Huawei Cloud releases ModelArts Pro: Teaching people how to fish is worse than giving them fish

Qian Ming from Aofei Temple
Quantum Bit Report | Public Account QbitAI

Huawei has sounded the clarion call for the implementation of AI in the industry.

During the Huawei Developer Conference 2020 (Cloud), Huawei Cloud launched ModelArts Pro, the world's first professional suite for enterprise-level AI application development.

The core goal is to solve problems such as high threshold, low efficiency and difficulty in implementation of industry AI development.

△ Yuan Jing, general manager of Huawei Cloud General AI Service, released ModelArts Pro, a professional suite for enterprise-level AI application development

As the culmination of Huawei Cloud's AI development platform, ModelArts can automate all aspects of training and deploying AI models, including data labeling, model parameter adjustment and debugging.

Since its launch in 2018 and its iterative upgrade in 2019, ModelArts' capabilities have continued to improve and it has attracted increasing attention.

Now, Huawei Cloud has launched ModelArts Pro, targeting the many pain points in the implementation of industry AI. This is good news for a number of industry AI solution providers, as it will enable them to complete industry AI application development and replication more quickly.

This is also a major benefit for industry companies, not only because they have one more choice, but also because ModelArts Pro can directly address industry pain points.

How should industry AI be implemented?

It is an established fact that the rapid development and iteration of technologies such as artificial intelligence, cloud computing, and big data have triggered a new round of industrial revolution and created new space for economic growth.

This has not only spawned a large number of startups, but technology giants are also actively embracing this trend, and concepts such as industrial Internet, industry + AI, and AI + industry have followed.

As for how industry AI should be implemented, all participants have given preliminary answers: AI must be combined with scenarios, and one must not only understand AI, but also understand the industry, accumulate industry know-how, and so on.

Although we understand the “big principles”, it is not that easy to implement them. Once you delve into industry AI, you will find that there are pitfalls everywhere:

The actual application of AI algorithms varies between different industries. Customizing a system requires a lot of work, and it is basically impossible to complete delivery quickly. Industry AI developers usually face tight cycles and business pressures, and it is difficult to have time to systematically and deeply learn all the basic knowledge in the AI ​​field. All these make it difficult for enterprises to quickly launch AI services that achieve the desired results.

Usually, after enterprises customize and purchase AI models, they are unable to self-iterate, and industry data is highly complex, making it difficult for expert knowledge to be effectively passed on;

This results in many similar projects being difficult to reuse and requiring a large amount of repetitive development work. Most general AI models in the industry, although universal, have limitations and do not support industry-specific business processes.

The advent of ModelArts Pro can directly solve these pain points in industry AI.

How to solve it in ModelArts Pro?

Yuan Jing, general manager of Huawei Cloud General AI Services, said that ModelArts Pro provides a new way to implement AI in the industry.

The underlying layer of ModelArts Pro relies on the basic platform and engine provided by ModelArts, including industry-leading intelligent labeling, automatic learning, neural network architecture search and other technologies to ensure efficient development of upper-level AI applications while reducing manpower investment.

Unlike the general pre-training services currently provided by major cloud service providers, ModelArts Pro also has built-in industry-specific algorithms based on Huawei's years of accumulation in the enterprise field.

In some scenarios, industry AI application developers only need a small amount of labeled data to obtain high-precision AI applications.

Throughout the entire process, ModelArts Pro can also consolidate the accumulation of algorithm experts and the knowledge of industry experts into corresponding suites and "industry workflows" for reuse.

In addition, ModelArts Pro also provides an AI market function, where industry AI developers can share AI workflows in different industry scenarios with each other, further activating the knowledge accumulation of different industries by Huawei Cloud EI. This will truly empower industry AI application developers and comprehensively improve the efficiency and implementation of industry AI development. In Yuan Jing's introduction, we can see that the capabilities of ModelArts Pro have been verified in actual applications.

During use, industry AI developers can easily complete AI application project development and continuous iteration with the help of pre-set industry workflows or industry workflows published by other developers on the AI ​​market.

The cost reduction and efficiency improvement are obvious. Huawei Cloud AI experts said that with ModelArts Pro, AI projects that used to take several weeks to deliver can now be completed in as little as one day, and the entire process is as simple as operating an everyday office software.

Models developed through ModelArts Pro can also be continuously iterated.

Yuan Jing said that for specific industries and scenarios, developers can use the atomic components (Atom) provided by ModelArts Pro according to their own needs to flexibly arrange workflows and realize AI application development.

ModelArts Pro implementation

The first case introduced is from the energy industry.

With the help of Huawei Cloud ModelArts Pro knowledge graph development kit, PetroChina quickly built professional models in the field of oil well logging and realized intelligent identification of oil, gas and water layers.

Effect: Evaluation time is shortened by more than 70%, and recognition accuracy reaches the level of well logging interpretation experts.

Moreover, with the help of AI, the previously dormant logging data, scattered stratigraphic data, etc., have also been quickly activated and mobilized, helping engineers quickly obtain valuable geological and oil reservoir information.

The second is a case in the logistics industry, from Yingzhi Technology, which applied Huawei Cloud ModelArts Pro text recognition development kit to develop various international logistics document recognition interfaces, and quickly realized the automated and structured output of document information.

Using traditional OCR interfaces, it takes 7 days to develop such a tool, but it only takes 3 minutes using ModelArts Pro.

Moreover, with the same manpower input, the efficiency of document information entry has increased by 50 times, which not only improves work efficiency but also saves 60% of labor costs.

Finally, there is the retail industry.

Miqi Cake uses the ModelArts Pro visual development kit to achieve zero-code AI model development. During the entire process, the clerk only needs to upload the cake picture to complete the image annotation, model training, and service update online, and the recognition accuracy rate can exceed 99% in seconds.