Nvidia Rides the Generative AI wave at GTC

The NVIDIA GPU Tech Conference (GTC), held this year, could not have been timed better for the company. Today, the hottest technology topic is Artificial Intelligence behind ChatGPT and other Large Language Models. NVIDIA GPUs are the foundation of all this new AI tech. NVIDIA CEO Jensen Huang has reaffirmed his support for LLMs, and the future of AI that is based on them. He calls it the \”iPhone moment of AI\”. AI computers can use LLMs to learn languages from people, programs or images. Generative AI is the ability to create unique, new works based on the knowledge they have gathered.

The latest GPT 4.0 (which was released just before GTC) is a good example of this. The training of these complex models requires thousands of GPUs. In order to apply these models to specific issues, more GPUs are required for inference. Nvidia’s latest Hopper GPU (H100) is well-known for its training capabilities, but it can be divided into up to seven instances, which Nvidia refers to as MIG (Multi Instance GPU), allowing multiple inference models on the GPU. In this inference mode, the GPU converts queries into new results using LLMs that have been trained.

Nvidia uses its position as a leader to create new business opportunities. It is a full-stack AI supplier, which includes chips, software accelerator cards, systems and services. The company has opened up its services to areas like biology. Pricing may be based either on time used or the value of products built using the company’s services.

Source:
https://www.forbes.com/sites/tiriasresearch/2023/03/31/nvidia-rides-the-generative-ai-wave-at-gtc/?sh=30fbb89356cc