Glossary
Who is Jensen Huang?
In short
Jensen Huang is a co-founder and the chief executive of Nvidia, whose graphics processors are used to train AI.
People · Jensen Huang
MMOOZZGG.com editorsChecked: 1 min read
For a child
Imagine a class where one top student solves problems one by one, while next to him a thousand kids each solve one small problem at the same time. Together they finish much faster. Jensen Huang’s company makes computer parts that work like such a big, friendly class. Programs that draw pictures and answer questions learn on them.
In plain words
Huang founded Nvidia in 1993 and has run it ever since. The company’s graphics cards were made for games: such a chip draws a picture on the screen by doing thousands of simple calculations at once. It turned out that neural networks need the same parallel computing, only for multiplying large tables of numbers. So Nvidia graphics cards became widely used to train and run models. In 2016 Huang personally brought a DGX-1 server, built specially for machine learning, to OpenAI.
For business
Huang’s work largely decides what AI costs for any company. Models are trained and run on graphics processors, and their price and availability set the prices of cloud services and the budget for your own server. If a company wants to keep a model inside its own network, the main question is video memory: the model must fit into the card’s memory, or it runs slowly or does not start at all. So before buying, count the model size, the number of simultaneous users and the load. There are risks too: delivery times and restrictions on importing equipment. Memory calculations and graphics card prices in Russia are collected in a separate breakdown.
For a specialist
Huang founded Nvidia in 1993 and has held the posts of president and chief executive since its founding; he sits on the board of directors. Before that he worked at LSI Logic (1985–1993) and AMD (1984–1985), and he earned a bachelor’s degree in electrical engineering at Oregon State University and a master’s degree at Stanford. In 2017 Fortune named him Businessperson of the Year. For AI the company’s key line is graphics processing units (GPUs) with thousands of cores for parallel computing, the CUDA software platform and servers for training models. On 15 August 2016 TOP500 reported that Huang personally delivered a DGX-1 system, Nvidia’s machine learning server called a “supercomputer in a box”, to OpenAI’s headquarters in San Francisco. The system was meant for OpenAI’s heaviest computing tasks.
Example
A company decides to run a language model on its own server so that customer data does not go outside. An engineer chooses an Nvidia graphics card by the amount of video memory: after quantization the model fits into one card, and without it you need two.
Related terms
Questions and answers
Sources
Not sure whether your company needs this?
AI solutions map