AI landing scenarios increase, but making money is still difficult. According to YEEO, nearly 90% of AI companies were in the red in the whole of 2018, and 10% of the companies that made money were basically technology providers . From concept, technology to scene and landing, artificial intelligence based on big data faces the challenge of data itself.90% of artificial intelligence companies in the red? The following and gold investment editorial to take a look!

We often mention big data, but in fact we do not need so much data, the future trend of AI is the rise of small data. At the GMIS2019 Global Data Intelligence Summit in the north of the city, Stanford University professor, founder and CEO of Landing.ai, Wu Enda said .

Specific case for factory cell phone screen scratch inspection. Currently, many people use their eyes to check their phones for scratches, and if there are 1 million scratched phones, AI can effectively identify phone scratches. However, the reality is that there are not millions of different factory injured cell phones. At this point, smallshotlearning, which uses less data to reach the same correct conclusions, could help advance the field.

The urgency of small-shot learning is that it is difficult for AI technology to realize its full value due to data silos and data fragmentation caused by data privacy protection.

Unlike AI used in competitions that require tens of millions of photos for training, AI penetrates deep into the industry and sees that the data is small and granular, which means that without connected data, it is difficult to use more advanced AI technology. Yang Qiang, chairman of the International Society for Artificial Intelligence, professor at the Hong Kong University of Science and Technology, and chief artificial intelligence officer of micro-banking, said.

In May this year, the State Network Information Office published the "Data Security Management Methods (Draft for Public Comments)," which puts forward the idea of utilizing the network to carry out activities such as data, storage, transmission, processing, and use, as well as the protection and supervision and management of data security within China.

Yang Qiang believes that China's version of the GDPR is coming soon, data privacy is being strict and comprehensive, and the dimension and scope of data that enterprises can use in practical applications is not large. Strict data privacy protection provides an opportunity to upgrade AI technology.

The next team building is not only to rely on star engineers, but also to build a perfect, cross-disciplinary, cross-functional team. At the same time, do not expect AI to work immediately, but rather multiple attempts to rationally budget for the benefit curve of AI development. Instead of using traditional processes to evaluate AI projects, you should set up appropriate KPIs and goals for the AI project team.

There are more and more applications about AI, but AI change in the enterprise is not as simple as developing an app, do not expect AI to solve all the problems, and do not expect AI projects to be successful at once. Wu Enda said .