Get the latest tech news
These Startups Are Building Advanced AI Models Without Data Centers
A new crowd-trained way to develop LLMs over the internet could shake up the AI industry with a giant 100 billion-parameter model later this year.
Nic Lane, a computer scientist at the University of Cambridge and cofounder of Flower AI, says that the distributed approach promises to scale far beyond the size of Collective-1. AI companies currently build their models by combining vast amounts of training data with huge quantities of compute concentrated inside datacenters stuffed with advanced GPUs that are networked together using super-fast fiber-optic cables. Creating an LLM involves feeding huge amounts of text into a model that adjusts its parameters in order to produce useful responses to a prompt.
Or read this on Wired
