Unpredictable Black Boxes are Terrible Interfaces
magrawala.substack.com · 2,409 words · saved by 1 readers
Why generative AI tools can be so difficult to use and how we might improve them
I recently decided I should update the profile picture on my webpage. Yes, I too have fallen into the academic trap of using a terribly out-of-date picture on my webpage. As a Computer Science Professor, I figured the easiest way to produce a high-quality photo would be to generate it using DALL-E2. So I wrote a simple prompt, “Picture of a Professor named Maneesh Agrawala” and DALL-E2 made an image that is … well … stunning. From the text prompt “Picture of a Professor named Maneesh Agrawala” a black box generative AI model (i.e., DALL-E2) can produce an amazingly coherent photo. From…
related reading
- Using Artificial Intelligence to Augment Human Intelligencedistill.pub
- Why A.I. Isn’t Going to Make Art | The New Yorkernewyorker.com
- The Shape of AI | UX Patterns for Artificial Intelligence Designshapeof.ai
- Using Artificial Intelligence to Augment Human Intelligencedistill.pub
- Why Chatbots Are Not the Future of Interfaceswattenberger.com
- Replicate - Run AI with an APIreplicate.com
- Natural Language Is an Unnatural Interfacevarunshenoy.substack.com
- Epistemic calibration and searching the space of truththesephist.com
- Lonely Surfaces: On AI-generated Imagestheconvivialsociety.substack.com
- Nano Banana can be prompt engineered for extremely nuanced AI image generation | Max Woolf's Blogminimaxir.com
- Sprachkunstwerkrobhorning.substack.com
- Doomprompting Is the New Doomscrolling.workingtheorys.com