How AWS gets ideas for its new AI products and services
At re:Invent 2019, Amazon executive Swami Sivasubramanian spoke about a commitment to democratizing machine learning, and making its benefits available to all.
- AWS re:Invent 2019: Leadership session: Machine learning
Swami Sivasubramanian, vice president, AWS machine learning, delivered a keynote presentation at the leadership session during AWS re: Invent 2019. Sivasubramanian spoke about why Amazon decided to launch products and services like Deep Java Library, Amazon SageMaker Autopilot, and Amazon CodeGuru (Preview) during the event. The common threads linking these products: a commitment to democratizing machine learning and making its benefits available to all.
Here are three key takeaways from Sivasubramanian’s presentation:
- The state of machine learning today is similar to art during the Renaissance
The clearest indicator of the explosion in the adoption of machine learning is the wide implementation across a variety of industry sectors and disciplines. NASCAR is giving fans a realistic race experience with AWS Media Services. Intuit has been able to generate a series of custom machine learning models — paired with other machine learning services like Amazon Textract — to create better, more personalized financial management services. And by using predictive modeling and machine learning techniques, CARE™ Disease Prediction analyzes more than 14 billion medical claims to identify early indicators of disease for local markets and disease onset across the United States.
- Machine learning is not a new phenomenon
Machine learning has been around for a long time. After all, the seminal deep learning paper, Gradient-based learning applied to document recognition was published twenty-one years ago. Then why did the widespread adoption of machine learning take this long? The explosion in adoption of machine learning is directly linked to the expansion of computing capability and data storage made possible by the cloud.
However, Sivasubramanian points out that we are still very much in early days when it comes to machine learning, as it is still largely a domain restricted to experts. Analogous adoption arcs can be found in other fields: for example, the first digital SLR camera was only introduced 150 years after the invention of photography. Just as the smartphone made (nearly) everyone a professional photographer, the key to accelerating the adoption of machine learning is to make it accessible to all developers.
- AWS’ new products and services aim to make machine learning accessible to all developers.
AWS’ new products are services help simplify machine learning on three fronts:
- Easier to build: Developers in more traditional environments have found it challenging to build apps that leverage machine learning. Though Java is the most popular language in enterprise, there are very few resources to work with deep learning. Python continues to be the language of choice. As a result, Java developers spend a significant amount of time interpreting and rewriting code to develop deep learning applications. AWS announced Deep Java Library (DJL) to bridge the gap between data scientists and enterprise developers, and make machine learning applications easier to build.
- Easier to scale: Machine learning deployment is not as easy as software development. Algorithm selection is still largely an explorative process requiring broad knowledge of the field. This is why AWS launched Amazon Sagemaker Autopilot that makes machine learning more transparent and explainable, while reducing the time it can take to train, deploy, and scale apps. Another examples is Amazon SageMaker Operators for Kubernetes a new capability that makes it easier for developers and data scientists using Kubernetes to train, tune, and deploy machine learning (ML) models in Amazon SageMaker.
- Easier to apply: This involves connecting the dots for customers in machine learning, and making it more real for business and IT decision makers. To this end, AWS launched Amazon Kendra, a highly accurate and easy to use enterprise search service that’s powered by machine learning. Kendra delivers powerful natural language search capabilities to internal enterprise websites and applications so end users can more easily find the information they need within the vast amount of content spread across a typical company. Customers can use Kendra's connectors for popular sources like file systems, web sites, Box, DropBox, Salesforce, SharePoint, relational databases, and Amazon S3. In a similar vein, Amazon CodeGuru (Preview) is a machine learning service for automated code reviews and application performance recommendations. Amazon CodeGuru (Preview) helps developers find the most expensive lines of code that hurt application performance, and gives them specific recommendations to fix or improve their code.
Products and services featured in the article include:
Amazon CodeGuru (Preview)