Interactions with AI (voice and text)

Using AI
, ,
1

Hi community, I want to share some articles that I wrote recently:

I appreciate your feedback and comments :slight_smile:

5 Likes
2

Wow, ok, a lot to unpack here @elba_sindoni! Iā€™ve got to come back to these to review. Great stuff.

Jumping into the ideas around stochastic gradient descent, which I think you are suggesting that AI interactions work better when they proceed in small, controlled steps rather than dumping everything at once. Love this

  • Ideas start small
  • Most ideas do not deserve equal weight
  • One idea earns depth
  • Synthesis happens later, after evidence
  • Movement between ideas is deliberate (not automatic)

On a personal note, hereā€™s how I see the value of AI in creation right now in my own work: idea evolution and convergence, rather than interaction control over time. Real-time synthesis limits the slop and controls where Iā€™m going with an idea.

image
image600Ɨ561 17.4 KB

3 Likes
3

@elba_sindoni ahhh, thank you for dropping these! I canā€™t lie, I had to parse them (coming from a non designer role) haha. BUT the one that stood out to me most was ā€œDesigning AI Interactions Using Progressive Structure and Explicit Rulesā€

I often find that AI outputs can be SOOO long. While sometimes I do think the long outputs help with opening up different angles, Iā€™d argue its information overload. What I hear you advocating for are more ā€œrulesā€ to make sure the outputs are focused and intentional, which I love!! haha.

On the other side of the coin, Iā€™d ask, how much structure does AI need to be trustworthy without becoming restrictive?

Anywho, thanks for sharing these :sparkling_heart:

2 Likes
4

Hi @elba_sindoni :waving_hand:

I really enjoyed ā€˜Designing AI Interactions Using Progressive Structure and Explicit Rulesā€™. Really helps break down an approach that feels actionable during a development process. Will definitely want this as reference for future projects.

Thanks!

3 Likes
5

Exactly, thatā€™s it, the user sets the direction, and the AI resolves on the fly the userā€™s previously declared needs.

1 Like
6

Yes, the idea is for the AI to respond to what the user actually needs, but it also depends on the personality you want to give the agent, and that goes hand in hand with the product.

In that sense, limits are important. The premise is not to restrict, but to ensure the AI does not make assumptions on behalf of the user. The user should explicitly declare their need, and the AI executes the request.

1 Like