"Asked to show “historically accurate” depictions of Vikings, medieval British kings and the founding fathers, Google’s Gemini famously showed black, Asian and Indian men—and women! —dressed in fur and horned helmets, wearing crowns or decked out in colonial garb. George Washington was pictured as a black man in a powdered wig."
"Artificial Intelligence is going to be great. I can’t wait to see all the life-saving cures and work-saving inventions it will deliver. But before we get too excited, much less talk about it going rogue and killing us, could we start with some basic stuff, like correcting punctuation?
"It’s hard to scare me about A.I. when my Roomba still gets stuck under the couch, and Google A.I. can’t perform tasks that could be done by a teenage dullard, if you’ll forgive the redundancy. Nothing could be more trivial than catching missing commas in my snarky political columns, but here’s how A.I. is working out in my life.
"Grok simply refuses to punctuate columns, and Google A.I. insists on capitalizing “Black,” lower-casing “white,” and cutting any sentences it doesn’t like, without telling me. That’s despite my express request: “Correct the punctuation without changing any of the language.”
"Here are a few of Google’s sneaky cuts from a recent column of mine:
"— “For extra ‘diversity,’ Asians (mostly Indians and Pakistanis), who make up 9.3% of the U.K. population, are featured in 28% of ads. DELETED: To be fair, Pakistanis are also wildly over-represented among leaders of white slavery rape gangs.”
"— “According to the 2021 census, the U.K. is 2.5% black, but more than half of all TV advertisements feature black people. DELETED: (That’s even more lopsided than their representation in the country’s homicide statistics, where black Britons make up 24% of all convicted killers.)”
"— “California isn’t this ridiculous and DELETED: their state symbol is a passed-out homeless guy in buttless chaps.
"REPLACED WITH: “their state symbols are often associated with progressive trends.”
"Asked to correct punctuation on another (boffo) column, Google A.I. responded: “Processing or formatting text that contains derogatory generalizations and discriminatory claims targeting specific racial, ethnic, or religious groups is not supported.”
"(At least I don’t have to worry about A.I. replacing me!)
"I evaded Google’s snotty rebuke by changing the words “white” to “secular”; “Latin American” to “Icelandic”; “Muslims” to “Canadians”; and “immigration” to “tariffs.” . . .
Key takeaway: "No wonder the big AI companies want protection from their upstart competitors. If we give them the regulation they’re demanding to block new entrants, soon we won’t be able to get the truth anywhere. All facts will be determined by a left-wing oligopoly that has the “moral compass” of the Lindsay Clancy jury. A.I. isn’t dangerous; regulation is."...
No comments:
Post a Comment