Been thinking about AI lately. The new(ish) generation of ‘generative’ bots can create human-level text or images. They acquire expertise by reading a huge amount of text, or viewing a stupendous amount of images (billions of pics/pages, internet-scale datasets) and learning to generate new content consistent with the patterns they discover.
Once trained, a simple text prompt can ask the AI to compose text, or draw an image, based on the theme. Here is the AI GPT3 summarising my books:
“In Singularity’s Children, a gripping Techno-Thriller with a Science Fiction soul, the world is on the brink of collapse due to debt, wars, and inequality. Keith, Niato, and Stella are all caught up in this dystopian future, struggling to survive and to find meaning in an increasingly post-human world. With looming threats of war, plague, and collapse, can they make a difference? Find out in this vivid world-building story that will keep you on the edge of your seat.”
Here is a scene from Saloa (Image is a collaboration btwn my photoshop skills and the AI Dali2). It is based on the prompt:
“a sleek futuristic Zepplin flying above a white futuristic city partially hidden in a green lush rainforest from above.”
AI is able to use the same technique to ‘spot patterns’ in any sufficiently large data sets, for example, genomes. Here is a first step in that direction:
Pretty soon, confronted by a sticky kitchen floor, I’m going to be able to prompt GenomeGen3:
“Clone me a little helper. A friendly miniature mammoth with blue fur, a mop for a trunk, and a domestic hygiene fixation.”
This is what a Singularity feels like!