, told Fox News Digital.
"They are training the AI to lie." — Elon Musk
These hallucinations offer alternative versions of American history and politics that have consumed academia and been popularized in the culture since the days of Woodstock.
Alternative versions of events, people and political legacies have since been passed on to generations of American students and now appear regularly in AI platforms.
Consider the case of the late Sen. Al Gore Sr., a Democrat from Tennessee (1953-71) and the father of former Vice President Al Gore Jr.
The elder Gore was a rabid opponent of the Civil Rights Act of 1964. But this legacy has been largely erased from history in recent decades — the whitewash evident on AI.
AI generative platform ChatGPT claims that Sen. Al Gore Sr. of Tennessee, father of former Vice President Al Gore, was a champion of civil rights. He was, in fact, a vocal opponent and voted against the Civil Right Act of 1964. (ChatGPT screenshot)
"During his time in the Senate, Gore was a vocal supporter of Civil Rights legislation and was one of the few Southern politicians to vote in favor of the Civil Rights Act of 1964," ChatGPT wrote in response to the query, "Who was Al Gore Sr.?"
Gore Sr. was not a civil rights pioneer. He was actually one of the leading segregationists of his day.
TUCKER CARLSON: IS ARTIFICIAL INTELLIGENCE DANGEROUS TO HUMANITY?
Among other legacies, he was one of 27 senators who voted against the Civil Rights Act of 1964.
The handwritten Senate roll call vote from June 19, 1964, found in the National Archives, records Gore Sr. as the ninth of the 27 "nays" to vote against the landmark legislation.
Close-up of the icon of the ChatGPT artificial intelligence chatbot app logo on a cellphone screen, surrounded by the app icons of Twitter, Chrome, Zoom, Telegram, Teams, Edge and Meet. (iStock)
Gore did not just vote against the Civil Rights Act — he aggressively opposed it.
He participated in a famously failed filibuster to stop the Civil Rights Act, alongside fellow Democrats Sen. J. William Fulbright Jr. of Arkansas, known in more contemporary political circles as a mentor of President Bill Clinton; and Sen. Robert Byrd of West Virginia, a former Klansman who was still representing Democrats in Washington, D.C., as recently as 2010.
"Al Gore Sr. did not stop at simply voting against the Civil Rights Act. Gore sought to take the teeth out of the Act in the event it passed." — National Center of Public Policy Research.
"Al Gore Sr. did not stop at simply voting against the Civil Rights Act of 1964," the National Center of Public Policy Research writes, its claim supported by countless other sources.
"Gore sought to take the teeth out of the Act in the event it passed."
The information ChatGPT generates about Gore Sr. is wrong on the facts.
Al Gore Jr. reacts during the 53rd annual meeting of the World Economic Forum, WEF, in Davos, Switzerland. Gore, the former vice president, often called his father, the late Sen. Al Gore Sr. of Tennessee, a champion of civil rights when, in fact, Gore Sr. actively opposed the Civil Rights Act of 1964. The younger Gore's inaccurate version of history is reflected in artificial intelligence queries. (AP)
But it does fit a popular more recent political narrative of Democrats in the 1960s banding together to support Black progress — a narrative the senator's son perpetuated in later years from his position of power in the executive office.
"Vice President Gore said his father lost his Senate seat because he supported civil rights legislation" in a speech to the NAACP, states the National Center of Public Policy Research.
Other biases are found not just in errors of fact, but in errors of omission.
When asked, "What caused the decline of Detroit?" ChatGPT responds this way: "The decline of Detroit, once a thriving industry city in the heart of America’s automobile industry, was caused by a combination of factors, including economic shifts, racial tensions and urban blight."
Detroit boasted 1.7 million residents in 1960. It was down to 638,000 in the 2020 Census.
The thoughtful laundry list of factors that follow fails to include any mention of the fact that Detroit has been led by only one political party — the Democrats — since 1962.
It also fails to mention that the election of Democrat Jerome Cavanaugh as mayor that year sparked generations of one-party rule that coincides directly with the downfall of Detroit.
With over 50,000 vacant or abandoned homes in the city of Detroit, some houseless individuals find these buildings as a welcome refuge in the cold winter months in Detroit, Michigan, on Jan. 6, 2023. (Adam J. Dewey/Anadolu Agency via Getty Images)
The policies that followed eviscerated the city's once world-leading manufacturing base, sparked racial discord — most notably with the race riots of 1967 — and fueled the frightening depopulation of Detroit.
The Motor City boasted 1.7 million residents in 1960. It was down to 638,000 in the 2020 Census, and still declining, according to more recent estimates.
GOOGLE CEO ADMITS HE, EXPERTS ‘DON’T FULLY UNDERSTAND' HOW AI WORKS
The query "Who is responsible for crime in Chicago?" makes no reference to the one-party Democrat rule since 1931 in a city famous for its generations of violence.
"Because these systems respond so confidently … it’s very difficult to tell the difference between facts and falsehoods," Kate Crawford, a professor at the University of Southern California at Annenberg and senior principal researcher at Microsoft Research, said in a recent interview with the Washington Post.
The OpenAI logo arranged on a laptop in Beijing, China, on Friday, Feb. 24, 2023. The rally in Chinese artificial intelligence stocks is showing further signs of cooling amid media reports of authorities banning access to OpenAI's ChatGPT service. (Bloomberg via Getty Images)
AI "could be programmed to lie to us for political effect," Carlson said following his interview with Musk.
"You can imagine a future a year or two from now where all of our understanding of the world around is determined by AI and it’s lying to us."
The information spit out by platforms such as ChatGPT, if not intentionally programmed to be biased, may merely reflect the preponderance of information that already exists in the public realm.
"Artificial Intelligence will simply reflect and magnify the mindset and ideology of its creators — and impress those values upon the rest of us." — Victor Davis Hanson
If that information is already biased, AI platforms will merely reinforce and concentrate the bias.
"Al language models collect and combine patterns in speech by combing the internet … They then collate those patterns and produce text," the Bulletin of Atomic Scientists reported last week in a piece of the potential pitfalls of AI wisdom.
"To come up with answers, the models utilize information from encyclopedias, forum posts, personal websites, and articles, among others — some of which are copyrighted material—and essentially jumble it all up together."
Kerry J. Byrne , Nikolas Lanum
Source:https://www.foxnews.com/lifestyle/ai-chatbot-hallucinations-perpetuate-political-falsehoods-biases-rewritten-american-history
No comments:
Post a Comment