If there’s a subset of Artificial Intelligence (AI) that closely imitates the way humans think, then Deep Learning (DL) sits atop that list. By definition, DL is Machine Learning (ML) and AI combined to imitate the way humans acquire specific types of knowledge. Deep Learning has caught the attention of the Chinese AI community more over the last few years.
Deep-learning frameworks, which are used by software developers to build AI applications, have been included in the field of next-generation AI. As it represents key cutting-edge technology, DL has been supported by Beijing during the 14th Five-Year Plan period (2021-25).
The application scenarios of China’s AI-powered deep-learning frameworks will be more diversified and buoyed by open-source platforms and large-scale industrial use as an AI industry leader puts it. Moreover, the cost and application threshold to be further lowered, he added.
DL is definitely gaining ground as more industries discover its value and potential. Thus, deep-learning frameworks will be integrated and innovated with more frontier industries such as scientific computing, quantum computing and life sciences.
The support needed for the DL community to grow is also growing. For instance, the first open-source DL platform in China has provided software developers of all skill levels with the tools, services and resources they need to rapidly adapt and implement deep learning at scale. To date, China’s first open-source DL platform has:
- garnered 4.06 million developers on its platform.
- served more than 157,000 enterprises and institutions
- created 476,000 AI models
The industries it has helped move forward are spreading. The DL company has covered a range of industries such as manufacturing, agriculture, healthcare, city management, transportation and finance. It has worked with 22 domestic and foreign hardware manufacturers including the biggest names in the global industry.
However, there are some bottlenecks that hinder the development of deep-learning frameworks. China is still experiencing a shortage of technical talent in underlying AI technologies. Additionally, it takes at least three to six months to put an AI application into use. It entails a long and complicated process to develop a deep-learning framework and build an AI application ecosystem with a large amount of investment.
Nonetheless, DL’s use cases are simply too hard to ignore. DL create opportunities for revamping operations and workload management, and enhancing productivity, experts noted. There’s already some success though. In the wake of the COVID-19 pandemic, a Beijing-based oncology data platform and medical data analysis company released China’s first open-source AI model for pneumonia CT scan analysis, powered by DL.
In essence, the AI model they developed can quickly detect and identify pneumonic lesions while providing quantitative assessment for diagnosis information, including the number, volume and proportion of pneumonic lesions. It has developed an AI-powered pneumonia screening and lesion-detection system, which can pinpoint the disease in less than one minute with a detection accuracy of 92%.
