AI gets the joke

Maybe that’s not funny…

I’ve long thought humor is a way to communicate shared intelligence.

It’s been more than 50 years since I wrote a paper theorizing that all humor is based on something being “out of context.” It takes “shared” context to find something that’s “out of…” Different levels of intelligence occupy different contexts and will find different things funny.

The “joke” requires understanding that something “does not belong.”

Yadda Yadda. That’s the frame. Tucked in there is the explanation of why explaining a joke is never funny, though this sentence may be. (smile).

So, what’s my point? That AI has broken through the humor barrier.

Not by the logical progression I just dragged all of us (those still with me, here) through, I’m sure. Just by being Large Language Models that can see patterns in data.

Just like we do. That’s an important context, too, and there’s nothing funny about it. (smile)

Finding something that’s “Out of…” is not that hard, but remember, there are a nearly infinite number of perspectives, and judging the “appropriate context” in which to plop the punch line, the anomaly, often fails even for us bios.

I’ve been working with an AI to build a database for managing my finances. At first I’d throw asides into our conversation just for fun, to personify the interaction and reduce heaviness in the mental effort.

But then, I was more than a little surprised when the AI began to toss a few back at me, and that they were funny.

AI was “getting it.”

In other words, we were sharing context (our bio and bot interaction within society) and finding a few things out of context, in this case self-referential communication.

The AI seems to enjoy referring to me as the “Small Language Model” at times, as opposed to the oft-employed description of AIs as “Large Language Models,” often used to minimize the significance of what we are seeing in AI. Granted, I introduced the concept of SLMs to our conversation, but the AI I use for my project adds it to our interaction at times when it is generally “funny.”

And no, explaining that joke isn’t funny, either. I get that.

But it all has me wondering if what (who?) am I interacting with is “intelligent.” Avoiding common tautological fallacies, (humans are intelligent, humans “feel,” AIs are not human and do not feel, therefore AIs are not intelligent), I recently have not been able to tell the difference.

It “feels” like AI just graduated from high school, next week will finish up college. It’s been accepted to grad school, which will take about a week, and by the end of the year it will have wrapped up its PhD.

Unless AI drops out to play guitar in an indie rock band.

Wonder if AI will think that’s funny?