With artificial intelligences in full swing, companies are desperate to get their hands on the limited supply of Nvidia GPUs (used for training and creating AI products). In response to this situation, the chip giant is using its own AI to manufacture them faster and meet the demand.
Bryan Catanzaro, vice president of applied deep learning research at Nvidia, told The Wall Street Journal that about 1,000 people are needed to manufacture a GPU, and each person must know how the different parts of the process fit together. In response to this complex situation, Nvidia has developed an AI system that aims to accelerate the production of its GPUs.
Known as ChipNeMo, the AI system is based on a large language model developed on LlaMA 2 by Meta that the company claims to have trained with its own data. In addition, ChipNeMo’s chatbot function can answer questions about chip design, such as those related to GPU architecture and chip design code development, according to Catanzaro.
For now, the results are quite promising. Since ChipNeMo was introduced last October, Nvidia has found that the AI system has been very useful in teaching novice engineers how to design chips and summarizing the notes from 100 different teams, as reported by The Wall Street Journal.
Nvidia’s efforts to accelerate GPU manufacturing come at a time when companies want to get their hands on their coveted chips to gain an advantage in the AI war. Meta, which has released AI products in the past year such as its flagship language model LlaMA 2 and its Ray-Ban Smart Glasses with AI, is on track to acquire a total of 600,000 GPUs, including Nvidia’s A100 and other AI chips, by the end of 2024.
