Jensen Huang, founder and CEO of Nvidia, reveals at a SIGGRAPH conference in Los Angeles that the strategic decision to invest heavily in artificial intelligence (AI) in 2018 was crucial to the company's future. By focusing on AI-based image processing, including ray tracing and DLSS, Nvidia has redefined its hardware and helped shape the industry. Huang emphasizes that this initiative was more than just a financial success and marked the beginning of an AI-powered future, with Nvidia hardware at the forefront.
It highlights the growing role of AI in various fields, from video games to industrial design, and predicts that AI will be at the heart of the future digital industry. Huang also presents recent developments such as the GH200, a dedicated AI hardware for data centers, highlighting their efficiency and cost advantage. Despite an optimistic, enthusiastic vision, certain questions remain regarding the regulatory challenges and ethical implications surrounding AI.
Once limited to fantasy and science fiction stories, artificial intelligence is now not just a reality but an integral part of our way of life. A few decades ago, AI was just a hypothesis about the potential capabilities of machines. Since then, it has quickly become an essential technological feature that powers various aspects of our daily lives, from our businesses to our vehicles.
Artificial intelligence with NVIDIA's Ampere A100
In 2020, Jensen Huang introduced NVIDIA's Ampere A100 GPU architecture, designed specifically for machine learning and HPC markets. Huang said at the time that the A100 represents the largest and most powerful GPU NVIDIA has ever developed, as well as the largest chip made using a 7nm semiconductor process. Equipped with 6912 FP32 CUDA cores, 432 Tensor cores and 108 SM (Streaming Multiprocessors), the A100 is coupled with 40 GB of HBM2e memory and offers a maximum memory bandwidth of 1.6 TB/s.
During the presentation, Jensen Huang discussed various topics, including NVIDIA's recent acquisition of Mellanox, new products based on the highly anticipated NVIDIA Ampere GPU architecture, and significant advances in software technologies. Huang announced that the NVIDIA A100, the first GPU based on the Ampere architecture, represents the largest generational performance leap among the eight generations of NVIDIA GPUs.
Designed for data analysis, scientific computing and cloud graphics, it is already in full production, shipping to customers around the world and boasts an impressive 54 billion transistors. Although the launch date was initially set for March 24, it was delayed due to the Covid-19 pandemic.
Earlier this year, researchers at Tsinghua University in China developed a hybrid chip called ACCEL that combines electronics and photonic computing. This chip, considered potentially more efficient than NVIDIA's A100 chip, uses both photonics (computing with photons) and electronics (computing with electrons). The fully analog ACCEL chip can perform 74.8 billion billion operations per second with just one watt of power.
The researchers say the technology could find widespread applications in wearable devices, self-driving cars and industrial controls, and could particularly benefit from fast and energy-efficient visual processing thanks to photonics. Despite the potential benefits, challenges such as nonlinearity, power consumption and reliability remain a problem in the practical implementation of systems based on this technology.
Silicon Valley praises: Nvidia leads with more than 20 investments in artificial intelligence
Nvidia is establishing itself as a leading investor in artificial intelligence, making more than twenty investments during the year, strengthening its dominant position as a major AI processor provider. The Silicon Valley-based company announced investments in more than two dozen companies during the year, spanning industries from new billion-dollar AI platforms to larger startups and small businesses in areas such as healthcare and energy.
According to Dealroom data, Nvidia participated in 35 deals in 2023, almost six times more than the previous year, cementing its position as the most active major investor in the AI space. This activity has exceeded that of the major venture capital firms in Silicon Valley. The investments, which totaled $872 million over nine months, went to companies using Nvidia's technologies.
In May, Nvidia briefly joined the club of U.S. companies valued at more than $1 trillion, becoming the first chipmaker to reach that figure. The stock's rapid rise in value, which has tripled in less than eight months, is a testament to the growing interest in artificial intelligence, especially given the rapid advances in generative AI, which is capable of conducting human conversations and producing diverse content, jokes and to create poems.
Today, Jensen Huang announced that Nvidia made a pivotal business decision in 2018 that few were aware of at the time, but that has since redefined the company's future and helped reshape a rapidly changing industry. These statements were made at a SIGGRAPH conference in Los Angeles.
