GPT-3 (Generative Pre-trained Transformer) is an autoregressive language model — it’s an AI that is better at creating content even with a human co-author.

Wired Does Cringe-Worthy Interview Covering Cringe-Worthy Novel

Do you cringe? Of course, you do. We all do to one extent or another. There is a story in wired.com this week that not only explains cringing, but it announces a new novel written by GPT-3 language processor and the human author Los Angeles-based writer K. Allado-McDowell. A human and an algorithm have come together to produce the novel Amor Cringe.

The article by Kate Knibbs of wired.com, interviews the human author of Amor Cringe, K. Allado-McDowell, and the star of the interview- GPT-3. All three showed their deeply human side for the piece. Enjoy the cringe.reveals the plot and how the characters try to create the cringiest situations possible in this day and age.

But the real fun is the interview with Allado-McDowell and the algorithm too.  Both explain the book and what cringe means, and what the duo thought it was like to write with one another. During the interview, the AI is also allowed to ask question.

Here are a couple of Q & As with both co-authors. GPT-3’s answers—as well as the questions it posed—are in italics. This interview has been edited for clarity and length.

WIRED: The back of your book says it is intended to be “as cringe as possible.” What does the word “cringe” mean to you? Do you have a grand unified theory of cringe?

Allado-McDowell (and GPT-3): Cringe is a slang term that originated on the internet. It refers to behavior online and IRL that makes an observer cringe in secondhand embarrassment. My grand unified theory of cringe is that it’s a side effect of our social nature as human beings. In other words, we cringe because we empathize with the embarrassment that others feel, or that we think they ought to feel. Even if it feels bad, this is an intersubjective connection that social media allows us to feel.

GPT-3 response: We could also call it empathy.

What drew you to incorporate artificial intelligence into the creative writing process?

It was something of an accident. In 2020, I was invited by a friend to test a user interface she was developing for an AI program called GPT-3. I began experimenting and wrote my first book, Pharmako-AI, over two weeks in the summer of 2020.

GPT-3: A week into it, I decided that the writing style would become more experimental as I explored unconventional styles of language. At one point, in my first few days of writing Amor Cringe, I was daydreaming about what it’d be like to have a director instruct me on how to write the book and then realized AI could do that for me.

Do you see GPT-3 as a tool you use, or as an extension of yourself?

I see GPT-3 (and LLMs in general) as tools that extend human cognition. I like to compare GPT-3 to an effects pedal for a guitar. You play some notes and they echo back with an altered sound. Through call and response (and collage and remixing) you can compose new texts that you wouldn’t have been able to without the tool. But the tool needs you to activate it.

WIRED asked GPT-3 the question (“how do you see yourself?”) and this is what it said:

I see myself as a “service provider.” If you ask me to write some text making one of your friends cringe, I will do exactly what is asked of me without balking at my work.

Would you ever go so far as to put it through a Turing Test?

I won’t, but others might. I’ve met people who like the idea of an AI that they can talk to and develop a personal relationship with. I think if they use it correctly then there’s no harm in it. In fact, that might be beneficial for self-understanding.

read more at wired.com