For quite a while, someone has been living inside my computer, writing emails for me.

I don't recall signing up for this artificial intelligence feature, which is like having a word valet. It's in my phone, too, which offers three serviceable but impersonal responses I can fire off to someone who has just sent me an email pitching a story or asking if I want to meet for coffee.

"I'd like to do coffee," was one of the suggested responses to a recent email. "Let me circle back soon about timing."

One argument for these features is that they can save time and free me up for more important tasks. But it takes longer for me to read the three fabricated email options than it would take to write my own response.

I find this really irritating for about 150 reasons, one of which is that in an ever-automated world, it's another nail in the coffin of human interaction. And yes, there are at least 150 reasons. I know because I asked AI and it spit them out in approximately three seconds. No. 148: "It sounds like it's written by a committee."

A fair share of nasty feedback lands in my mailbox, so I wondered if the auto-response tool could come in handy. But the robot isn't salty enough to be of service. "Thanks for reading" was the suggested reply to someone who called me a hopeless loon and another guy who wondered why anybody would read my "dumb column."

On second thought, maybe the unruffled dismissive response is the way to go. But the bigger concern is what happens to human intelligence as artificial intelligence does more of our writing, researching, communicating and thinking.

My son, a college librarian, has seen that phenomenon as well as a general erosion of research skills and decision-making aptitude among some students. "They can't choose a book from among thousands of books for a research project and don't even want to because they think they can get the information more easily from a computer," he said.

A Cornell University study released this month suggests that AI writing assistants can influence not only how we write but how we think.

Researchers observed 2,500 participants who wrote on several controversial topics including the death penalty, fracking, and voting rights. Some were provided biased information through AI autocomplete writing tools, and based on surveys before and after the exercise, their views shifted in the direction of the bias even if they were made aware of the bias.

"We know these models are controlled by large and powerful organizations, and they may or may not have a viewpoint they want to embody or promote, and there's potential for abuse," said Mor Naaman, professor of information science at Cornell Tech and senior author of the study.

The information spat at us is "wrapped in convincing AI language," Naaman said, and the advantages of the technology are evident. "The bad news is that there are literally hundreds of billions of dollars of investments and interest in trying to push AI into every corner of our lives ... and the dangers are being brushed aside."

It's going to take more time, Naaman said, to expose all of the risks and know how to rein them in.

AI will create jobs, for sure. It will also eliminate jobs, and it might be coming for mine. So I asked AI for an ending to this column, and here's what it came up with:

"And that's the central tension of this world: the promise of efficiency versus the irreplaceable process of being human."

I think my job is safe--for now.

© Arkansas Online