According to the news on December 26, Baidu and NVIDIA have reached an agreement to cooperate to build an AI meta-universe. In addition, at the Baidu AI Developer Conference held today, Raymond Teh, global vice president and president of NVIDIA Asia Pacific, will be invited to attend and deliver a keynote speech.
In fact, since the beginning of this year, chip giant Nvidia's march into the metaverse has been very strong, as early as November this year, NVIDIA launched the core Omniverse of the layout metaverse. But the initial sensation was caused by Huang Jen-hoon's statement at NVIDIA GTC 2021 that 14 seconds of his speech in August were virtual.
Regarding the layout of the metaverse, the senior vice president of Intel has said that it is necessary to increase the computing power by 1,000 times to achieve the metacosm, and the support of computing power is an important aspect of the metacosm, so NVIDIA, as a chip giant, is an indispensable and important component in this regard.
So in addition, how many trump cards does Nvidia have under the wave of the meta-universe, let's find out.
01, the key to the realization of the metacosm: computing power
Although the concept of metaverse is considered to be one of the final forms of Internet 3.0, in fact, the relevant technology to realize the metaverse is still relatively scarce, the core of the metacosm concept is the mapping of the virtual world to the real world, and this mapping must be complete and figurative, and the presentation of the mapping results must be to make everyone who enters the metaverse think that it is "real".
Then, in order to achieve the "real" effect, we must start from two aspects: on the one hand, the visual effect of the virtual real world is presented - VR; on the other hand, the virtual information is added to the real world - AR. But today's human technology, although it can partially implement virtual reality of data for certain specific scenarios, cannot be used on a large scale.
It may be easy for a technology company to make a fictitious avatar, but it is impossible to make a fully realistic simulation of the environment. For example, people can now make a "virtual avatar" similar to a real person, but they can't create a "virtual range" similar to a real town on the Internet.
The virtual and mapping of the scope cannot be carried out, and the layout of the constituent metaversm is to fish for the moon in the water, so the key to realizing the virtual mapping of the range is undoubtedly the computing power. The shortage of computing power has caused VR/AR technology to remain relatively crude, and the same shortage of computing power has caused difficulties in range mapping, and devices cannot run such huge data.
Computing power is actually not only in the meta-universe implementation is the key, in most of the Internet field computing power is the most core driving force, the blockchain field mining most need is a strong computing power support, Nvidia graphics card to provide this strong computing power support.
At the beginning of 2018, there was an unprecedented shortage of GPUs, because of the influx of miners in cryptocurrency mining booms such as Ethereum and Bitcoin, gamers were difficult to find a card, and Nvidia even implemented a "single user purchase limit of 2 cards", the importance of computing power is self-evident.
So, after finding the real core of the problem, it is to solve the problem, and what is the most effective means to improve the computing power? At present, the constraints of cloud computing are still large and unstable. The way to steadily and effectively increase the computing power is a chip that is lighter, more precise, and has a strong enough computing power. And the chip is precisely NVIDIA's strength.
At NVIDIA GTC 2021, the avatar of Huang Jen-hoon can fluently communicate with people in natural language. The technology uses megatron 530B, the industry's largest pre-trained natural language processing model, and also "borrows" Huang's own voice, image and speech posture, and the whole person also has ray tracing effects, even the reflection of glasses as the light changes can be accurately captured.
In terms of intelligence, during the demonstration, it was also asked difficult questions such as climate change, astronomy and biological proteins, and Huang Jenxun of the "Q version" answered smoothly. And most importantly, all the actions of all avatars are generated in real time, which is a completely different concept from pre-rendering, which requires a lot of computing power.
Compared with Facebook(META), which is more inclined to the office and entertainment metaverse, NVIDIA's metaverse with the Omniverse platform as the core is more inclined to build a virtual world that is equal to the real world, and the construction of this virtual world will show unprecedented application scenarios in architecture, engineering and construction, manufacturing, supercomputing and other industries.
Through the above elaboration, improve the performance of the chip to improve the computing power and then enhance the VR, AR, range mapping capabilities, and then complete the technical logic of the construction of the entire metaverse has been formed, obviously, in terms of computing power, Nvidia has made some preparations, although the current "bottom card" of all the metaversms of Nvidia has not yet been fully revealed, but only from the current technical strength, Nvidia has the ability to compete for the future of the metaverse.
02, metaverse layout core - Omniverse
The core of NVIDIA's layout of the metaverse is the Omniverse Avatar platform, which has established cooperative relations with bmw, Ericsson, Foster Construction, WPP and many other large enterprises.
Omniverse was officially launched in 2019 and launched in October 2020 with more than 17,000 customers having tested the experience, and numerous companies including BMW, Ericsson, Volvo, Adobe, Epic Games, and many others are participating in the Omniverse collaboration, initially based on NVIDIA RTX GPUs and Pixar Universal Scene Description (USD) is a real-time graphics and simulation platform designed to transform workflows in the engineering and design industry and accelerate project design and productivity.
According to data provided by Nvidia, after the Launch of the Omniverse Open Beta beta version, about 50,000 users have downloaded it as of mid-November 2021, including nearly 10,000 users in the Chinese market. The Omniverse platform features two systems, Omniverse Avatar and Omniverse Replicator.
Among them, Omniverse Avatar integrates NVIDIA's accumulated technologies in speech AI, computer vision, natural language understanding, recommendation engine and simulation technology for generating interactive AI avatar technology platforms, and Omniverse Replicator is a synthetic data generation engine that can continuously generate synthetic data for training based on existing data.
Omniverse's application is more inclined to engineering and construction, which can build a "virtual reality" digital space through the improvement of computing power and reasonable proportions, and in the space can simulate many characteristics of the real world. Creating shared virtual spaces for designers, engineers, etc. in different fields, and then changing work processes, etc., actually represents the application market for their technologies in the future metaverse. According to NVIDIA, Omniverse can help solve many of the industry's core pain points (quoted from the original text):
1. There are already many 3D collaboration software on the market, such as 3dsMax, Maya, Substance, Unreal Engine, Blender, etc., but the data lacks a common data flow carrier in these software, so Omniverse chose to embrace a lightweight language such as USD. 2. Teamwork Omniverse provides solutions that can collaborate with employees around the world to present WYSIWYG rendering effects, combined with push streaming to achieve instant viewing. This has allowed many teams that can't work face-to-face due to the pandemic to see immediate results and make feedback and modifications based on immediate effects. 3, big data because Omniverse embraces technology such as USD, so it has a natural advantage in opening up large scenes. Avoid many pieces of software from loading too long when opening highly coupled data. 4. Security of data assets In the past, several different data assets needed to be imported, exported, and transmitted using carriers such as the Internet or mobile media, which faced the encryption problem of digital assets. Omniverse enables you to work and access the same platform, whether public, private, or hybrid. This makes it easy to regulate digital assets and introduces many end-to-end encryption mechanisms.
03. Conclusion
For the metaverse, the computing power of the GPU platform, including iterations of various algorithms, will be greatly improved and improved in the next 5 years. Under the improvement of computing power, presentation, and virtual reality technology, the metaverse will gradually spread from the mapping of smaller scenes to the mapping of the scope of scenes, and as the virtual world becomes larger and larger, the real metaverse concept will also be realized. (This article was originally published in ChainDede, authorized to be published by Titanium Media App, author: vernacular blockchain)