ryan_greenblatt's Shortform

ryan_greenblatt

This is a special post for quick takes by ryan_greenblatt. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.

People seem to use the terms 'slop' or 'AI sloppiness' to refer to a wide variety of different issues with pretty different causes, consequences, and solutions. It seems good to typically be more precise, at least in long-form discussion.

These all differ:

AIs have poor capabilties at conceptual / very hard-to-check tasks
AIs aren't trying on conceptual / very hard-to-check tasks but could do better
AIs "intentionally" produce outputs that are low-quality but appear good
AIs try to make their (low-quality) outputs look good
Many people specifically like AI outputs with a low-quality vibe so the internet fills up with this
AIs make it cheap to produce (mediocre) output so it will floods many places

2silentbob6mo

Can you elaborate? My impression is that at least Claude models struggle immensely to avoid their typical way of speaking (which I find annoying as hell), and I never managed to find a prompt that works to avoid that.

Just treat it as a combination of low standards and paranoia.

3Greg Price6mo

Here's yet another, quite different, meaning, which I heard used by a researcher at an AI company: * AIs produce output with recurring tics ("it's not A — it's B", "genuinely", etc.) that people notice and that start to grate on them Or perhaps more abstractly: * AIs produce output that many people learn to tell at a glance is from AI I don't like that meaning because it's not really about quality at all, or at most it's about one narrow aspect of the output's quality. But it seemed clear that those verbal tics are what the person had in mind when they explained they were working on a new "anti-slop" initiative. And it certainly fits that that's an aspect which an AI company might want to prioritize. Personally, when I say "slop" I normally mean something that overlaps with some of the meanings you list but is a bit distinct: * An AI produced something low-quality, and a human chose to post that without bothering to bring it up to reasonable quality. So it's about the human's behavior as much as the AI's performance. Partly that's because the contexts where I use the term tend to be something to the effect of "don't post slop here, it's a waste of our time and yours". (And because I know the term is ambiguous, I include a few words defining what I mean by it.)

ryan_greenblatt's Shortform

19

Ω 5

ryan_greenblatt's Shortform

19

Ω 5