[RMWildly]

The human in the mix is definitely the interesting part here. UI design inevitably has a more subjective element than other tasks, so I suppose it would be tricky to build a solution like this because, even if what it outputs is objectively a good UX, the end user will base their judgements on their own design preferences, so it'd be hard to satisfy everyone enough to get something working that's commercially viable.

I wonder is there also a factor that - since there are so many low/no-code and WYSIWYG UI tools on the market already - the potential time savings are lower compared to back-end tasks that are being targeted by AI tools - ie providing language prompts to get a tool to output UIs and then needing to tweak them to implement your own branding/messaging/design preferences might not be that much faster than building the same UI from scratch with the kinds of solutions already on the market.

Very interesting to see how this space could play out over the coming years anyway.