The following editorial originally appeared in The Seattle Times:
Experts have warned for more than a year that artificial intelligence-fueled misinformation is poised to upend American elections. Washingtonians have already seen it.
Washington Secretary of State Steve Hobbs recently joined three other Democratic secretaries of state from New Mexico, Minnesota and Michigan calling out Elon Musk and his Grok AI search assistant for peddling election misinformation.
Grok is a large language model deployed in association with Musk-owned X, formerly Twitter.
Before Washington’s primary election, Grok provided incorrect dates when people asked about voting deadlines. Whether that was due to malfeasance or AI hallucination remains unknown.
Bad information coming out of AI systems either inadvertently or at the request of users is not unique to Musk and Grok.
Unlike the owners of other artificial intelligence systems, however, Musk not only doesn’t appear interested in reducing misinformation, but he also himself propagates it. Notably, he shared a video with a fake voice-over of Vice President Kamala Harris saying things that the Democratic presidential nominee most assuredly would not really say.
Hobbs and his fellow secretaries requested that the next time users ask for election guidance, Grok direct them to the nonpartisan voting information site CanIVote.org. That’s the eminently sensible approach taken by ChatGPT, a different AI system.
In these hyperpoliticized and hyperpartisan times, there’s no good reason to risk a large language model getting it wrong when there’s a readily available and reliable source.
With hotly contested presidential, congressional and legislative elections just weeks away, voters are sure to encounter more misinformation.
Hobbs and his peers can and should highlight bad practices and bad actors who peddle falsehoods about something as important as elections, but it’s up to voters to educate themselves and seek out reliable sources like official election sites and the local free press.
Musk’s choosing to name his AI system Grok provides a geeky, probably inadvertent cautionary message.
Science fiction author Robert A. Heinlein coined the term “grok” in a 1961 novel, and it soon entered popular usage. It means to understand and empathize with something so thoroughly that it becomes part of you. To grok something is to be changed by it and to see the world differently as a result.
If people grok Musk, his AI search assistant or really any social media, they risk internalizing things that aren’t true and seeing the world through a distorted lens.
Voters, beware.
