he/him
You probably heard about me because I am the founder of the Twisted python networking engine open source project. But I’m also the author and maintainer of several other smaller projects, a writer and public speaker about software and the things software affects (i.e.: everything), and a productivity nerd due to my ADHD. I also post a lot about politics; I’d personally prefer to be apolitical but unfortunately the global rising tide of revanchist fascism is kind of dangerous to ignore.
| posts | https://blog.glyph.im/ |
| disclosures | https://blog.glyph.im/pages/disclosures.html |
| code | https://github.com/glyph |
| patrons | https://www.patreon.com/creatorglyph |
Ah. So this is why I have seen several mentions of this over the last few days. Somebody posted my article to lobste.rs with an edited title, and there has been some Discourse I didn't notice.
I've now replied in the comments, but will update the post itself later, too: https://lobste.rs/s/rvgvgj/best_line_length_is_88#c_mhal5b
Thus, when an LLM absorbs some stolen data, what is happening cannot be 'learning'; it's something else. When we call it 'training', that's a metaphor, not a description. In reality, it is a parasitic activity that requires fresh non-LLM-generated information from humans in order to be sustainable.
A teacher “learning more from their students” is such a common observation that it is a cliché. Colleagues mutually learn from each other in professional settings. Actual artists are in conversation with one another, not just learning from a static historical canon. Etc, etc.
LLMs cannot do this. The output that an LLM produces contains a sort of poisonous residue that makes it destroy the reasoning capacity of other LLMs; this is a well-known problem in the field, known as "model collapse".
"LLMs learn the same way a person does, it's not plagiarism"
This is a popular self-justification in the art-plagiarist community. It's frustrating to read because it's philosophically incoherent but making the philosophical argument is annoyingly difficult, particularly if your interlocutor maintains a deliberate ignorance about the humanities (which you already know they do). But there is a simpler mechanical argument you can make instead: "learning" is inherently mutual.
Let's ask the real question:
Firefox users,
do you want any AI directly built into Firefox, or separated out into extensions?
| I want AI built into Firefox | |
| I want AI separated into extensions | |
| Mozilla should not focus on AI features at all |
I have in fact made all my code open source; I write updates about it, I try to stream myself coding once a week, and while I wouldn't say I get *zero* engagement, it's… closer to zero than any other number.
The thing about actual human beings is that they have shit to do. It's *very* hard to get their attention.
In a way, I'm lucky that LLMs just don't work for what I'm trying to do, because it feels like standing on a trap door that could be opened by glancing in the wrong direction.
Josh Simmons
Duke of Germany 💫
Robert Kingett
Anil Dash