A Human Halo for Mistrusted AI
A lifetime ago, when I was a “rising senior” in high school, I picked up a remaindered copy of Woe Is I while my family was on summer vacation. I loved the book. I read it as we drove around the rural Midwest. I chuckled at its wit, and I learned, finally, what our high school’s proofreader for AP English courses was looking for in comma placement.
With that backstory, it will not surprise you that I don’t use Grammarly.
Yet the tool pushed its way into my notice last week via stories about its “Expert Review” tool. For those who haven’t encountered this particular nightmare yet, Grammarly has rolled out a new generative AI feature that reviews your content from the alleged perspectives of AI personas that use the names of real people, both living and dead.
The living, of course, have not been compensated in any way for their names being used to give credibility to generic AI-generated “feedback.” The use of the names of the dead (some very recent, some more distantly historical figures), has led critics to call this feature “necromancy.”
This is just the latest salvo, of course, against intellectual property in the great generative AI roll-out. It’s also yet another instance of peoples’ likenesses (intellectual or physical) being used in generative AI without their consent.
It likely isn’t even the start of the kinds of tricks we’ll see as companies look to for ways to get users to trust their products’ generative AI features.
Only around 35% of people in the US say that are even “somewhat” willing to trust generative AI (KPMG study). So it’s not surprising that companies that have hugely invested in the technology are looking for ways to shortcut the path to trust and their products’ adoption.
You can see the corporate logic: people don’t trust Claude or ChatGPT or Grammarly (which I have learned has rebranded to the name Superhuman) sufficiently. They do trust Neil deGrasse Tyson, Stephen King, and Gary Marcus (the irony of Grammarly’s decision to include him among their genAI zombies is…wow). So Grammarly has decided to try to steal some trust by association.
This is a tangled web of problems, and I’m not a lawyer, so I can’t address most of them. But I am a marketing professional and, well, a person, so I can say that this attempt to add a human halo to an AI brand 1) goes to show how far the latter has to go still with the general populace and 2) is gross.