Every Corpus Was a Body Once: Putting the [body] language into LLMs

Let's review and discuss LLMs and code generation through an internet linguistic lens. Not just examples of the cues that tip us off that someone has sent us their Chat response unedited and our thoughts about the etiquette of that (although this is welcome!); but consideration of changing internet language norms. Because language.

Corpus (noun)

• body in latin
• a collection of writings, conversations, speeches, etc., that people use to study and describe a language in linguisitics
• training or evaluation data in ai

“Language is humanity's most spectacular open source project.” ― Gretchen McCulloch, Because Internet: Understanding the New Rules of Language

The internet changed how we write. Not by replacing formal writing, but by giving informal writing greater visibility and international influence. In doing so, it changed language itself. New words, new conventions, new communities, new registers. AI is doing the same thing to code.

I've lived through a lot of these patterns before as a global studies major turned world language teacher turned software engineer; but you probably have as well. Drawing on McCulloch's linguistic research, my own survey of the local dev community, and 15 years of building software, we'll look at the shift we're living through and ask questions about who code is written for and what "readable" means.

In the spirit of keeping Minnebar weird, come ready to draw with crayons and experiment with how we communicate through (and without) AI. I'll bring the supplies.

• All levels
• Development
• AI/ML
• Community
• Other