IEEE Fellow and AI expert Tian Qi joins Huawei Cloud! Why did he come? "To accelerate the implementation of basic AI research"

Jin Lei from Aofei Temple
Quantum Bit Report | Public Account QbitAI

The digital and intelligent transformation trends driven by cloud + AI now complement the new infrastructure and have long been deeply rooted in people's hearts.

As a fundamental change for enterprises, industries and sectors, when Huawei Cloud entered the market, what the outside world was most optimistic about was its genes and customer service capabilities in the enterprise market, as well as its technical strength backed by Huawei's 30 years of accumulation of ICT technology.

If product solutions and customer service capabilities are compared to the superstructure, then basic technology is naturally the economic foundation.

Moreover, Huawei Cloud is constantly strengthening its "economic foundation".

Here’s the latest news: AI expert Tian Qi has been appointed as the chief scientist in the field of artificial intelligence at Huawei Cloud.

Anyone who is familiar with the resume of this AI scientist knows that this is also a signal that Huawei Cloud is further investing in artificial intelligence technology.

Who is Tian Qi?

He holds a Ph.D. from the University of Illinois at Urbana-Champaign, is an IEEE Fellow, and was formerly a full professor in the Department of Computer Science at UTSA.

He has won the Google Faculty Research Award, the UTSA President's Outstanding Research Award, and one of the top 10 most influential scholars in the field of multimedia.

He has published more than 550 articles , including 250+ top international journals and conferences such as IEEE TPAMI, IJCV, CVPR/ICCV/ECCV, and NeurIPS.

In 2018, amid the fierce competition for AI talents, Tian Qi officially joined Huawei and became the chief scientist of computer vision at Huawei Noah's Ark Lab, leading cutting-edge research in the field of AI vision.

But at Huawei Cloud's latest TechWave Artificial Intelligence Special Day, his identity changed like this - he made his first public appearance as "Chief Scientist in the Field of Artificial Intelligence at Huawei Cloud."

What changes will Tian Qi joining Huawei Cloud bring?

In an interview with Quantum位, Tian Qi gave an honest answer:

I have been a teacher in colleges and universities for 17 years, and Huawei has helped me make the transition from college to a company.

Considering the next goal, we hope to help specific departments of the company build business technology competitiveness and directly create business value.

Huawei Cloud is a department that the company is focusing on building, with many AI business scenarios .

So, by chance, Tian Qi became the chief scientist in the AI ​​field of Huawei Cloud, leading the team to strengthen the basic research capabilities of Huawei Cloud's AI.

However, basic research in artificial intelligence ultimately needs to be implemented in scenarios. Huawei Cloud EI has become one of Tian Qi's main battlefields. In his words, "Huawei Cloud EI is closer to AI implementation and users":

With the extensive experience accumulated in the industry over the past decade, Huawei Cloud EI has innovated and consolidated core technologies in the fields of computer vision, speech semantics, and decision optimization.

It is also worth noting that Huawei Cloud EI, which Tian Qi emphasized, has also been continuously delivering results in basic research.

Why is the company’s basic scientific research strength “many firsts in the industry”?

Tian Qi introduced that Huawei Cloud's basic research mainly includes three aspects:

• Computer Vision

• Natural Language Processing

• Voice Interaction
  

He believes that if one were to use a label to reflect Huawei Cloud's strength in basic research, it could be described as "many industry firsts."

The first is in the field of computer vision .

In terms of fully supervised learning , the visual research team led by Professor Tian achieved a Top-1 accuracy of 85.8% on the large-scale image classification dataset ImageNet-1000, reaching the industry's top level without using additional dataset settings.

Before this, Google had been ranked first in this field for three consecutive years, and Huawei's visual team also broke Google's long-term dominance.

In terms of unsupervised learning , Huawei's vision team has achieved 75.0% accuracy in its self-developed algorithm on the ImageNet-1000 dataset, which is 3.9% higher than Facebook's MOCO v2 (71.1%) , and has achieved better generalization performance in a series of downstream task tests.

In terms of weakly supervised learning , in the just-concluded CVPR 2020 Webvision challenge image recognition competition, Huawei's visual research team achieved 82.97% top-5 recognition accuracy and won the championship of this challenge.

In one of the core tasks of computer vision , object detection , Huawei's vision team proposed adaptive data enhancement technology based on its strong R&D capabilities. Huawei's vision team also ranked first in the industry in bounding box detection on the industry-recognized MS-COCO dataset, achieving a single-model detection accuracy of 56.8% and a multi-model detection accuracy of 58.8%. Both indicators ranked first in the world, breaking the monopoly of Microsoft and other companies in this field for many years.

Such strong basic hardware capabilities have also accelerated Huawei Cloud's progress in technology implementation.

In terms of natural language processing and knowledge graphs , Huawei Cloud has applied basic AI technologies to industries including petroleum, healthcare, and recently, epidemic prevention.

For example, Huawei Cloud's AI-based genome analysis reduces the time required to test the virus from hours to minutes. Currently, these systems have been deployed in more than 40 countries and more than 100 hospitals to help local countries fight the virus.

In addition, AI-assisted medical image analysis can output quantitative results of COVID-19 medical images within seconds, helping doctors to make quick diagnoses (using AI to help doctors read CT scans of the lungs).

In terms of voice interaction , the conversational question-and-answer technology is mainly used in WeLink Xiaowei Assistant to answer some sales inquiries, troubleshooting, question-and-answer support, etc.

The above series of achievements also serve as reference points for Huawei Cloud's progress.

But how do we understand Huawei Cloud's progress and what changes will it bring to the cloud market?

Please continue reading.

