AI Made Friendly HERE

Fast computing the powerhouse of future for AI

Locals in Tehran march after prayers on Friday in a show of anger at the Israeli attack on Isfahan.

For Nvidia cofounder and chief executive Jensen Huang, computers are the single most important instrument in society today and that recent fundamental transformations in the sector impacts all industries.

Nvidia, a chipmaker, is an integral force in artificial intelligence due to its powerful graphic processing units that accelerate calculations and are well suited for machine learning and massive data crunching.

From a niche firm that made chips for video games that has become the third most valuable tech company after Microsoft and Apple, it has a market capitalization of US$2.15 trillion (HK$16.77 trillion), according to data from CompaniesMarketCap as of a week ago.

At its GTC conference last month in San Jose, Huang shared profound perspectives on three key tech advances that will shape the future of computing and innovation: accelerated computing, AI factories and humanoid robots.

Starting this week, I want to explore why and how the three concepts will transform industries, fuel innovation and shape a more efficient and data-driven society.

The first advance, accclerated computing, is a far cry from the general purpose or traditional computing created in the 1940s, which has inherent limitations due to a lack of human-like cognition and intuition.

Traditional computing relies solely on central processing units that run all tasks sequentially, regardless of complexity.

Around the 1980s, accelerated computing emerged as an alternative to scale tasks efficiently, reduce costs, and consume more computing power sustainably.

It combines CPUs with specialized processors such as GPUs and data or language processing units in an architecture known as heterogeneous computing that distributes workloads to the most suitable processors. GPUs are designed to all work in parallel, rather than one after another, to achieve fast overall processing.

These processors work together harmoniously, creating a balanced system design that delivers both better performance and improved energy efficiency than one that features just CPUs.

Imagine a bustling kitchen at a fine-dining restaurant where the head chef can focus on complex creations while sous chefs and apprentices handle repetitive tasks of dicing, sauteing and simmering.

Accelerated computing is a catalyst for innovation across domains, shaping our world in profound ways.

It plays a pivotal role in various domains, enhancing efficiency and enabling breakthroughs, such as deep learning, scientific simulations and financial risk analyses.

It is the backbone of modern deep learning. GPUs are tailor-made to run many simple operations in parallel and empower neural networks to process vast amounts of data efficiently.

Nvidia recently announced the Blackwell platform, designed to unleash real-time generative AI using trillion-parameter large language models. This accelerated computing platform, using its most powerful single-chip GPU with 208 billion transistors, promises to revolutionize diverse industries, from extreme weather forecasts to precise risk assessments.

Present forecast models can accurately predict the path of storms but are limited to 25-kilometer resolutions, which can miss key details. Scientists use accelerated computing for complex simulations and scientific modeling to explore hypotheses, predict and mitigate natural disasters, and optimize designs more efficiently.

Using a new generative model trained on high-resolution radar-assimilated weather forecasts can resolve extreme events like super-typhoon Chanthu in 2021 to two-kilometer resolutions, with 1,000 times the speed and 3,000 times the energy efficiency of conventional weather forecast models.

Financial stability, fraud detection and economic growth can also emerge from accurate risk assessments powered by accelerated computing that will see traders make informed decisions, regulators monitor market stability and ensure compliance with anti-money laundering, and customers know regulations.

Financial institutions such as RAZE banking and Bridgewater rely on accelerated computing for risk assessment, portfolio optimization, and algorithmic trading.

Accelerated computing has reached a significant tipping point and become a fundamental part of cutting-edge AI technology.

Dr Jolly Wong is a policy fellow at the Centre for Science and Policy,

University of Cambridge

Originally Appeared Here

You May Also Like

About the Author:

Early Bird