AI AIn't Funny

This fall I dropped my son off at Kennesaw State University - Hooty Hoo! He graduated High School this past spring, but didn't walk the stage. He had been dual enrolled for a year, and had not been a part of high school campus life, so our family watched the proceedings together from the comfort of home.

Artificial Intelligence Chatbots had just sprung onto the map with ChatGPT's launch, and I was definitely interested in the technology after a few interactions. As we watched the graduation and valedictorian address I wondered aloud to the family how many such speeches would be generated by AI in the spring graduations. I made no individual calls at that point, but, along with the rest of my family, well, I'll just say the percentage estimation (with no scientific backup) was high.

So I started experimenting, using AI to create Valedictorian speeches. I asked AI to make the speeches funny, or heartfelt, etc... There was standard fare in the speeches generated, but, when I asked it to be funny, it was mostly corny "Dad Jokes" of the worst kind...not bad enough to be good, and not good enough to be funny. Very predictable.

I'm also a huge fan of "The Office", and any time I think about speeches, my mind races to an episode that featured a speech that Dwight Schrute gave at a sales convention. His co-worker, Jim, convinced him to give the speech in the style of an authoritarian dictator. I couldn't resist prompting the AI along similar lines of thinking. So I asked it to generate a valedictorian speech as it would be given by an authoritarian dictator.

Now the content generated out of that prompt was, well, the results of that prompt had our whole family rolling on the floor. Here is an excerpt:

"Throughout our tumultuous years in this institution, we, the chosen few, have weathered countless trials and tribulations. Our journey has been rife with challenges, each one serving as a test of our mettle and unwavering allegiance. In this hallowed hall of knowledge, we have been groomed, disciplined, and molded into obedient subjects, ready to serve the purposes for which we have been destined."

My son then wondered about Salutatorian speeches. So I prompted Chat GPT to generate a speech for the Salutatorian. Only I asked it to make the style even more autocratic, and a tinge jealous. The results left us breathing hard...tears in eyes. Here is an excerpt:

"In this autocratic regime of academic hierarchy, success is a cruel mistress, favoring the unworthy while casting aside the deserving. We must toil incessantly, not for personal fulfillment, but to prove ourselves to a world that will always view us as second best. Our efforts will forever be overshadowed, our achievements forever belittled, but we must press on, fueled by our bitter jealousy."

In my opinion these speeches were way...way better than the corny "Dad Joke" speeches.

It brings the point of this article into high relief. For the moment, as far as I can tell, AI still needs context and human guidance (desire) to be effective. It can produce content like a fiend. It can be used to generate questions about any topic by brute force. It can create images with no copyright protection (see the picture that headlines this article). However, at least for now, it still needs context and guidance to be useful. First efforts require knowledgeable human "tweaking" to make a good product. Just take a close look at the faces in the crowd of the picture for this article. I didn't tweak the image because it fools the eye from a distance, but, upon close inspection, it really needs some work.

Why It Matters

To get the best performance, the guidance (at least at the time) was to instruct it with a prompt to pretend. In a separate instance, on a professional related chat, I asked it to pretend, and that was exactly what it did. I asked it where the information used to guide its interaction came from, and it sited 2 sources...one of which turned out to be completely fabricated. When asked why I should trust it, it said that I should trust it because it had 10 years of experience. HHHHMMMM. It was only doing what I asked it to do…and some of its reasoning was plausible, but not accurate.

It is a tool that verges on being a partner - but there are important differences. It is my understanding, confirmed from use of the tools available, that AI, like other tools, amplifies human potential - but does not replace it. In the story above I riffed off of a very funny scene from the Office, written by human writers. I had to work with the AI (me for initial and creative direction as well as refinement of the final product, it for content generation) to create a funny speech.

For quality as well as direction, AI still needs a knowledgeable human to interact with it throughout the process. Content will suffer if AI is left to develop said content on its own.

AI is quickly evolving. There is real concern, and it is not unfounded. I’ve only scratched the surface of what one chat bot can do. There are people who spend all their time exploring what AI can do…I’m not one of them. But from my limited use, I don’t think we have to worry too much that AI will take over in its present form.

I will worry when it can make me laugh as hard as the writers from the Office by itself.

Until that happens, AI Ain’t funny – at least on its own.