#84 Hands on with DALL-E
Saturday, October 01, 2022
Last week OpenAI announced that the DALL-E waiting list is no more and the access is open to everyone. Anyone can use DALL-E and let their imaginations go wild creating AI-generated images. I spent the entire Thursday afternoon playing with DALL-E and challenging my creativity. Three years ago, I captured a scene in Van Vihar, Bhopal where two deers were grazing across a river standing in a dense forest, facing each other. Guess what is the input to my first AI-generated image?
This is the prompt I entered - “A realistic photograph of two dotted golden deers across the river, facing each other, grazing the grass in a semi dense dry forest“
The result I got -
If you go and check out my post on instagram, the two pictures are quite similar. The difference exists largely because of my inability to explain to the AI what I exactly want.
Halfway through my experimentation, I realized how dumb machines are. The AI system reminds me of that kid (read every human on this planet) who has a lot of potential but never utilizes it. The agent's performance depends on how articulate we are in our ability to explain what we want. The output(my behavior) you get depends on the input(your behavior) you give. We often say that we are in the midst of a technological revolution, with state of the art AI systems coming up. What we fail to take into account is that we are also in the middle of a human revolution, a revolution in human capabilities. In the span of a few decades, humans have learned to think with their fingers, to imagine that a flat screen is really as big as the world, to create new personae for themselves, to expand their social interactions in number and kind, to write and design in new ways, to visualize complex concepts, to find information in seconds, and incorporate that information into a constantly evolving awareness. And there’s still a lot going on!
The quality and the detail of your AI generated image depends on the quality of your prompt. Yes, a picture does speak a thousand words. But in this case, a picture will only speak a thousand words if your 400 words are capable enough to speak for the picture. Maybe there will be a marketplace for buying and selling prompts(Update: Something like this already exists) just like we trade crypto and stocks. This reminds of what Naval Ravikant said in one of his interviews. In future, all jobs will be creative jobs, the non-creative and repetitive jobs will be done by the AI agent in a better and effective manner. What are your thoughts on this?
While you put on your thinking caps, I’ll leave you with a few more prompts that I tried -
Prompt: “Surfing in the pacific ocean with my dog”
Prompt: “Driving in a Tesla on the lanes of Los Angeles during night”
Prompt: “Humans leaving Earth in a spaceship, flying towards Mars.”
Prompt: “A serious writer writing a horror fiction novel on a humor street with the sun setting“
A FEW THINGS KEEPING ME AWAKE
Article: Is there anything bigger than infinity?
YouTube: Life After Death by PowerPoint
Song I am listening to: Radioactive
Thought of the week: "To grow faster is to make better mistakes."
MEME OF THE WEEK
SARCASTIC REFLECTION
Here are the last three posts if you were too occupied to read them -