Generative/Creative AI & Bret Victor’s Immediate Feedback Principle
I recently came across a 2012 video by Bret Victor called “Inventing on Principle.” I was moved by Bret’s idea of identifying and dedicating oneself to a single guiding principle in order to drive innovative and meaningful work.
Bret’s principle can be summarized as immediate feedback for creators. The main point of immediate feedback is that when creators can see the results of their actions instantly, they are empowered to do their best work. In his canonical presentation, Bret demoed (1) a game development environment that lets you adjust its parameters “on the fly”, then (2) a code visualization tool, and then (3) an animation tool where a falling leaf can follow your hand’s motion.
In the picture above, adjusting the jump parameter instantly changes the trajectory of the player character.
In this article, I’ll give you my take on what the future of Creative/Generative AI could look like if the immediate feedback principle is implemented.
How it Works Today
As of today, GenAI tools like ChatGPT and CreativeAI tools like DALL-E or image generation in Bing offer a vast space to boost your creativity. However, the response time is significant — I can almost feel the enormous amount of computation it takes GPT 4.0 to generate a response to my prompt, and it is coming in chunks, one word at a time.
Very often, after reading a couple of paragraphs produced by ChatGPT, I realize that my prompt was not accurate or has been misinterpreted. I need to press “Stop generating”, edit my query, and wait for the new response to get generated. And if I edit the middle of our back-and-forth conversation, it generates a parallel branch.
Until the last word of the new answer is in, it is hard to tell if the new prompt is better or worse!
It is the same story with image generation: first, you craft a prompt, then press “Enter” to get the system started. It takes a short (totally appreciate the tools available almost at no cost to the user!), but significant time to generate your image and display the results.
If the resulting image is not what you were looking for, you need to change your query (a bit or a lot) and restart the process, hoping the next iteration is a hit! Each prompt to the Creative AI engine is like asking a question to the oracle and hoping to get the desired answer.
Generative AI + Immediate Feedback Principle
What would it look like if Generative AI was providing immediate feedback? Maybe it would look like a text document with highlighted changes.
Seeing how small changes are affecting the result helps the user get a better understanding, and perhaps even a “feeling” of how to best structure the prompt.
Creative AI + Immediate Feedback Principle
Applying the immediate feedback principle to image generation feels like an even larger leap in human-AI interaction.
When reality does not give an operating model you’re looking for, it is always helpful to turn to sci-fi examples!
Remember Dolores from the Westworld HBO series? She was sitting at her desk as a story writer, narrating a story off the top of her head, and the machine in front of her was visualizing that narrative, co-creating together with her, making a one-dimensional text into a 3D-rendered picture.
Important Disclaimer: Dolores from Westworld is a machine, so it was rather a machine-to-machine interaction…
A similar scene appears in Blade Runner 2049 when a replicant named K meets Dr. Ana Stelline, a replicant memory designer. The scene shows her in a room creating a memory of a child’s birthday party. Her ideas a instantly visualized and enhanced by the machine she’s operating.
It certainly feels less like “prompts to Creative AI to render 3D models and animate them” and more like painting a picture and breathing life into it.
Additional Thoughts
Following the principle of immediate feedback for Generative and Creative AI opens a new horizon to creativity, but it also comes at a great cost. As with any real-time feedback tool, it must consume more resources than a reactive one.
What do you think about Bret Victor’s “Inventing on Principle”? What are your thoughts about this article? Please share your vision of the future of Generative and Creative AI in the comments or reach out on LinkedIn!