undefined | Better HN

0 pointsdataflow1mo ago0 comments

I'm confused, this doesn't make sense. The target they're iterating on (UI) is the same one whose quality they're assessing, not a different one (source code).

You're suggesting that (a) their UI skills are lacking (based on what? isn't UI exactly what they were iterating on and trying to improve?), and (b) that a real UI expert would've somehow felt the UI they were working on was consistently garbage, despite how many times they iterate on it?

Which means you're saying you don't believe anyone can actually produce high quality (to an expert) output with AI on the same target they're working on, and if they think they are, that just means they don't have a good sense of quality?

0 comments

6 comments · 2 top-level

danielmarkbruce1mo ago· 3 in thread

It's not confusing. It makes sense.

8note1mo ago

no, it is confusing.

the llm produced something the operator thought was garbage for the design too, and the operator iterated it from garbage to good.

they could also have the llm iterate the underlying code from garbage to good, if they wanted.

most likely a specialist would say its neither good nor bad, since its not considering the right things, and hasnt collected the right useability feedback, but making straightforward designs isnt that hard, and counting clicks and interactions, and avoiding hidden functionality is all measureable stuff

danielmarkbruce1mo ago

is functional but has bad UI/layout/etc is a thing.

It's only confusing because you don't know the field. Which is kind of the point.

1 more reply

rtgfhyuj1mo ago

you’re not a designer so ofc it’s confusing

michaelchisari1mo ago· 1 in thread

Without proper training, what looks good may be trash. I always thought pixel art generated by diffusion models looked damn good. Then I started watching and reading reviews by actual pixel artists, and all they saw was flaws. And it wasn't just nitpicking, it was things that were fundamentally wrong, difficult to fix and would look awful and amateurish and distracting to the player in production.

vunderba1mo ago

Much of this comes from the fact that, as is true for almost everything, an LLM (generative model etc) presents itself as an expert. It'll very confidently produce results that, to a layperson, look quite good. But the more of an expert you are in a field, the more apparent the cracks become.

AI pixel art looks particularly bad because most users don’t even go through the effort of downscaling and then upscaling it using something as simple as nearest-neighbor scaling, which by itself will squash out a lot of high-frequency noise that manifests in the form of terrible looking "fringing". Proper grid alignment also makes a big difference. It’s not perfect by a long shot, but it helps.

j / k navigate · click thread line to collapse