Amazon Augmented Artificial Intelligence Goes GA
Amazon Web Services (AWS) announced the general availability (GA) of a new, fully managed service aimed at AI developers who need to add human review to machine learning (ML) model predictions to improve model and application accuracy.
The Amazon Augmented Artificial Intelligence (A2I) service was designed to make it easier to build the review system, structure the review process and manage the human review workforce.
The accuracy of ML predictions (known as "inferences") can be of critical importance in many applications -- those used to scan financial documents, for example, or perform facial recognition for law enforcement. Apps that require especially high levels of accuracy can also require human insight.
But human review systems can be expensive to build and operate at scale, and often involve multiple workflow steps, the operation of custom software to manage the human-review tasks and results, and recruiting and managing large groups of reviewers. Developers can end up spending more time managing the human review process than building the applications -- often enough, they simply forgo human reviews, which results in reduced confidence in the ML apps.
"In situations where the confidence score is lower than desired and/or human judgment is required, reviews can be used to validate the prediction," AWS explained in a statement. "This interplay between machine learning and human reviewers is critical to the success of machine learning systems."
A2I provides developers with more than 60 pre-built workflows and allows them to tap the insights of human reviewers from Amazon's Mechanical Turk crowdsourcing Web site, third-party vendors or their own employees.
Using A2I, developers could, for example, quickly spin up and manage a workforce of humans to review and validate the accuracy of ML predictions for an application that extracts financial information from scanned mortgage documents or an application that uses image recognition to identify counterfeit items online, so that the quality of results improve over time, the company says.
Developers can initiate a human review process for any ML prediction that falls below a certain confidence threshold. They can set the number of reviewers for each prediction and choose how to source those reviewers. A2I also lets devs post instructions for reviewers to ensure consistency, and to adjust those instructions as needed to address recurring reviewer errors.
A2I automates the human review process for such AWS solutions as Rekognition (image recognition) and Textract (character recognition), as well as the SageMaker machine learning platform.
There are no upfront commitments to use Amazon A2I, and users pay only for each review needed. More information on Amazon A2I is available here.
Gladys Rama is the senior site producer for Redmondmag.com, RCPmag.com and MCPmag.com.