“Hey, what do you know about this ChatGPT stuff?” my brother says to me from the living room when I’m over there recently.
“What do you want to know?” I say.
“No no—I’m asking you. What do you know?” he says.
“No, I know, but what do you want to know?” I say.
“I wanna know what you know about it,” he says.
“I mean, it depends. What are you asking?”
“Does it work?”
“What do you mean?”
“Can it pretend to be you?”
“Yeah. I mean, what?” I said. “You want it to send your wife texts while you’re off having drinks with that twenty-five-year-old ladder-climber of the…female persuasion? In your office, there?”
“Yeah, Don,” my sister-in-law says, cutting vegetables next to me in the kitchen. “Do you?”
“Nah, I already got someone for that,” he says. She turns. “I want to know if it can be me, you know. Like, on paper.”
“That’s a lot to live up to,” my niece says, typing on her phone.
“You mean, like sit in for you at an interview with Goldman Sachs or something?”
“Like, can it write a g–damn self-evaluation report for someone at work so they don’t have to?” he says.
“I don’t know about that.”
“So, what do you know?”
“I know that some guy who makes 300 bucks an hour helping rich kids write letters to schools they probably wouldn’t get into if they weren’t rich kids tried it. And, in the end, he said it basically writes like a sixth grader. I mean, a decently smart sixth-grader, you know. But it can’t do what human beings can do,” I said.
“Which is what?”
“Be a human being. And then write about that.”
“I was hoping it could get me out of this self-evaluation report I gotta write about myself, so I don’t have to,” he says.
“Yeah, I kind of figured that,” I said. “I mean, maybe? Some guy was talking about how he could side hustle a second job writing because ChatGPT does like 80% of his first one, where all he does is just market crap for tech companies.”
“Well, I know her cousin in New York had it banned from their school,” he says, motioning to his niece, still on her phone. “So, I figured people are at least somewhat onto it being used for cheating. But can they sniff it out?”
“The makers themselves already released a tool for that. So, probably.”
“I doubt a plumbing company would have something like that, Don,” my sister-in-law says.
“And a paper was just published talking about not-so-cutting-edge science being able to theoretically sniff out 99% of all possible AI bullsh–,” I said. “Plus, like, half the tech community is talking about putting a moratorium on it, at least for about six months, ever since the latest version got rolled out, back in March.”
“Anyone been caught yet doing it?” he says, kind of sheepishly.
“One dude in China got his doors kicked in by the Chinese SWAT team. But that was because he had it publish some fake story about a bunch of people getting pancaked in a train wreck. And, for some reason, instead of just once, he uploads it 25 times to some blog site, where like 15,000 people see it all over the country.”
“How about in the States?” he says, even more sheepishly, rubbing his chin.
“Well, just yesterday there was a story about a lawyer who used it to file a brief in federal court that was full of a bunch of fake citations and legal opinions.”
“Is he f—ed?” my brother says.
“Yeah, he’s f—ed,” I said.
“It’s not that hard to use, though,” my niece says. “Here, I just used it to type up this self-evaluation about myself.”
“While you two were talking.”
“Get the f— out of here. Lemme see that,” he says. He looks at her phone. “Hey, this is pretty good.”
“You just ask it a bunch of questions, and then if you don’t like what it says, you ask it more until you do,” my niece says.
“You mean, I gotta argue with the f—ing thing until I get what I want?” he says.
“Oh, that shouldn’t be a problem,” my sister-in-law says.
“You also need a bit of an imagination,” I said.
“Never mind,” she says.
“Like how?” he says.
“Like, say you want to know how the radioactive chemical market’s holding up in Uzbekistan. But you wanna know from the point of view of a hedge fund guy who’s looking to corporate raid their biggest radioactive chemical company, with some Saudi backing. It won’t tell you that outright, so you gotta modify your questions. And then tell it that you want it to answer you like Paul Lynde in the Hollywood Squares, with Emmanuel Lewis sitting on his lap doing his best Gollum impersonation of something Chewbacca would say to Captain Picard if they were flying over the planet Dune.”
“Huh. Yeah, I think I could do that,” he says.
“Careful, though,” I tell him. “It remembers everything you say and do.”
“That I say? You mean, with the writing?”
“And if you use it for too long, it starts to adapt your personality.”
“Nuh-uh. What do you mean? To my personality?”
“No, like, your actual personality,” I tell him. “It pretty much becomes you. Like the Alexa or Siri does—like how they learn about you the more they listen to you talk? Only it’ll go online and speak on your behalf to places that request it of you. It basically becomes your spokesperson. Banks, lenders, the cops—stuff like that. But it’s so good that people can’t tell the difference. And it’ll start doing it without telling you, if it thinks you’d think it was a good idea.”
“You’re f—ing lying,” my brother says after a moment of staring at me.
“Yeah, I’m making that up,” I said. “Everything else is true, though. If you don’t believe me, just run everything I said through that AI bullsh– detector.”
“Yeah, sure, okay,” he says, turning back to the TV. “Christ, I don’t know what to believe anymore.”
“Probably for the best,” I said.