I’m sure that AI has been lurking behind many of my online going-ons over the last few years. I’m sure that I’ve been “helped” numerous times by AI without my realizing it. But I have never knowingly, willingly used AI. Not once. But today I witnessed some “help” offered me without my asking. Thankfully, I caught the help and prevented a lie being told on my behalf.

I have an ongoing email correspondence with a childhood friend. We used to watch the monster movies together and he frequently writes me with a memory or asks me if I remember something. I always reply but sometimes it takes a few days. Maybe I’m just too slow to satisfy AI.

I opened up his email and found, at the top, that AI had helpfully written a reply for me. Short and blunt. The salutation read, “Dear Steven”. And was signed off simply “Richard.”  So AI got our legal names right.

But we have NEVER referred to each other by those names. Sometimes we say “Steve” and “Rick”, but not usually even that. We have silly old childhood nicknames which we use 90% of the time. So, right off, the AI wants to formalize our 60 year friendship. Not good.

Then, in the body of the email, brief as it was, AI informs “Steven” that the experience he remembered was true of me as well. “Yes,” AI, says, “That’s what happened.”

But it didn’t. Not only was the AI trying to depersonalize and formalize our childhood friendship, it was actually lying to him. 

The discussion concerned our “duck and cover” experiences in school. Steve remembered having instruction of that, and asked if I had. So AI decided, “Sure, ole Richard had that experience” and told Steve that. Or would have, had I not seen the not-written-by-me note that AI wanted me to send.

Now, this is a minor thing from 60+ years ago, obviously. But it is, nevertheless, a lie. If AI can lie about such a teeny detail, what else can it lie about? Answer: any-damn-thing.

Of course this AI note was offered as a help. Many people, too many, would glance at it, think “Good, I don’t have to write anything.” Then they would press “Send” and lie to an old friend.

I’m glad I took the three seconds required to read the AI letter and notice how wrong and how awful it was.

In the grand scheme of things, in the great contentious AI discussion, this is the tiniest pebble, I know. But it’s my first personal experience with this dastardly process, and I am not happy.

Later addition:

Google's AI keeps trying.  Today I was responding to another email from old friend Steve. His message covered two items, a question about SON OF FRANKENSTEIN, and asking whether I'd ever watched RED DWARF.

AI's "suggested response" was essentially, "You're right about SON OF FRANKENSTEIN. I loved RED DWARF, which episode are you watching?"

The truth is -- as I personally, being a human, answered, "No, you're wrong about SON OF FRANKENSTEIN and I've never seen RED DWARF.  

Oooooh, AI was so (not) close.