How an AI native uses AI in his workflow

With the current euphoria around ChatGPT, many forget that Conversational UX is not the holy grail of computing soft design. We will always need custom Interfaces for better control.

This article showcases a brilliant series of applications for Information-intensive activities (research...) :

1- Search the Web / KBs: It is not either Google or ChatGPT [4]. There is a lot of potential in the space between, especially for particular domains like academia KBs (Elicit [0]),or personal KBs/data (revery [3])

2. Assist Reading / Understanding : this includes highlighting/auto-linking Text (to rapidly scan it in context [1]) and summarizing/explaining [2]

OP also shows the need for AI text tools to give us more control over the output (the way AUTOMATIC1111 does it with Stable diffusion). He gives 2 examples for text rephrasing (Protopolis, Clay)

Essentially: it is the UX, stupid : AI is too raw a power, it needs to be channelled.

[0] https://elicit.org/

[1] https://costructure.vercel.app/

[2] https://www.explainpaper.com/

[3] https://github.com/thesephist/revery

[4] https://news.ycombinator.com/item?id=33817682