One of the stories this week was a natural to paste into Chat-GPT for analysis, and it surprised me. It declared my tagline “nasty” (twice)!
As usual, we start with the story:
Escalation Required
When his grandfather died, Mark Culbertson of Los Angeles, Calif., ordered several small cremation urns from Amazon so his family could share his remains. “My grandmother was confused and asked my mom, ‘Why does this urn already have ashes in it? Did you order it pre-filled?’” Sure enough, one seemed to be filled with cremains. Amazon customer service didn’t seem to grasp the issue. “Amazon wasn’t working with us at all,” Culbertson said. “They just wanted to give us a refund and stop right there.” He explained in the text chat that there are human remains inside. The rep replied that they could give him a $19.99 promotional credit, and “You can keep or get rid of the item.” Culbertson tried to explain that some family was missing their loved one’s remains, but couldn’t get Amazon to understand, so he contacted a local TV station to intercede. They quickly found someone with a brain at Amazon, who promised they would do what they could to get the ashes back to the correct family. The seller was a third party, the company said, but it promised to “put in a procedure” for customer service reps should it happen again in the future. (RC/KABC Los Angeles) …That moment when it suddenly becomes clear that A.I. can be better than some humans at comprehension and empathy.
Specifically, I asked it for its observations regarding the tagline.
“That’s a strong angle because it reverses the expected anxiety,” it said. “Instead of ‘AI lacks empathy compared to humans,’ you’re pointing at ‘some human systems have become so script-bound they already behave worse than the AI we fear.’ The line has a nice ambiguity: Is the criticism of AI hype? Of dehumanized corporate systems? Of poorly trained humans? Answer: yes.”
(Aside: nice use of “we” — “the AI we fear.”!)
“One subtle strength in your line: ‘can be better’ is measured. You’re not claiming AI is better than humans, just that some humans have set a remarkably low bar. That restraint makes it more credible.”
Ouch! Yet true.
When I read the story out loud to Kit, she was confused about where A.I. came in. Only in the tagline, I said. cG noticed too:
“The story doesn’t actually establish AI was involved at all. Your line works as commentary on the comparison, but it asks the reader to make a small inferential leap. Which may be perfectly fine, because the deeper point is: if a human behaves indistinguishably from bad automation, what exactly are we defending as uniquely human? That’s a nastier and more interesting observation than a straight Amazon joke.”
It was definitely not an A.I. chatbot, I replied: because of the typing, or really, almost certainly the click-click-click to send various sentences, I’m confident it was a human. That prompted me go back to screen-shot that specific paragraph from the TV station’s video to use as a story illustration.
“That actually makes your observation sharper, not weaker. Because then the indictment is: A human being, presented with ‘there are human remains in this package,’ performed worse than the AI stereotype. That’s much nastier. The real target becomes not AI, but human de-skilling inside rigid systems — people trained to process tickets rather than think.”
Ah, thinking. A.I. can’t do it, but it “knows” how important it is. Satisfactory!
But “nasty” twice? Touchy! 😆
And a Title?
I told cGPT I was blogging this, and asked how it would title it. Here’s how it presented the first of the 14 titles it suggested:
When Humans Fail the Turing Test ← probably my top pick
…and then it explained at the bottom, “It immediately communicates the inversion and is compact enough for a blog title.” I went with what I had in mind instead. But I put its suggestion in as the Excerpt for search engines. 🙂
—
Style Note: When I’m writing, I put periods in A.I. as it is an initialism, not an acronym (a pronounceable word formed by combining the initial letters of a multiword name, typically presented without periods — my favorite being “NASA”). The periods also make A.I. much more searchable: searching for “ai” would stop on quite a few normal words, like “said”. But when I’m quoting actual written material, I preserve the style that others — including A.I. systems themselves — have chosen. Typographically fussy, perhaps, but intentional and consistent.
– – –
Bad link? Broken image? Other problem on this page? Please Let Me Know using the Help button in the lower right, and thanks.
This page is an example of my style of “Thought-Provoking Entertainment”. This is True is an email newsletter that uses “weird news” as a vehicle to explore the human condition in an entertaining way. If that sounds good, click here to open a subscribe form.
To really support This is True, you’re invited to sign up for a subscription to the much-expanded Premium edition.
Interesting that Chat GPT recoginized so much of the nuance in your post. Makes one wonder if eventually, it’ll develop the ability to play upon such nuance itself, in an objective fashion.
Like many, I’m concerned about the possible ramifications of A.I. in everyday use — not just as displacing humans, but more in a sense of manipulation of us. We’ve certainly shown over the ages how easily we can be manipulated… hardly seems very challenging.
—
Child’s play. -rc
absolutely pedantic comment incoming
A.I. can’t “recognize nuance.” It can’t “understand” anything. But I fully admit that’s a pedantic statement since it is programmed to appear to “understand” or appear to “recognize” things/meaning/etc. It’s quite a testament to the programmers that we have created such amazing machines that can definitely pass the Turing test for many people, but it is scary because only when we recognize that they are just really well written programs that reflect a creator (and a learning corpus) can we stop treating them as if they think.
—
I completely agree with the over-anthropomorphization of the obviously misnamed “A.I.” That said, the software is “recognizing” nuance. It correctly catches subtlety; humans love shorthand to communicate, and borrow familiar terms to do it. I’m confident I understood Doc’s intent, and yours — which, to summarize, is “Don’t forget this is just software, not actual intelligence.” A fact that is important to remind people of now and then. -rc
That someone would return an urn with ashes in it (guessing here) is disturbing. The pile-on of absurdities is kafkaesque.
The A.I. generated photo is indeed powerful. The light vs. dark, A.I. vs. humanity, reverence vs. blind obedience.
—
When I replied that the second one was “superb,” its response was, again, very interesting:
All emphasis from the original. -rc
I was looking for a ZT tie in here. Well, more of a Zero Thought than tolerance. I often experience this in large corporations where there are low trust environments. In this case there is low trust of the agent to respond in way that will not harm the company. CYA rules the day the customer is an afterthought. The results are always less than satisfactory for those on the receiving end. BUT leadership is protected. That is what matters.
Honestly whenever I read or type A.I. without periods it just looks like Al. My Vietnam vet uncle or Weird Al.
—
Yet another good reason. -rc