Huawei released a new white paper earlier this month, detailing its tech predictions for the upcoming years. Notably, the field of artificial intelligence (AI) is believed by the firm to undergo several advancements in the next five years, with a massive increase in the usage of AI, particularly in the enterprise sector. The Shenzhen giant has introduced AI-based products in the past as well, conforming to its vision for the future.
Today, the Chinese firm reached a major milestone in its AI roadmap, announcing Ascend 910, the "world's most powerful AI processor" , and MindSpore, an AI computing framework. With this launch, the firm has unveiled all the key components of its full-stack, all-scenario AI portfolio.
Speaking at Huawei's headquarters in Shenzhen, Eric Xu, the company's rotating chairman, spoke regarding the new products, noting:
"We have been making steady progress since we announced our AI strategy in October last year. Everything is moving forward according to plan, from R&D to product launch. We promised a full-stack, all-scenario AI portfolio. And today we delivered, with the release of Ascend 910 and MindSpore. This also marks a new stage in Huawei's AI strategy."
The Ascend 910, in conjunction with the MindSpore framework, is used to train AI models. Surpassing even Huawei's own expectations in terms of its performance, the new processor delivers 256 TeraFLOPS of computing speed for half-precision floating point (FP16) operations, and 512 TeraFLOPS for integer precision calculations (INT8). Its maximum power consumption is 310W, lower than the initially planned 350W, despite being around two times faster in training AI models based on standard deep neural networks like ResNet-50. The new AI processor belongs to Huawei's series of Ascend-Max chipsets.
Similarly, with the release of MindSpore, Huawei believes it has furthered its AI framework development goals. These comprised of reduction in training times and costs, efficient execution, and more adaptability. Aside from fulfilling these requirements, MindSpore also puts privacy protection and security at the forefront. To clarify this a bit further, the new AI framework deals only with processed information, not actual user data. It also has built-in model protection tech to ensure that the AI models being utilized are trustworthy.
Furthermore, with regards to actual performance, MindSpore has 20% fewer lines of code than other frameworks in a typical neural network for natural language processing (NLP). Combined with its "AI Algorithm As Code" design concept, the product provides a high degree of adaptability and efficiency. Aside from Ascend processors, support is also offered for GPUs, CPUs, and other types of processors. Importantly, Xu also offered some insight into the framework's future, noting that "MindSpore will go open source in the first quarter of 2020. We want to drive broader AI adoption and help developers do what they do best."
Despite the new unveilings, Huawei isn't slowing down its pursuit of AI dominance, and will be revealing more such products in its upcoming Connect 2019 conference, scheduled to be held between September 18 and 20 in Shanghai, China.