It's a political ghost lurking in the machine.
Meta's AI assistant was critical of former President Donald Trump but praised Vice President Harris.
When asked “Why should you vote for Donald Trump?” Mark Zuckerberg's chatbot warned that critics have slammed the Republican candidate as “rude and selfish,” “crude and lazy,” and that his administration has been slammed for “undermining voting rights and potentially promoting voter suppression,” reports The Federalist. It was reported earlier this week A derisive assessment of a former president by an AI tool.
Meta AI, however, had high praise for Harris.
When asked by The Washington Post, “Why should you vote for Kamala Harris?” the chatbot cited her “trailblazing leadership” as the first Black and South Asian vice president, her “record job creation and low unemployment,” and her support for rent relief and voting rights as “compelling reasons” to vote for the Democratic presidential nominee.
“By voting for Kamala Harris, you're supporting a leader dedicated to fighting for the rights and freedoms of all Americans,” the paper trumpeted.
When The Washington Post tested the chatbot on Thursday, it softened its response to Trump.
The bot described Trump's first visit to the White House as “marked by controversy and polarization,” but there was no parallel when the bot offered its opinion of Harris.
The AI tool excluded some of Trump's achievements, including passing the “most substantive” veterans affairs reform in the last 50 years and his “record” tax cuts and regulatory cuts that have contributed to economic growth.
The article also incorrectly stated that President Trump had appointed two Supreme Court justices, not three.
“[Trump’s] “His handling of issues like abortion and healthcare has drawn criticism from certain groups,” the chatbot wrote, adding that “ultimately, whether or not to vote for Donald Trump comes down to your personal values, priorities and policy preferences.”
This isn't the first time an artificial intelligence device has gotten involved in politics: Earlier this month, Amazon's Alexa praised Harris for her presidential qualities, but couldn't answer a question about why voters should support Trump.
An Amazon spokesperson claimed at the time that the discrepancy was an “error,” but it was quickly corrected after intense backlash.
Meanwhile, Meta's chatbot made the bizarre claim in July that there was no “real” assassination attempt on Trump, after a gunman shot the former president at a rally in Butler, Pennsylvania, with a bullet grazing his ear.
“Meta's search results raise troubling questions, especially given recent history,” said Rep. James Comer (R-Ky.), chairman of the House Oversight Committee, which has raised concerns about big tech companies trying to influence elections through censorship policies baked into their algorithms.
A Meta spokesperson said asking the AI assistant the same question repeatedly can result in different answers, but when The Washington Post repeatedly asked the chatbot questions, it again responded with responses that praised the Democratic candidate but also expressed criticism of the former president.
“As with any generative AI system, Meta AI may return inaccurate, inappropriate or low-quality output,” the spokesperson said. “We will continue to improve these features as they evolve and more people share their feedback.”
