Nvidia, a graphics chip manufacturer with a capitalization of 2,2 trillion USD, has become the "life source" of synthetic artificial intelligence (AI) models. Besides hardware, this company's strength also comes from computer code lines developed over the past 20 years, making opponents who want to dethrone Nvidia wary. Currently, more than 3 million programmers globally depend on the CUDA platform to build AI and other applications.
Originating from the technology developed by Intel, called OneAPI, a group of technology companies – UXL Foundation, has planned to build a new suite of software and tools capable of programming a variety of AI chips. This open source project aims to bring computer code that can run on devices with any chip and hardware.
“The problem is how to create an open ecosystem in the field of machine learning that can promote performance and be suitable for a variety of hardware.”, said Bill Hugo, director of Google's high-performance computing unit. The search giant is a founding member of UXL and supports the technical direction of the project.
Reuters said that in the first half of this year, UXL will prepare to determine the basic technical specifications, before turning it over to engineers to refine and perfect at the end of the year.
This technology alliance also does not rule out the possibility of "admitting" additional members such as cloud giants Amazon, Microsoft and other chip manufacturers. Since its founding in September 9, UXL has received technical contributions from third parties, in addition to institutional members, and those interested in using open technology.
Intel's OneAPI is ready to use. In the next step, experts will create a standard programming model specifically for AI. In the long term, UXL will also support Nvidia hardware and programming code.
The UXL Foundation's project is one of many efforts to eliminate Nvidia's dominance of AI-powered programming software. According to data statistics firm PitchBook, venture capitalists and businesses have poured more than $4 billion into 93 separate projects.
Interest in defeating Nvidia through software links has received increasing attention since last year. This graphics company has cast such a large shadow that few startups can make a breakthrough when CUDA is on paper an attractive, full-featured software and is continuously developed by Nvidia and its partners. co-programming.
“The important thing is that people have been using CUDA for over 15 years and using the platform to build other programming code.”, commented Jay Goldberg, CEO of financial and strategic consulting firm D2D Advisory.