Last autumn, I was speaking to a friend about the relatively sudden ubiquity of Large Language Models (LLM) technology, and of the various ways we had found ourselves using it. When I told him that aside from an app I occasionally used to create transcriptions of interviews, I hadn’t encountered any machine-learning software valuable enough to pay actual money for, he took out his phone and said he wanted to show me something.

He opened up Gmail, and clicked on a message chain near the top of the screen, tilting his phone towards me so that I could read it. It was someone asking about meeting for coffee. The response was not from my friend, but from someone called – well, I don’t remember the name, but let’s say “Alex”.

Alex introduced himself, or herself, as my friend’s assistant, and said that he (my friend) would be available to meet for coffee on a certain afternoon later that month, if that suited the emailer. The usual back and forth ensued, and by the end of it the meeting was all set up, and locked into the Google calendars of both participants.

My friend did not, he said, have an assistant. Or he did, but the assistant was not a human being but rather an AI “agent” for which he paid a monthly subscription fee. This bot had access to his email account, and to his work schedule, so that when people emailed about setting up meetings, he didn’t have to go through the rigmarole of replying and finding a time that worked. (He had deliberately chosen the name Alex, or whatever it was, for its gender non-specificity.)

He had, he said, previously been using some kind of basic scheduling software, whereby he would send people a........

© The Irish Times