• It will be simpler for programmers to discover errors in the data they process for AI applications thanks to a new capability for SageMaker Studio Notebooks.
  • AWS is releasing a few trained AI models that can use geographic data for use cases including agricultural production monitoring and urban planning.

By releasing an improved version of SageMaker, its neural network creation platform, Amazon Web Services Inc. recently increased the scope of its Artificial Intelligence (AI) offerings.

More than a dozen AI development tools are included in SageMaker, which was first released in 2017. Software teams use the platform to build neural networks, train them, evaluate their performance after deployment, and carry out associated tasks. AWS recently revealed in conjunction with the changes that tens of thousands of customers use the platform.

Bratin Saha, AWS’s vice president of artificial intelligence and machine learning, said, “Many customers are using ML at a scale that was unheard of just a few years ago. The new Amazon SageMaker capabilities announced today make it even easier for teams to expedite the end-to-end development and deployment of ML models.”

Amazon SageMaker Studio Notebooks

Developers can build new neural networks using a feature of SageMaker called Amazon SageMaker Studio Notebooks. It is a managed version of the well-known AI development tool, Jupyter. Developers can use Jupyter to prepare a dataset for analysis, build a neural network to process that dataset, and then analyze the processing results in the same interface.

A new feature for SageMaker Studio Notebooks will make it easier for programmers to find mistakes in the data they process for AI applications. According to AWS, the feature pinpoints data quality problems and suggests solutions. The tool automatically generates the software code required to apply a remediation approach when a developer chooses one of the suggested remedies.

It can take a long time to deploy a Jupyter-built neural network. The neural network and any external components that the container might need to function must be packaged by developers into a software container. The next step is to set up the cloud infrastructure that will house the program.

AWS says that SageMaker Studio Notebooks can now assist with that task too. Neural networks can now be packed automatically into software containers without requiring manual labor on the developers’ side. Additionally, the feature provisions the infrastructure necessary to operate neural networks and destroys hardware resources that are no longer needed.

AI applications are frequently created by one or more software teams in the organization. Teams can exchange AI model code and other software components more easily because of the collaboration feature coming to SageMaker Studio Notebooks. The functionality, according to AWS, enables the organization of an AI project’s components in a shared workspace.

Efficient AI development

One of the many AI development tools that AWS offers as part of SageMaker is SageMaker Studio Notebooks. AWS is upgrading several additional platform elements with today’s improvements.

SageMaker is a tool that some businesses use to create neural networks that analyze geospatial datasets or datasets that contain location-specific information. A logistics company, for instance, may create an AI that studies traffic flow in a specific location and identifies the quickest delivery routes. SageMaker will soon have new features added by AWS that will make it simpler to build AI models that can analyze geospatial data.

With just a few clicks, SageMaker users can now easily add geospatial data from outside sources into an AI project. The company claims developers may access data via AWS’s Amazon Location Service map platform, open-source datasets, and proprietary sources like satellite constellations.

Geospatial data is sometimes difficult to assess in its raw form due to its complexity. AWS has equipped SageMaker with functions that can transform geospatial data into a format that is easier to process automatically. In addition, AWS is introducing a selection of pre-trained AI models which can leverage geographical data for use cases like urban planning and crop production monitoring.

New AI testing capabilities

After building a neural network with SageMaker, programmers may use a brand-new feature called shadow testing to ensure it performs as planned. The capability, according to AWS, is more efficient than conventional techniques for determining the dependability of AI systems.

A company’s current AI software is used for shadow testing, assessing fresh neural networks’ dependability. It duplicates the user requests submitted to an organization’s existing AI software. After that, the feature sends the copy to the new neural network under test by the corporation to see if it can process it consistently.

AWS claims that the shadow testing function in SageMaker automatically generates a monitoring dashboard for assessing AI applications. The dashboard monitors statistics like mistake rates and latency. Using this capability, programmers can evaluate a new neural network’s performance compared to current software before releasing it for use in real applications.

AI governance is made more accessible

AWS unveiled new development features along with a set of AI governance tools. The cloud giant claims that businesses may use the tools to ensure that AI projects powered by SageMaker adhere to internal requirements and cybersecurity regulations.

Administrators can more easily control user access to a company’s SageMaker environment using the first tool, Amazon SageMaker Role Manager. Through a centralized console, administrators may control who has access to what SageMaker features and how.

Amazon SageMaker Model Cards, another recently launched AI governance tool, will assist software teams in managing the data generated as part of machine learning initiatives. These data include neural network reliability test outcomes and AI training datasets. AWS claims that SageMaker Model Cards allow engineers to save this data in a central area for quick access.

The Amazon SageMaker Model Dashboard completes the AI governance features that AWS unveiled for SageMaker in the event. After AI models are used in production, it offers a console for checking their dependability. The technology can aid administrators in spotting mistakes and resolving them more rapidly, such as a significant drop in the processing results accuracy of an AI program.