According to Huang, the turning point five years ago was the decision to adopt AI-based image processing, embodied by ray tracing and intelligent upscaling, known as RTX and DLSS, respectively. (Quotes are from my notes and may not be verbatim; minor corrections can be made after reviewing the transcription.) “We realized that the traditional technique widely used for 3D rendering, dithering, had reached its limits “, he said. The year 2018 was a pivotal moment when we decided to bet on the company. We had to rethink the hardware, software and algorithms. By reinventing synthetic imaging with AI, we have also reinvented the GPU for AI.
NVIDIA RTX
NVIDIA RTX technology represents one of NVIDIA's most significant advances in graphics cards, opening the door to a new generation of applications that can simulate real-world speeds. Thanks to significant innovations in artificial intelligence, ray tracing and simulation, RTX technology can generate extraordinary 3D models, photorealistic simulations and stunning visual effects faster than ever before.
NVIDIA RTX technology integrates the full power of artificial intelligence with visual computing, enabling developers to build AI-powered applications to accelerate user workflows unparalleled. This significantly increases the creativity of graphic artists and designers and gives them more time and resources through innovative functions for intelligent image manipulation, task automation and optimization of computationally intensive processes.
RTX technology offers advanced real-time cinematic rendering capabilities and leverages optimized ray tracing APIs such as NVIDIA OptiX, Microsoft DXR and Vulkan. Real-time rendering of photorealistic environments and objects, combined with unprecedented precision of shadows, highlights and reflections, gives artists and designers the ability to create exceptional content at unmatched speed.
NVIDIA DLSS
NVIDIA DLSS (Deep Learning Super Sampling) is a neural graphics technology that multiplies performance by using AI to create completely new images, display higher resolution through image reconstruction, and improve the image quality of intensive ray tracing content while producing premium images deliver quality and responsiveness.
DLSS leverages continually improved AI models through continuous training on NVIDIA supercomputers, delivering better image quality and performance in more games and applications. DLSS leverages the unmatched power of deep learning and artificial intelligence to improve gaming performance while maintaining visual quality.
DLSS allows gamers to experience smooth rendering at the highest graphics settings, without frame rate drops during the most intense gaming sequences. It is available on the following GPUs and supports DXR ray tracing at the resolutions listed below:
3840 x 2160 All RTX GPUs
2560 x 1440 – RTX 2060, RTX 2070 and RTX 2080
1920×1080 – RTX 2060 and RTX 2070
Technical efficiency or ethical concerns
Nvidia founder and CEO Jensen Huang's announcement of a massive investment strategy in artificial intelligence in 2018 appears to be characterized by a consciously positive attitude that represents more self-promotion than analysis. Critical. Highlighting AI-based image processing, particularly ray tracing and DLSS, Huang highlights the financial success of this initiative and portrays Nvidia as a major player that has redefined its hardware and shaped the industry.
However, behind this triumphant rhetoric, there remains a gap in the discussion about the real challenges and consequences of AI. The ethical issues and regulatory implications are mentioned superficially, but the speech does not adequately address these crucial issues. Concerns about data privacy, automated decision-making, and the profound societal impact of AI deserve further attention.
Huang focuses on the GH200, a dedicated AI hardware for data centers, and presents this tool as an efficient and economical solution. However, this raises doubts about the dominance of commercial optimism over careful ethical reflection. Efficiency and cost gains cannot ignore the potentially harmful consequences and risks associated with large-scale use of AI.
Although Jensen Huang talks about an AI-powered future and positions Nvidia as a pioneer, it is important to deepen the discussion beyond commercial enthusiasm. For a comprehensive understanding of the impact of this technological revolution on society and industry, a more critical analysis of the ethical implications and regulatory challenges is required.
Source: Jensen Huang, CEO of Nvidia, during a conference
And you ?
Do you agree with Jensen Huang that Nvidia's massive investments in AI in 2018 had a significant impact on the industry?
Jensen Huang cites ray tracing and DLSS as key elements of Nvidia's approach. How do you assess their effectiveness and added value?
See also:
Nvidia is reportedly the largest investor in AI companies, with more than two dozen investments in the field over the course of a year
Nvidia introduces GPU Ampere A100, an artificial intelligence chip with 54 billion transistors and 5 ptaflops of performance
China claims to have developed an AI chip more powerful than the American Nvidia, the Chinese ACCEL would be 3000 times more efficient than the Nvidia A100