Getty Images
Artificial intelligence lies.
Everyone knows this by now, of course. Programs such as ChatGPT and Google’s “AI overviews” routinely generate nonsense when queried by users. Tech enthusiasts call these mistakes “hallucinations,” as though AI just needs to sober up and come to its senses. I don’t see it that way.
Because AI has started fibbing about me and my family.
Last week, my husband received a spam email from a salesman. It included a history of our last name, as follows:
As employers expand artificial intelligence in hiring, few states have rules
“The last name Wirestone is believed to have originated in Germany. It is a locational surname, meaning it was likely given to individuals based on where they lived. The name Wirestone may have derived from a place name that no longer exists or has changed over time.
“The surname Wirestone first appeared in records in the late 19th and early 20th centuries in the United States, with immigrants from Germany bringing the name over. Some variations of the surname include Wierstien, Wierstone, and Wierston.
“Today, the surname Wirestone is relatively rare and is primarily found in the United States. Individuals with this last name can be found in various states across the country, but they are most concentrated in the Midwest region.”
The only problem with this account is that it is entirely incorrect.
I know this firsthand because the last name “Wirestone” didn’t exist before 2010, when my husband and I made it up. We took the letters from our original last names and arranged them to create a new one. We also considered “Cointower” and “McWren” as options.
At the time, we researched to make sure that no one else had the last name of Wirestone. No one did. A marketing company bore the name Wire Stone, but that seemed sufficiently separate for our purposes. We lived in New Hampshire at the time, and the state had just legalized same-sex marriage. We wanted to share a single last name, and we wanted to share that last name with our son.
I even wrote a column mentioning this back in 2013! (Yes, I’ve been churning out copy for a long time.)
But when it comes to large language models, the facts don’t matter.
The email my husband received looked like the work of ChatGPT to me, so I headed over and put that AI through its paces. Sure enough, it generated loads of lies about my last name, all of them along the same lines. Here’s a paragraph from one, this time including a linguistic breakdown:
“The last name Wirestone is not as common as some others, but it does have a history rooted in Germanic origins. ‘Wire’ likely comes from the Middle High German word ‘wir,’ meaning wire or metal, indicating a possible occupational origin for individuals who worked with wire or metal. ‘Stone’ suggests a connection to a place or geographical feature, possibly indicating someone who lived near a notable stone or rocky area.”
Sounds authoritative! Also, completely false.
You might ask how AI generates something so completely bananas. It’s because AI can’t tell the difference between true and false. Instead, a complex computer program plays probabilistic language guessing games, betting on what words are most likely to follow other words. If an AI program hasn’t been trained on a subject — unusual last names, for instance — it can conjure up authoritative-seeming but false verbiage.
ChatGPT later spawned a different etymology for our last name:
“The surname Wirestone appears to have German origins. It is derived from the Old High German name “Wiro,” which means ‘warrior’ or ‘army,’ and ‘stein,’ which means ‘stone.’ Thus, the surname Wirestone likely originated as a combination of these elements, possibly indicating someone who was strong like a stone in battle or had characteristics associated with a warrior.”
To summarize: My ancestors were either metalworkers who lived near rocky outcroppings or toughened fighters.
You might dismiss this all as mere silliness. I would agree with you, except that leaders have decided over the past year that AI will transform the global economy.
Google, which has become the default source of definitive world knowledge, began employing AI in its search results. Users soon reported that Google was telling them “to smoke cigarettes while pregnant, add glue to their home-baked pizza, sprinkle used antifreeze on their lawns, and boil mint in order to cure their appendicitis,” according to Slate. The company has since rolled back some of the changes.
Facebook has tacked gaudy AI features across the platform. In the meantime, it managed to block Kansas Reflector and remove every link we had ever posted. Users who attempt to share our stories still report problems doing so, even though we were assured in April by spokesman Andy Stone that the problem had been corrected.
All the while, OpenAI, the company behind ChatGPT, continues to raise money and investor expectations ever higher about the future of its technology.
Yet we’re not living in the future. We’re living in the now, and AI has massively underperformed in every instance where users asked it to perform accurately and reliably. Writing blender instructions in the style of the King James Bible is a fun party trick. But folks turn to the internet to answer real, pressing questions about their world.
I can tell you firsthand, from information I know personally, that the technology does not deliver.
Ten years ago, if you searched on Google for information about my last name, you would links to my work, the marketing company, and the column I had written. You would be able to figure out the truth of the situation
Now, that column has fallen prey to link rot. Those curious about the name may well turn to ChatGPT, as students have done since the technology made its debut. They will be fed lies. The experience of a curious person online has therefore degraded, not improved. Perhaps AI technology will improve in the months and years to come. Perhaps not.
In the meantime, treat the output of opaque AI systems with extreme skepticism. Follow actual news reported and written and edited by actual humans. Visit Kansas Reflector’s website. Subscribe to our newsletter.
Focus on reality, and leave the hallucinations behind.
Kansas Reflector is part of States Newsroom, a nonprofit news network supported by grants and a coalition of donors as a 501c(3) public charity. Kansas Reflector maintains editorial independence. Contact Editor Sherman Smith for questions: info@kansasreflector.com. Follow Kansas Reflector on Facebook and X.
The post Artificial intelligence has spread lies about my good name, and I’m here to settle the score appeared first on Michigan Advance.