The Large Language Model Sea Shanty

Like everyone else, I’m playing with GPT3 chat. I asked it to generate a sea shanty that explains how large language models work, in the style of Stan Rogers.

Prompt: Explain large language models in the style of a Stan Rogers sea-shanty.

Completion:

Verse 1:

Hark ye sailors, come listen to me

I’ll sing a tale of a machine so grand

A large language model, ye see

To predict the words that we’ll command

Chorus:

Oh, the large language model

Truly a marvel to behold

With its deep neural nets

It never forgets

A treasure trove of knowledge untold

Verse 2:

Trained on vast amounts of data

It can understand any text we feed

From simple sentences to complex queries

It can provide answers with impressive speed

Chorus:

Oh, the large language model

Truly a marvel to behold

With its deep neural nets

It never forgets

A treasure trove of knowledge untold

Verse 3:

With its ability to generate words

It can create stories never heard before

From news articles to poems and more

It’s a tool that we’ll use and explore

Chorus:

Oh, the large language model

Truly a marvel to behold

With its deep neural nets

It never forgets

A treasure trove of knowledge untold.

Prompt:Compose the melody for the shanty. Write in ABC notation.

Completion:

X:1
T:Large Language Model
C:Traditional
M:6/8
L:1/8
K:G
D | “G”GGG “C”GEG | “D”FDF “G”DGD | “G”GGG “C”GEG | “D”FDF “G”G2 D |
GGG GEG | FDF DGD | GGG GEG | “D”FDF “G”G2 D |
“G”GGG “C”GEG | “D”FDF “G”DGD | “G”GGG “C”GEG | “D”FDF “G”G2 D |
GGG GEG | FDF DGD | GGG GEG | “D”FDF “G”G2 D |
“G”GGG “C”GEG | “D”FDF “G”DGD | “G”GGG “C”GEG | “D”FDF “G”G2 D |
GGG GEG | FDF DGD | GGG GEG | “D”FDF “G”G2 D ||
D | “G”GGG “C”GEG | “D”FDF “G”DGD | “G”GGG “C”GEG | “D”FDF “G”G2 D |
GGG GEG | FDF DGD | GGG GEG | “D”FDF “G”G2 D |
“G”GGG “C”GEG | “D”FDF “G”DGD | “G”GGG “C”GEG | “D”FDF “G”G2 D |
GGG GEG | FDF DGD | GGG GEG | “D”FDF “G”G2 D |
“G”GGG “C”GEG | “D”FDF “G”DGD | “G”GGG “C”GEG | “D”FDF “G”G2 D |
GGG GEG | FDF DGD | GGG GEG | “D”FDF “G”G2 D ||

So of course I turned it into a mp3; listen here.