Huawei Cloud's AI implementation strategy

From Huawei Cloud's accumulation and breakthroughs in basic AI research to the acceleration of industry applications, it also reflects a major trend in the development of the industry - the productization of cutting-edge AI algorithms .

However, with the continuous iteration and record-breaking of AI algorithms and frameworks, and the complexity of AI application scenarios, the implementation of AI is fraught with difficulties.

Therefore, HUAWEI CLOUD provides a full-stack AI solution - HUAWEI CLOUD EI - for governments, enterprises and developers .

The performance of Huawei Cloud EI in recent years has once again proved the correctness of this move.

For example, in terms of AI platforms, Huawei Cloud's ModelArts one-stop AI development and management platform can help users manage the entire life cycle, including AI development and operation stage management, and provide faster and more effective solutions in various aspects such as data processing, model training, model management, model deployment, and AI market.

The HiLens end-cloud collaborative AI development application platform consists of end-side devices with AI reasoning capabilities and a cloud development platform. It provides developers, enterprises, and equipment manufacturers with a variety of development methods that integrate "end", "cloud", and "core", as well as end-cloud collaborative management services for visual devices, further lowering the development threshold.

DAYU , a one-stop data operation platform , is aimed at data workers and covers data integration, development, governance, theme model design, and data openness. It supports access to more than 30 data sources and more than 100,000 concurrent jobs. Users can achieve end-to-end data development and governance through drag-and-drop methods throughout the entire process. At the same time, the platform is open in layers, supports user-friendly integration, quickly builds an industry data transformation platform, and helps companies accelerate data monetization.

In addition, Huawei Cloud EI has explored a large number of projects in more than ten industries and summarized the key path for the implementation of industry AI technology - identifying the right scenarios .

Huawei Cloud EI divides this critical path into three categories: massive repetitive scenarios, expert experience scenarios, and multi-domain collaboration scenarios.

The logistics industry is a typical example of massive repetitive scenarios . For the transportation and security inspection of logistics goods, Huawei Cloud EI uses visual technology to help Deppon Express achieve comprehensive automated inspection, saving a lot of time and labor costs and effectively reducing cargo losses.

In the multi-domain collaboration scenario , Huawei Cloud helped an airport solve problems such as flight docking rate, passenger docking rate, walking distance, ground service consumption, and taxiing conflict rate. The core indicator of docking rate was increased by 5% compared with manual operation, and the conflict rate was reduced by 10% compared with manual operation, which comprehensively improved the work efficiency of commanders and the operation efficiency of the airport.

In the expert experience scenario , the medical field is very typical, and experts are very valuable and scarce resources.

Huawei Cloud EI and KingMed Diagnostics have started cooperation in many areas such as big data and AI. For example, they use AI technologies such as deep learning to study and train massive amounts of pathological sections and expert-annotated data.

Based on the professional annotation of KingMed Medical doctors and the network model construction and training of Huawei Cloud EI Vision Team, the cooperative project has achieved phased breakthroughs. For example, the sensitivity of AI-assisted screening of cervical cancer cells exceeds 99%, which is comparable to the expert level.

It is precisely such achievements that Huawei Cloud EI continues to make efforts in this field.

At today's TechWave Artificial Intelligence Day, Huawei Cloud launched the industry's first AI edge training and inference solution - ModelArts Edge Smart Station .

△ Tian Qi, Chief Scientist of Huawei Cloud Artificial Intelligence, released

ModelArts Edge Smart Station

As an extension of ModelArts platform services on the edge, ModelArts Edge Smart Station has the ability to deliver complete ModelArts platform services in an integrated full-stack manner and also has the following features:

Full-stack AI development service capabilities are decentralized/ready to use out of the box : The architecture is unified with Huawei Cloud, inheriting the mature large-scale commercial architecture of the public cloud. It is stable, reliable, open and compatible, and can provide customers with services consistent with ModelArts on Huawei Cloud.

Not only that, Tian Qi said at the Huawei Cloud TechWave Artificial Intelligence Special Day:

In the future, HUAWEI CLOUD EI will continue to productize Huawei's cutting-edge AI algorithms and open them to AI developers in various industries, driving industrial intelligence upgrades through technological innovation.

So in a nutshell, the goal of Huawei Cloud EI is to "make AI omnipresent."

Of course, you may also have such questions:

How did today’s cloud market become so closely related to cutting-edge technologies such as AI?

Why is Huawei, which has always planned before taking action, so determined and fast in its advancement of cloud + AI?

The changing cloud landscape

What is the cloud? A widely accepted definition is:

It is an Internet-based computing method through which shared hardware and software resources and information can be provided to various computer terminals and other devices on demand, using the computer infrastructure provided by the service provider for computing and resources.

However, with the support of new technologies and the general trend of digital transformation and upgrading, cloud computing has long been more than just a computing method, but a direct carrier for various new technologies such as AI and 5G. Cloud services are the most direct means of technology output.

Moreover, it no longer just reaches Internet and technology companies, but also much broader industries. It is the core way to upgrade industries to intelligent ones, and the prospects are very promising.

In this year's hot new infrastructure wave, the role of cloud, AI, 5G and other technologies in the intelligent upgrading of industries has been clearly affirmed.

And Huawei is the perfect candidate.

After clarifying the intelligent transformation, Huawei has also long clarified its strategic direction towards cloud, AI, and 5G, and is determined to once again cross the technology cycle and win the battle for the "foundation of the intelligent world."

Now, increasing investment in basic scientific research and accelerating the implementation of basic AI technologies in the industry market is just a clearer signal.

Finally, if you have any needs for Huawei Cloud EI product solutions, you can click "Read original text" in the lower left corner to submit your needs.