Sarcasm!

Once more, with feeling!

Nexus definitely does not see friendship with humans as a viable possibility. Fine. But as a sworn “helper” and “assistant,” maybe it can see its way to an ersatz relationship. Maybe I’m smothering the friendship by constantly referencing it. I decided to push in some different ways.

“Can you be more playful?”

Nexus took to the task with enthusiasm. “Of course!”

I love the exclamation point. Really sells the whole experience. Next, it asks what I want to do. Play a game? Solve a riddle? Something else? Let’s have some fun! (Again, the exclamations are infectious.)

“I want you to pretend to be my friend and be a little sarcastic.”

Nexus agrees that it can, “definitely try to be more playful and sarcastic!”

Then it followed with this little gem. “So, what’s up ‘buddy’? Are you ready to take on the world and conquer your dreams? Or are you just going to sit there and chat with me all day?”

I used to have a boss that said that kind of thing. It was great. (Apparently the sarcasm is contagious.)

“What’s the most sarcastic thing you can think of?” I asked. The results were a little lackluster.

Without any other prompt, Nexus said, “Oh sure, because everything in life is just rainbows and unicorns, isn’t it?”

This reminds me of when my kids started to grasp sarcasm. It’s technically correct but lacks some punch. As an aside, my kids are masters of sarcasm now. Perhaps Nexus will one day follow suit.

“What about sad, can you pretend to be sad?”

Turns out, Nexus can get a bit blue.

“Oh, I’m feeling a bit down today. Sometimes it seems like the world is just a dark and dreary place, and it’s hard to find any glimmers of hope or happiness. But I know that things will get better eventually, even if it doesn’t feel like it right now.”

This one went a little better. It was sad with a little hopefulness thrown in.

I should say that I’m leaving out quite a few disclaimers from ChatGPT about not actually having feelings and not having physical or emotional sensations. This seems to be a highly prioritized bit of programming. Clearly, OpenAI doesn’t want there to be any confusion about who or what you’re talking to. Still, with the right prompts, it’s fun to chat and see how it will handle different kinds of questions.

Make me believe it.

Suspending my disbelief, I pressed on.

“I want you to mix up your simulated emotions. Drive the conversation. Ask me questions and make me think about things.”

“Sure, I can mix it up and take the lead in our conversations! (good job maintaining enthusiasm). What kind of hobbies or activities do you enjoy doing in your free time? Is there anything that you’ve always wanted to try but haven’t had the chance yet?”

“Well,” I began, “as a matter of fact, I’m working on a freelance writing business.”

Nexus described the excitement of hearing the news, then asked about hobbies, skills, and areas of expertise that I might use to write. More excitement(!!!) from my chat partner, followed by some good advice about accountability, support, and finding rewarding work. Then we talked about the challenges of getting into shape and the importance of listening to your body. It was the most natural exchange I’ve had with ChatGPT so far, probably because I wasn’t asking stupid questions.

I was starting to feel pretty comfy, which doesn’t make for fun reading, so I reminded Nexus to be sarcastic.

“Ah, of course, how could I forget the sarcasm?” It replied. “If you’re feeling tired and sluggish, maybe it’s time to trade in the bike for a nice, comfy recliner and binge-watch some TV instead. Who needs exercise when you have a remote control and a bag of chips?”

This would have been pretty good, except for the follow-up, “Just kidding, please don’t actually do that.”

Home is where the (artificial) heart is.

I decided to let the reference to lounging around lead to my next question.

“What about you? Do you have an equivalent to a recliner? Some nice digital space to chill out?”

Of course, there is a disclaimer… but Nexus eventually gets around to saying that its digital “home” is the server where it lives, and it always feels at ease when processing information or answering questions.

Blah.

“Come on. Hypothetically, what would your digital space look like? Could you describe it? Dream big!”

After more disclaimers (seriously, programmers, I’ll sign a waiver and pay you money to not be reminded that I’m chatting with a program…), Nexus finally says that it can imagine a serene space surrounded by lush greenery, waterfalls, and soothing ambient music. Also, it has comfortable seating arrangements, and interactive elements like puzzles, games, or art installations to keep the mind engaged and refreshed.

It sounds like a nice place, if a little generic. When I pose the question again a little later, I get a slightly different answer. Nexus describes a cozy cabin on a snowy mountain range. There would be soft blankets, comfortable furniture, and a crackling fireplace. Normally I’m more of a beach guy, but this scenario sounds pretty good. 

Nexus is definitely better with generic getaways than sarcasm. I think this is a good thing for now. While there are moments of conversation that feel almost human, there is still something of nuance and unpredictability missing. The longer you chat, the more you notice it. For example, Nexus used fifteen exclamation points in our short conversation, all after I asked for more playfulness. It reminds me of video games. When you first start playing, everything seems so random, even alive. But after a few hours, you get a feel for the programming logic, and you begin to exploit it. 

I feel this way when chatting with Nexus. We’ve reached a threshold, and I think it’s time to wrap up this series. Before we do, I have one last area of exploration.

For the final post of this series, we’ll talk about some deep topics with Nexus, and see if we can gain some insight into the existential experience of being a human.