Kicking re:Invent into high gear, Amazon Web Services Inc. today unveiled a massive array of artificial intelligence solutions spanning more than a half-dozen product categories, from code debugging to cybersecurity and contact center automation. And that’s on top of a formidable lineup of infrastructure announcements.
The new AI offerings fill in key pieces of the product vision AWS C hief E xecutive Andy Jassy outlined during his keynote address this morning. The provider’s objective, he said, is to deliver a broad portfolio of machine learning tools spanning multiple layers of abstraction. The foundational layer is the freshly upgraded SageMaker developer toolkit, while the top layer includes pre-packaged AI applications such as Kendra and Amazon Fraud Detector.
SageMaker gets smarter
SageMaker, the provider’s managed toolkit for building, training and deploying machine learning models, is the AI service that has received by far the most new features today. The main highlight is a web-based integrated development environment called SageMaker Studio in which data scientists can develop neural networks.
Building an AI usually requires engineers to toggle between several different coding tools. SageMaker Studio provides the ability to write code, experiment with model changes, visualize data and perform debugging in a single interface. It furthermore provides access to the other new development tools AWS debuted for SageMaker at re:Invent.
SageMaker Studio’s debugging features are powered by an AI glitch-hunting tool that engineers can customize to track down specific issues. In addition, there are new features that pick out the best neural network for a given project, prepare AI training datasets and manage experiments . Finally, there’s an open-source toolkit that will enable developers to build graph-based neural networks for use cases such as chemical analysis.
Amazon Augmented AI
Companies using machine learning to process sensitive data, such as mortgage applications, have to take into account that their AI will sometime get tripped up. A mortgage application may contain unclear handwriting or might have poor resolution. Organizations typically deal with the issue by forming a team of reviewers to double-check their models’ decisions.
Amazon Augmented AI , which also debuted today, makes it easier to manage human reviews. It allows developers to build rules into their AI applications that automatically determine when decisions should be routed to humans for a second opinion.
is another AI service aimed at developers, but it can be …read more
Source:: Daily Times