Google is trying to bridge the gap between them and OpenAI in the world of artificial intelligence. Among the steps taken is to restructure the Google Assistant section so that it can help the development of Bard's artificial intelligence. Today Google shared information about their latest supercomputer which is claimed to be more powerful than the supercomputer produced by NVIDIA.
Google's latest supercomputer uses the fourth generation Tensor Processing Unit (TPU V4) chip. As many as 4000 TPU V4 chips are combined in a supercomputer used to train Google's PaLM artificial intelligence. This chip has actually been used since 2020 and only shared about its capabilities today.
When compared to the NVIDIA A100 chip produced at the same time, the TPU V4 is up to 1.7 times more powerful, and 1.9 times more energy efficient. In addition to Google PaLM, this supercomputer with TPU V4 chip is also used by Midjourney to train their generative artificial intelligence.
Although the TPU V4 is claimed to be more powerful, the supercomputer for training the world's artificial intelligence is still Perlmutter equipped with 6159 NVIDIA A100 chips.