More

    AWS Extends AI Reach to Custom Algorithms

    Amazon Web Services (AWS) at an AWS Summit event revealed it is extending a managed service it provides for training algorithms to include custom algorithms developed by organizations investing in artificial intelligence (AI) applications.

    Most AI applications are based on algorithms that are in some cases decades old. It’s only now that the cost of compute and storage in the cloud has made it cost effective to employ those algorithms to drive development of AI applications. However, now that those compute and storage resources are readily available, organizations are also starting to develop a new generation of custom algorithms.

    An Amazon SageMaker Streaming for Custom Algorithms now makes it possible to train those custom algorithms using data stored on AWS. Previously, AWS only supported a narrower range of algorithms it had approved for use with the managed Amazon SageMaker service.

    Dr. Matt Wood, general manager for AI at AWS, told conference attendees that Amazon SageMaker reduces the time required to train AI models by as much as 90 percent. That service can be invoked via application programming interfaces (APIs) that make the resources required for developing AI applications more accessible to a larger number of developers, says Wood.

    AWS now offers a broad range of algorithms spanning everything from computer vision to deep learning algorithms that can all be embedded within an AI application. Those algorithms become more accurate as they get exposed to massive amounts of data in the cloud.

    “Machine learning is experiencing a Renaissance in the cloud,” says Wood.

    AWS also announced it has updated its Snowball Edge appliance by adding support for an instance of an EC2 virtual machine that makes it possible to pre-process data before physically shipping that appliance to AWS to upload data into the cloud.

    It’s unclear at this point how many AI applications will wind up residing on-premises versus the cloud. But as more data is being generated in the cloud, chances are high that the algorithms required to apply AI to that data will follow suit.

    Mike Vizard
    Mike Vizard
    Michael Vizard is a seasoned IT journalist, with nearly 30 years of experience writing and editing about enterprise IT issues. He is a contributor to publications including Programmableweb, IT Business Edge, CIOinsight and UBM Tech. He formerly was editorial director for Ziff-Davis Enterprise, where he launched the company’s custom content division, and has also served as editor in chief for CRN and InfoWorld. He also has held editorial positions at PC Week, Computerworld and Digital Review.

    Latest Articles