Artificial intelligence (AI) applications requiring access to massive amounts of data to drive machine and deep learning algorithms are clearly an ideal workload for public clouds. Amazon Web Services (AWS) this week made it clear that it views AI as a critical workload for the cloud by launching no fewer than 13 different additional AI services.
Launched at the AWS re:Invent 2018 conference, those services span everything from tools that make it easier to train AI models to an additional instance of the AWS P3 service that provides access to 100G Ethernet connections.
In addition, AWS promised to deliver a new type of processor, dubbed AWS Inferentia, optimized to provide AI applications with access to hundreds of teraflops of processing horsepower per processor. AWS CEO Andy Jassy promised conference attendees that AWS Inferentia will be able to support multiple data types and AI frameworks, including TensorFlow, Apache MXNet and PyTorch.
“We don’t believe in one tool to rule the world,” says Jassy.
AWS also announced the general availability of AWS Elastic Inference, which provides a more efficient approach to running the AI inference engines on which AI models depend in order to process and analyze data.
The most pressing issue, however, remains simply making it easier for the average IT organization to train AI models. To solve that issue, Jassy says the tools AWS announced this week will, for example, automate the tagging of data to accelerate the training of AI models. AWS is also making available as a technology preview Amazon Textract, which uses machine learning to instantly read multiple types of documents to accurately extract text and data without any manual intervention required. AI developers can use Amazon Textract to now extract data from millions of document pages in a matter of hours.
AWS also announced that is opening an AWS Marketplace for Machine Learning, which provides access to over 15 algorithms and AI models that IT organizations can download to jumpstart development of AI applications.
When it comes to all things AI, it’s clear that size matters. AWS is betting that as more AI applications are developed and deployed on its cloud, the overall size of the AWS cloud will continue to exponentially increase. The larger the AWS cloud becomes, the less expensive it becomes for AWS to operate its cloud at scale. It’s not clear just yet when all these AI applications will find their way into production environments. But as the cost of building these applications continues to decline, it’s now only a matter of time before AI models are pervasively embedded in every business process imaginable.