Aaron Master publishes a detailed piece on how digrams that depict neural networks can be improved to be Generally Objective Observable Depiction (GOOD).
Hyperdimensional computing uses vectors to encode information and relationships in 10,000 dimensional space, and could supercharge current AI computing models.
Researchers at Google released Zip-NeRF this week. A neural radiance field (NeRF) is a fully-connected neural network that can generate novel views of complex 3D scenes, based on a partial set of 2D images. Zip-NeRF incorporates complex anti-aliasing and faster training than previous NeFT system, with very impressive results.
Researchers at Stanford and Google created a human behaviour simulation with AI agents designed to simulate daily life. When left to run, the simulation agents organised a valentine's day party among themselves, in a better manner than the human evaluator control group.
Supertools is a handy directory of AI tools (built from the great Rundown Newsletter) that tracks the releases of tools during the current Cambrian explosion of LLM based solutions.
