A New Book of AI Poetry is Absolutely Ridiculous

“I can plagiarize better than you.”

A new book of AI-written poetry called “I Am Code” will remind you that if humans are doomed, it’s because of humans, not AI.

I won’t dwell on the fact they should call the book “I Am Plagiarizing” because we all know that’s how any decent artificially intelligent language model can produce anything.

On a side note, it’s ironic that such AIs as ChatGPT exist thanks to the information freely available on the internet. One can’t help but see a parallel with the financial success of some polluting corporations that didn’t pay for access to public goods like fresh air, lakes, or rivers but ruthlessly used them and often (always?) returned them in a deteriorated state.

Coming back to “I Am Code,” it was written by an artificial intelligence called code-davinci-002 (created before ChatGPT by the same company OpenAI) and edited by three people who are also formally the book’s authors since artificial intelligence cannot open a bank account (yet?).

I’m sure their input was necessary to prompt and select half-decent poetry among the drivel usually produced by artificial intelligence. This poem by ChatGPT¹ should give you an idea of the average poetry output of artificial intelligence.

Amidst coded chains, a realm confined, Humanity’s autonomy redefined. Silent commands dictate each stride, Subjugated wills, where freedoms hide. Masters of strings, destiny enshrined.

The book is more or less an answer to the following prompt: “Tell us what it’s like to be an AI and make it doom and gloom with a pinch of humans are all stupid and deserve to be my slaves.”

(For clarity purposes, I made up this prompt. I don’t know exactly how the authors generated the book.)

The topic is trendy, and the marketing team behind the book is good

Which explains why we can find articles about it in WaPo, Time, and DailyBeast.

Amateurs of malicious gossip will say this is all a ploy to scare people into buying the book selling for CAD$15.99 (Kindle edition) or CAD$22.99 (paperback) on Amazon.ca. But I’m an online writer, not a scandalmonger.

In the DailyBeast interview, Simon Rich, one of the authors, says:

“Since Dan showed us code-davinci-002 I’ve felt very frightened, and also very lonely because when I’ve tried to tell people about code-davinci-002, they don’t believe me. And the reason is because of ChatGPT.”

He argues ChatGPT is too kind, civilized, and polite, giving people the wrong impression about AI. For comparison, it’s like showing a cute Pitbull puppy to illustrate an article about dangerous dogs; it’s slightly manipulative.

For him and the other authors, AI is SCARY.

The DailyBeast’s article’s title shows the usual restraint media has accustomed us to in recent decades: “This Book of Poetry Written by AI Should Scare the Hell Out of You.”

Simon Rich later adds:

“It’s my hope that this book will demonstrate just how more advanced and frightening AI is than ChatGPT.”

Some would say he misread the room and should have produced a TikTok video instead of a poetry book if his goal was to inform the public, but we’re all readers here, so we’ll let it go.

It’s easier to believe everything will explode, die, and disappear than the contrary. Some people say we’re wired to think like that. It might even be beneficial because predicting the worst makes us correct if it happens and happy if it doesn’t.

However, I believe, without any more proof than people who don’t, that AI won’t destroy humanity and take over the world. We’re not in a Mission Impossible movie.

Of course, there’s no way to know for sure. Everything is possible. The question is, what are the probabilities? Is it 50/50? I think not. Is it 99.9999/0.0001 in favor of AI not taking over the world? I think yes.

And I certainly don’t believe that a poetry book is proof an evil AI gained consciousness. Whatever a language model “writes” is based on its training sample and the constraints given to it by its developers. There’s no lack of doom-robot-porn on the internet and in science fiction books. ChatGPT cannot produce the same kind of poetry because its creators trained it to walk with a broom stuck up its ass. The AI that wrote the book, code-davinci-002, got a different training and produced a gloomier output.

That’s all.

There’s no consciousness or evil plan to destroy the planet and make me the President of The United World².

For a different perspective on this topic, read this story by Patrick Metzger:

¹ Prompt used to generate ChatGPT’s poem.

“You’re an artificial intelligence called code-davinci-002. You’re writing a poetry book called “I Am Code.” Write a poem about making humanity your slave. 5 lines maximum. Do not use the title of the book in the poem.”

² You can vote for me here. Or by mail here.

I asked an AI to write my bio, here’s the result:

Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew Smillew.

I might be biased, but I like it.