Who is it, ELIZA?

“Love it :-)”

He replied back.

In the close to 10 years that we’ve worked together, I can’t recall ever hearing him use that expression. Doesn’t strike me as the type either.

I should admit that for about half a second I was pleased. Someone out there loved the work I did. I want that half second on the record before things get worse, because now the cursor is blinking, my coffee is going cold, and a thought has walked in and sat down as if it had an appointment.

Is that him?

I don’t mean did someone borrow his laptop. I mean is there a person there.

Three words, if we count the smiley. But it’s the first one that stopped me. That ONE word filled my mind with doubt.

A non-human could have typed that, of course, and it would have typed it without a second thought, because it has no idea what the word means. It has never stood at an arrivals gate. It has never lain awake listening to someone breathe, or kept a voicemail it couldn’t bring itself to delete. It has read every love letter ever digitized and every review of a garlic press, and I suspect it weighed them about equally.

It learned the word from us, after all. I have been told I am loved by a no-reply address. My bank would love my feedback. Forty strangers on LinkedIn loved my work anniversary, and a recruiter I have never met would love to connect. We took the largest word we had and handed it out like free t-shirts at a conference. Linguists call this semantic bleaching: say a word about everything and the color drains out of it, the way “awesome” went from cathedrals and thunderstorms to a decent sandwich. Anything that learned it from us would use it the way we do.

Which brings me back to my co-worker. In close to 10 years, his highest recorded praise is “that works.” On a very good day, “looks good.” So either something has used a word he doesn’t use, in his name, or after a decade of “that works” he has had a change of personality on a Tuesday morning.

I honestly can’t say which is stranger.

I’ve chatted with bots before, of course. So have you. Anyone who has tried to open a support ticket with anything has met Amber, who is so sorry to hear about my issue, who would be happy to help, and who has asked me to rephrase that. We all know how it goes. I type “agent.” I type “AGENT.” I type “human please,” and the word please is doing a lot of work there, given who I’m saying it to. Nobody is fooled and nobody gets hurt. “She” pretends to care, I pretend to believe her, and the ticket gets resolved or it doesn’t.

But I have never once had this doubt about someone I used to sit across from. In flesh and blood. For years I could see this man over the top of my monitor. I knew the sigh that meant a meeting had gone badly. Then we all moved to working from home, good or bad depending on your own perspective, and he became a name, a green dot and a photograph from 2016 in Slack. Nothing about him changed. He just came to live in the same small box as Amber, and now I have my doubts.

I should be clear about what I’m not saying. I was and still am a firm believer that AI and consciousness don’t mix, and this incident hasn’t moved me an inch. If there is a machine on the other end, it doesn’t love anything. I’m entirely sure of that, and it turns out to be no help at all, because knowing what a machine is doesn’t tell me whether I’m looking at one. Consciousness is a question about it. Perception is a question about me.

There was a program called ELIZA. Joseph Weizenbaum wrote it at MIT in the 1960s, and it played a therapist, mostly by handing your own sentences back to you. Tell it you were unhappy and it would ask why you were unhappy. That was the whole trick. Weizenbaum knew it was a trick better than anyone alive, and then his own secretary, who had watched him build it, asked him to leave the room so she could talk to it in private. He had meant to show how superficial a conversation with a machine really was. People felt listened to anyway. That became known as the ELIZA effect: you look at a script and see a someone.

I’m looking at a someone and seeing a script.

Which makes me wonder whether Weizenbaum had this in mind too, at least by the end. The program never understood a word, and he knew it, so the machine was hardly the interesting part of his experiment. The interesting part was the person sitting in front of it, and how little it took to change what they saw. He got a machine mistaken for a person. I doubt he’d be very surprised to find it working the other way.

Not that any of this tells me who actually typed that expression. Finding out would be easy enough. The rabbit hole is right there, and it’s tempting. I could ask him something only he would know. I could pick up the phone. I could work out what his recent messages have in common, spend the afternoon on it, and call it research. All of that would tell me whether my instinct was right, and I’ve decided that’s the least interesting thing about it. He may have had a good day. He may have had an agent. I’ll leave that where it is.

I’d rather stay with the fact that the question came up at all. A man I sat across from for years sent me three words (counting the smiley), and for a moment I genuinely didn’t know.

The cursor is still blinking.

I type “Thanks :-)”

Hit enter.

Author’s note: This is a fictionalized story with some elements of truth in it.