Is Plasmic building anything like this?
If you could use GPT/AI to automatically convert descriptions into editable Plasmic UI designs (that could then be converted into code via Plasmic) then this would be an absolute game changer
Yes and no, AI is not to be praised to the sky, it’s just a randomization and probability algorithm is all it is, try telling it twice the same thing, and see how the results are a thousand miles apart, I can’t imagine anything useful and consistent UI design coming out of AI
It’s super good for hyping things up for non-UI designers though
True UI designs need solid design methodology, componentization, color logic, UI logic, surface logic, userflow logic, interaction logic, token logic, and data logic. AI could give you a ton of ideas for those that don’t go to dribbble, but when it comes to precision and wholistic systematic approach, you’re simply not looking at the right tool
ChatGPT is just dealing with plain texts, and even then it’s probability of an accurate answer is somewhere between negative100%~75% averaging around 55%.
Using a roll of dice for building actual sensible UI —— No
Use it to suggest wild freeform design ideas for the design exploration phase —— Yes
You are so wrong you are due for a big reckoning once Figma or someone with a huge dataset of good designs does this.
All you need is data and training. Including courses on all the subjects you just mentioned for the AI to learn .
If a human can learn it, so can an AI at a much higher level.
Just a matter of time, and not much