,
10 tweets,
2 min read
Read on Twitter
Most of my papers contain very little mathematical notation. Is this because, unlike many AI researchers, I'm very uncomfortable with reading and writing mathematical notation? Or is it because, unlike many AI researchers, I know how to write readable English?
These two modes of writing and reading - mathematics and natural language - are so different that I wonder if being good at one of them is actively bad for being good at the other. That is, the two modes of thinking interfere with each other.
I'm very narratively oriented. I need rhythm and cadences when I write, and when I read. I have the feeling that people whose primary language is math instead see the notation as something like visual puzzles, that they work out as they process it.
Of course, I can't confirm this. I have to painstakingly work my way through mathematical notation when I encounter it, like l would work my way through a labyrinth. Often I just skip it and hope that the authors provide a legible English translation as well. Often, they don't
I wonder if this is to some extent dependent on culture, or on native language. My native language (Swedish) is quite similar to English, in particular grammatically. But maybe some languages predispose the learner to read mathematical notation? I could be very wrong about this.
The fact that papers by authors from some parts of the world are more math-heavy could of course simply reflect that fewer researchers from those parts of the world write good English.
One thing I'm pretty sure about, though, is that mathematics is not the universal language of science. It's just a historical accident that it's so widely used. In particular, mathematical notation is not better than natural languages for writing about AI.
If there is such a thing as a "true" language for writing about algorithms, it would be actual source code. In some commonly used language, such as c, Python or Java.
Finally, here's a poll: how do you prefer to read explanations of new algorithms? (In particular in AI/ML)
(of course, we will see a selection effect here, as people who answer the poll are those who made it to the end of the thread...)
