Skip to content
What is AI? An application, or a technology?
Today, as AI becomes more prevalent, it resembles a capability, a capability that permeates various industries and scenarios, enabling intelligent applications.
As AI enters the “capability era”, its “three essentials” (data, algorithms, computing power) indicate that algorithms are becoming increasingly important, as they are the core element that transforms data into business decisions. The creation of a good algorithm model relies heavily on machine learning.
Clearly, AI developers need a powerful and user-friendly machine learning platform. The industry leader Amazon SageMaker recently launched in Ningxia and Beijing, China, and brought its latest integrated development environment, Amazon SageMaker Studio, along with a series of the latest development and debugging tools.
A Complete Family from Foundation to Application
AI is very popular now; in fact, AWS rarely talks about AI as a term; they prefer discussing specific machine learning applications and services.
However, to provide more understanding, before discussing AWS’s machine learning service Amazon SageMaker, let’s first take a look at AWS’s machine learning service family.
AWS‘s machine learning services are divided into three layers, which are basically consistent with the three-tier architecture of cloud services.
The bottom layer is the foundation layer, which includes machine learning frameworks and infrastructure, such as GPUs, FPGAs, containers, etc.
The middle layer is the platform layer, which is the machine learning development platform for developers, data scientists, and data engineers (AWS calls them Builders) called Amazon SageMaker, which is the focus of today’s discussion. This is not a simple service; it includes numerous tools and functions.
The top layer is the application layer, which consists of various AI applications that can be directly invoked. For example, text translation, speech recognition, video analysis, and even cool applications like financial fraud detectors.
All these services are provided to developers and end-users in a cloud service manner, facilitating usage while opening up infinite possibilities, allowing AI as a capability to penetrate various industries.
Dr. Zhang Xia, Chief Cloud Computing Enterprise Strategy Consultant at AWS China, stated that the AI capabilities brought by machine learning are influencing many industries. For example, content classification and subtitle generation in the media and entertainment industry, auxiliary diagnosis and drug discovery in healthcare, personalized recommendations and marketing tools in e-commerce, and risk management and potential customer discovery in finance, etc.
These extensive AI applications are becoming a crucial driving force for enterprises’ digital transformation, enhancing corporate competitiveness while promoting the development of the digital industry and upgrading the national economy.
Providing Better Environmental Support for Developers
Implementing machine learning is a very complex task that involves a lot of trial and error and requires specialized skills. Developers and data scientists must first visualize, transform, and preprocess data so that it can be formatted for algorithms to use in training models. Even simple models require significant computing power, extensive training time, and many specialized personnel.
Dr. Zhang believes that three major factors restrict the further widespread application of machine learning: the scarcity of talent with AI expertise, the difficulties in building and scaling AI technical products, and the time-consuming and costly deployment of AI applications in production. Therefore, the market calls for a cost-effective, easy-to-use, and scalable AI product and service.
Amazon SageMaker was created for this purpose. It is a fully managed machine learning service that allows developers to easily integrate machine learning-based models into the production practices of intelligent applications.
To enable developers to more easily build, train, interpret, inspect, monitor, debug, and run machine learning models, AWS recently launched Amazon SageMaker Studio. This feature has also landed in two regions in China, becoming one of the first regions globally to deploy it, ranking fifth and sixth.
Amazon SageMaker Studio is an integrated development environment (IDE) for machine learning that centralizes all components used for machine learning in one place.
Similar to common software development IDEs, developers can view and organize source code, dependencies, documentation, and other application assets in Amazon SageMaker Studio. Unlike traditional software development IDEs, it does not require purchasing software licenses; instead, it operates on a pay-per-use basis like cloud services.
Today’s machine learning workflows involve numerous components, many of which come with their own set of tools. The Amazon SageMaker Studio IDE provides a unified interface for all Amazon SageMaker features and the entire machine learning workflow, allowing developers to create project folders, organize Notebooks and datasets, and collaborate on Notebooks and results.
Reducing the Development Costs of Machine Learning
Simplification is the design philosophy of the entire Amazon SageMaker, which brings significant value to customers.
“Amazon SageMaker greatly simplifies the process of building, training, and deploying machine learning systems, allowing us to avoid building infrastructure. With its built-in algorithms, we only need to prepare data, and we completed the entire system construction within three months, achieving a breakthrough from 0 to 1. Compared to building models independently, using ECS’s Spot Instance ( bidding instance ) during the training phase can save 70% of costs.”
This evaluation was made by Su Yingbin, Director of Machine Learning Technology at Dayu Unlimited. Dayu Unlimited is a mobile application development company that originated from Wandoujia, and it owns internationally popular short video mobile applications such as Snaptube and Zapee. While building short video services based on AWS technology, Dayu Unlimited is also using machine learning systems for short video recommendation features. Achieving online recommendations for short video content poses a significant challenge for its development team, as the process is complex and requires substantial manpower and time. By adopting Amazon SageMaker, they achieved up to 70% cost savings.
To further assist developers in reducing development difficulty and saving development time, AWS has also launched many new features for Amazon SageMaker.
Amazon SageMaker Notebooks provides one-click enabled Jupyter Notebooks with second-level elastic computing capabilities; Amazon SageMaker Experiments helps developers organize and track iterations of machine learning models; Amazon SageMaker Debugger is used for debugging and analyzing model training, improving accuracy and reducing training time, allowing developers to better understand the model; Amazon SageMaker Autopilot is the industry’s first automated machine learning feature that allows developers to maintain control and visibility over their models; and Amazon SageMaker Model Monitor allows developers to detect and correct concept drift…
These features are also integrated into Amazon SageMaker Studio, allowing developers to easily call them on demand.
The Journey to Becoming a Market Leader
In February of this year, Gartner released the “Magic Quadrant for Cloud AI Developer Services”, and AWS was placed in the leader quadrant, ranking high in both vision and execution dimensions.
It is understood that AWS was rated as a leader by Gartner primarily due to the widespread acclaim received by its automated machine learning model generation tool Amazon SageMaker AutoPilot. This tool can automatically train and adjust the best machine learning models based on user data, lowering the barriers to machine learning development.
Amazon SageMaker Autopilot automatically checks the original data, applies feature processors, selects the best set of algorithms, trains multiple models, tunes them, tracks their performance, and ranks the models based on performance. With just a few clicks, users can obtain recommendations for deploying the best-performing models, requiring minimal time and effort for training.
Amazon SageMaker Autopilot provides developers with 50 different models. Those lacking machine learning experience can easily generate models based solely on data, while experienced developers can quickly develop foundational models using it, allowing teams to iterate further based on that.
It is worth noting that AWS’s AI research team in China has also contributed to Amazon SageMaker. Its DGL graph neural network framework is an open-source code library developed by AWS’s Shanghai AI Research Institute, aimed at simplifying the implementation and deployment of graph neural networks.
Leading products and services also rely on the support of partners to gain user favor.
Yikeluode is a cloud-native consulting service company and a core partner focused on AWS. Product manager Chen Changyou introduced that while helping users make good use of Amazon SageMaker, Yikeluode is also using Amazon SageMaker to build its industry solutions, creating AI applications that truly solve business problems for clients.
AWS’s machine learning services can also seamlessly integrate with other AWS cloud services to form complete application solutions. Chen Changyou stated that Yikeluode is assisting clients in accelerating business innovation, promoting the Amazon SageMaker platform, and utilizing Yikeluode’s industry experience to quickly implement AWS’s SaaS applications, making AI applications readily available to industry clients.
