SCOTT SIMON, HOST:
Composing emails has never been easier. AI assistants can craft messages with very little guidance from, you know, real human beings. But now some real human beings worry that emails are reading a little too perfect. And as NPR's Chloe Veltman reports, that's a concern shared by some of those working in tech.
CHLOE VELTMAN, BYLINE: Email assistants have taken over our inboxes. The AI-powered tools can do things like compose messages in response to simple prompts and complete people's sentences. AI-assisted email is big business, too. The market for robo-missives is expected to grow to more than $4.5 billion by 2029. However, a recent Columbia University study reveals that more than half of spam emails are written by AI, and that's one reason why some tech workers are worried their AI-assisted emails will be treated as such for seeming too machine-like.
PETER SLEIMAN: I sometimes do little typos on purpose if I want to seem human.
VELTMAN: This is Peter Sleiman. He's the creative director of a video agency.
SLEIMAN: Forgetting a dot or maybe misspelling a word, maybe adding a smiley face the old way instead of actual emoji like ChatGPT does. I could add the two dots with a parenthesis.
VELTMAN: The topic of inserting deliberate errors into email came up during a session Sleiman co-led about digital marketing at the recent TechCrunch Disrupt conference in San Francisco. Fellow entrepreneur Luca Oake attended the session.
LUCA OAKE: One of my favorite T-shirts is a T-shirt that says I am secretly judging your grammar.
VELTMAN: Oake says she prefers that people stick to good grammar and spelling, plz.
OAKE: Because it actually can make it really hard to take in your message when I have to wade through, what did you mean here, because of your typos.
VELTMAN: Yet even Oake has her own method for humanizing email.
OAKE: I lowercase all instances of I.
VELTMAN: But beating the AI's drive for flawless sentences takes commitment.
OAKE: I'm autocorrected all the time.
VELTMAN: Grad student Daniel Zock was also at the conference checking out the latest innovations. He says people in his network are even asking AI assistants to write in less perfect ways in an effort to seem more human.
DANIEL ZOCK: I say, hey, ChatGPT, please avoid, like, the em dash, the en dash, all that stuff.
VELTMAN: But Zock says he's personally fine with AI composing slickly worded emails on his behalf.
ZOCK: What are we trying to do? We're trying to train these systems for reaching that level of humanness. So it's not bad if AI generates something that's very close to what a human will be outputting, right?
VELTMAN: Zock says people shouldn't get so hung up on style. It's the content of the message that counts.
Chloe Veltman, NPR News.
(SOUNDBITE OF GHOSTOWN DJS SONG, "MY BOO") Transcript provided by NPR, Copyright NPR.
NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.
