High-quality output at low latency is a critical requirement when using large language models (LLMs), especially in ...

"We designed a new language to talk about these new systems," Zardini says. This new diagram-based "language" is heavily based on something called category theory, he explains.