Getty Images

Amazon boosts CodeWhisperer, AI and ML tools

AWS rolled out updates to a series of AI tools and services, highlighted by improvements to its CodeWhisperer coding system and Textract AI service for extracting document text.

AWS on Tuesday introduced a slate of updates and additions to its Amazon AI services suite, led by broadened support and ease-of-administration advances for CodeWhisperer, the cloud computing giant's machine learning-based coding platform.

While Amazon CodeWhisperer is still in preview, the system and AI coding assistants like it have the potential to make software developers significantly more productive and make coding accessible to many more non-technical and semi-technical enterprise users, said Kashyap Kompella, an analyst at RPA2AI Research.

"However, CodeWhisperer is likely to have the same challenges as Github's Copilot and the contentious issue of using open source code as training data for the underlying machine learning models has to be resolved," Kompella said.

Among the challenges in using open source code for AI models are security, bugginess, obtaining support when an enterprise can't fix a problem on its own, and the abandoning of support for some open source products.

AWS launched the new features and product upgrades at its re:Invent user conference this week.

Automating mortgage lending

Among new AI features are a new loan processing tool, Amazon Textract Analyze Lending, for its Textract document text extraction system.

The lending tool lets mortgage companies automate the classification and extraction of mortgage loan data, which can include many document types such as W2 forms, pay stubs, and bank and tax statements.

National home mortgage lender PennyMac, an early user of Amazon Textract Analyze Lending, uses the tool to process a 5,000-word mortgage application in minutes instead of hours, according to AWS.

Similar technologies have become available for other markets to fill in labor gaps and move people out of repetitive tasks, said Matt Mullen, an analyst at Deep Analysis.

"Textract Analyze Lending looks especially interesting in that it recognizes the need for specifically verticalized industry solutions," Mullen said.

Amazon CodeWhisperer AI updates

CodeWhisperer is likely to have the same challenges as Github's Copilot and the contentious issue of using open source code as training data for the underlying machine learning models.
Kashyap KompellaAnalyst, RPA2AI

New in Amazon CodeWhisperer, originally introduced in June, is added support for the AWS Builder ID authentication system. This lets AI developers securely sign on with an email address for their integrated development environment with the AWS Toolkit.

CodeWhisperer also now supports the TypeScript and C# programming languages, which accelerates code development, AWS said. CodeWhisperer previously supported Python, Java and JavaScript.

Another update is that the platform now makes code recommendations for several widely used AWS services, including Amazon Elastic Compute, AWS Lambda and Amazon Simple Storage System.

The company said it has made administering Amazon CodeWhisperer easier by adding it on the AWS Management Console so any authorized AWS administrator can enable the system for their organization.

More Amazon AI features

In another evolution to one of its existing systems, AWS tacked on a new capability to its Amazon Kendra enterprise search service that supports tabular search in HTML.

Using natural language, users can now find and extract precise answers from HTML tables. This ability addresses the widespread problem of finding information and revealing insights from unstructured datasets like HTML tables. That normally requires pulling information from two-dimensional formats, such as rows and columns, according to AWS.

Meanwhile, with its call center system updates, the vendor also introduced Amazon Transcribe Call Analytics, an addition to Amazon Transcribe, the tech giant's automatic speech recognition service.

AWS said the tool improves the customer experience with speech models that provide real-time conversation insights and identify call sentiment. It also automatically generates call summaries, eliminating the need for contact center agents to take notes.

Dig Deeper on Software design and development

Cloud Computing
App Architecture