Useful Poetry

Ease the pain in your strained brain

I believe we should bring back poetry.
Not in the rosy way that most poems are written in. But rather as a superior tool for communicating amongst ourselves.

Modern language is mostly judged by whether it is useful.

How clear can you write?
How concise are your explanations?
Are you good at creating visuals in your reader’s brain?

When people in industry or academia think of “good” language, they think of polished software documentation or professional-sounding conference papers.

Thesis titles like “Direction Symmetry of Wavefield Modulation by Tidal Current” or “A negative feedback loop between TERMINAL FLOWER1 and LEAFY protects inflorescence indeterminacy” are commonplace.
Indeed, everything that does not immediately confuse others can be deemed “unsophisticated” in academic circles, whereas industry is quick to denounce informal language as “unprofessional”.

Here are a couple of the highest-cited Machine Learning papers of the last 11 years:

• Deep Residual Learning for Image Recognition - He et al. (2015)

• Attention Is All You Need - Vaswani et al. (2017)

• Learning Transferable Visual Models From Natural Language Supervision - Radford et al. (2021)

Now, which of these titles could you recite if I asked you about them tomorrow morning?

Only “Attention Is All You Need” sticks. Why? Because we are human!
Our brains like simple language. They like language they can grasp easily and which they can fill with meaning. They like language that evokes associations beyond the characters on the page.

What does “Attention Is All You Need” make think you of?

The Charlie Puth song?
The person you met at the last party?
Or maybe your attention-adoring Golden Retriever?

No matter what associations popped up, there were certainly more of them than if you thought about “Direction Symmetry of Wavefield Modulation”.

Now let’s try something else: If you have read the “Attention Is All You Need” paper, then you know, that it is about a Machine Learning architecture — the Transformer —that allows parallel training in contexts in which it was not possible before. This makes training much faster.

If you wanted to remember that, you probably could.
However, If you are no Computer Scientist and don’t know the context, then you’ll likely forget the information at some point.

What if we write it down the following way:

“Attention that is all we need.
It gives our model training speed.
And why, you ask, it works so well?
The answer: It runs parallel.
The world’s computers now run warmer,
and all is due to the Transformer.”

I don’t claim to be a great poet, but in my humble opinion, the above is infinitely more memorable (and concise!) than the vast majority of things that people will ever tell you about the paper.

Again:

Poems are concise and memorable. That is exactly what we want from a lot of texts that we write in our every day lives. Additionally, the process of creating poems forces the writer (you!) to think about a topic in a different way than you usually would.
This is makes writing poetry great tool for learning!

That’s all I got. Maybe it motivates some people to write more poetry.
Apart from being a great form of text, it also makes the process of writing (and hopefully reading) a lot more enjoyable.

Let’s finish with another poem:

If you you think about it you’ll
soon realize that as a rule
when writing down some stuff that’s cool
poetry is a useful tool.