Data and algorithms are still the biggest obstacles to the application of AI

KJ2003X1-BA2 Whether it is an industrial brain or an industrial large model, it is the application of AI algorithms. However, the current dilemma is that industrial data is complex and difficult to publicly obtain, and the amount of industrial data accumulated is not enough, and the data quality is not high enough; Second, the algorithm is immature and needs to be continuously optimized.

Relevant people in the first financial industry learned that the applicable algorithms for some application scenarios are difficult, such as chemical reactions, etc., and the output results are very uncertain, and can only be explored through experience, but the results may only meet 80% of the scenarios. So, how are companies in the industry coping?

Liu Weichao told First Finance that from Saiyi KJ2003X1-BA2 Information’s own experience, industrial big data has more dimensions than traditional big data, and the meaning contained in the data is more complex. “In order to solve these problems, we specially built a company-level AI application platform to sort out all the problems generated by the delivery team in the project, and after the expert answers, the knowledge will be returned to the application platform simultaneously, and in some application scenarios,

KJ2003X1-BA2

the training corpus of our platform exceeded 200w characters in a month.” In addition, Saiyi Information also focuses on the use of industry-university-research cooperation, joint laboratories and other diversified forms, and works with multiple teams to jointly build expert AI that ADAPTS to specific fields.”

Huang Geng, COO of Hande Information, believes that, on the one hand, enterprises should have rich experience and domain expertise, so that it is easier to understand the needs and challenges within the industry, and be able to provide customers with more accurate AI solutions. At the same time, the industrial Internet field has a high demand for a large number of data support and analysis, so enterprises with rich data accumulation and data resources may be easier to KJ2003X1-BA2 achieve the landing of AI applications. On the other hand, the establishment of a wide range of partnerships and a sound ecosystem can provide enterprises with more resources and support, and accelerate the promotion and application of AI applications in the field of industrial Internet.

“Companies that are really deep in the customer front line, have a large customer base, and in the process accumulate a lot of data, will be the critical capability providers of the future.” Liu Weichao said.