tailwinds

• Problem: Current workflows still decouple the human/AI. It looks something adjacent to this: AI reasons, provides probabilities, then gives a summary. Humans respond approve/disprove.
• Reactions: We have become more interested in this idea of fluidity as a core embed into our interaction workflows. AI acts as a collaborator in human thoughts and creative activities instead of a separate actor. Tasks will feel more fluid since we will no longer create new apps, but build workflows on top of the user’s existing apps.

theses

thesis 1: AI as an enabler of human creativity

• texture: AI’s role is fundamentally tool-like — think the modern version of an avocado peeler or icing piper where we are supporting the user’s creative process by streamlining menial tasks. Move the interaction paradigm from converging to an answer to diverging to more potential ideas
  • cares about increasing the quantity of ideas: “design space thinking in creativity support tools”
• another example: think of when you’re planning a wedding. Oftentimes we are looking at multiple software platforms to find what we need — this is the grunt of the work but also removes from the creative process. Thus, we should remove this
• features: interfaces should inject AI as a tool to help humans iterate over ideas and arrive at a solution

supporting papers:

• Human-AI co creation — uses dimensional reasoning to encourage users to reason on the problem at a higher level. llm generates multipke dimensions to a prompt instead of a single output. This essentially constructs the “design space” the user can then explore in a systematic manner

  

• Concept of engaged interfaces — for humans to explore knowledge rather than completing a defined task

• Surfacing the craftsman: “Compare the feeling of doomscrolling to kneading dough, playing an instrument, sketching... these take effort, but they're also deeply satisfying… but we should have made typing feel like painting.”

thesis 2: AI as an automation tool to offload human tasks

• texture: AI should remove most/all human involvement from the equation other than as an observing agent to mitigate when the AI makes a mistake. Believes AI should remove humans need to perform
• examples: nearly any military targeting system is aiming for this.

thesis 3: AI as a tool for thinking

• texture: AI should integrate deeply into existing workflows as an invisible productivity layer that can support cross analysis and thinking. Offload menial tasks and stitch together the disparate knowledge representations of apps into one pane of glass — a thought slate. Remove the need to learn new interfaces