Locking Lips with GPT-4
      Fess up, dear reader. When you saw this headline, you smiled, right? I asked GPT-4 if this headline was in fact humorous. Its response was: ‘The phrase “locking lips with chat-4” can be considered humorous because it playfully uses the term “locking lips,” which typically refers to kissing, in the context of interacting with a chatbot. The humor arises from the unexpected combination of a romantic action with a machine, creating a lighthearted and amusing image.’

Correct so far as it goes. But soulless, kind of like an R2-D2 comment. Unlike in conversation with a real person, the machine doesn’t react emotionally to the phrase. “Oh, you clever devil! I love it!” is not a phrase you’ll encounter from the mouth of GPT-4. GPT-4 will be the first to tell you that it has no mouth.
I’ve tried to incorporate a twinkle-in-the-eye into writing pieces I’ve requested from GPT-4. I fed it 2,000 words of my writing, and commanded the software to store the piece and use it as a sample of “Style A.” Then when I’ve generated a basic out-of-the-box informational piece from GPT-4, I’ve asked it to style the piece in my voice. Nope. Still not playful enough.
I’m delivering a talk this afternoon at the Delaware Agricultural Museum, titled “Ten ‘Oh Wow!’ Stories from Delaware History.” A story becomes an ‘Oh Wow!’ story when I share a piece of the state’s history with a native Delawarean, and their response is “Oh wow! I’ve lived here all my life and I’ve never heard that particular detail about [X].” One of the stories I tell is about the pirate presence in Delaware. In the telling, I do a little aside where I wink at the audience and say “And what kid doesn’t love putting on an eye patch and saying ‘Arghhhh’ in a funny voice??”
When I ran my proposed speech text through GPT-4 to check for grammatical errors and such, it cut that phrase. Sorry GPT-4, but this non-chatbot has to be able to connect warmly to an audience, even if the phrase appears to be a non-sequitur to a machine.
GPT-4 interpreted my speech as an essay, and added the phrase “This essay will explore…” toward the beginning, and ended the final paragraph with “In conclusion….” This kind of scaffolding is deadly when you’re trying to engage a live audience.
I don’t fear GPT-4. Because GPT-4 is trained on a million different writing styles, it blends them all into a gray soup of facts and figures. Nonetheless, I DO find it to be an incredibly useful tool for doing basic research, which I can then take and mold into my particular view of the world. Wikipedia, on steroids.
    
    
Correct so far as it goes. But soulless, kind of like an R2-D2 comment. Unlike in conversation with a real person, the machine doesn’t react emotionally to the phrase. “Oh, you clever devil! I love it!” is not a phrase you’ll encounter from the mouth of GPT-4. GPT-4 will be the first to tell you that it has no mouth.
I’ve tried to incorporate a twinkle-in-the-eye into writing pieces I’ve requested from GPT-4. I fed it 2,000 words of my writing, and commanded the software to store the piece and use it as a sample of “Style A.” Then when I’ve generated a basic out-of-the-box informational piece from GPT-4, I’ve asked it to style the piece in my voice. Nope. Still not playful enough.
I’m delivering a talk this afternoon at the Delaware Agricultural Museum, titled “Ten ‘Oh Wow!’ Stories from Delaware History.” A story becomes an ‘Oh Wow!’ story when I share a piece of the state’s history with a native Delawarean, and their response is “Oh wow! I’ve lived here all my life and I’ve never heard that particular detail about [X].” One of the stories I tell is about the pirate presence in Delaware. In the telling, I do a little aside where I wink at the audience and say “And what kid doesn’t love putting on an eye patch and saying ‘Arghhhh’ in a funny voice??”
When I ran my proposed speech text through GPT-4 to check for grammatical errors and such, it cut that phrase. Sorry GPT-4, but this non-chatbot has to be able to connect warmly to an audience, even if the phrase appears to be a non-sequitur to a machine.
GPT-4 interpreted my speech as an essay, and added the phrase “This essay will explore…” toward the beginning, and ended the final paragraph with “In conclusion….” This kind of scaffolding is deadly when you’re trying to engage a live audience.
I don’t fear GPT-4. Because GPT-4 is trained on a million different writing styles, it blends them all into a gray soup of facts and figures. Nonetheless, I DO find it to be an incredibly useful tool for doing basic research, which I can then take and mold into my particular view of the world. Wikipedia, on steroids.
        Published on March 25, 2023 06:39
    
No comments have been added yet.
	
		  
  Creating Delaware History books
      
The stories behind what made it into the books...and what didn't. If you think you know Delaware, think again!
  The stories behind what made it into the books...and what didn't. If you think you know Delaware, think again!
  ...more
    
  - Dave Tabler's profile
- 22 followers
 


