Implementation of specific task models by AI21 Labs on AWS

In a move towards improving business operations through artificial intelligence (AI), AWS has introduced a new solution based on AI21 Labs’ Task Specific Models (TSM). These models, specifically designed to answer questions, summarize text, correct grammatical errors, and paraphrase, are now available on Amazon SageMaker JumpStart.

The integration of AI21’s TSM into AWS allows companies to easily access these models by subscribing to them through AWS Marketplace and setting up their environment on Amazon SageMaker. According to AI21 Labs, their models offer an improved rejection rate to ensure that answers are not provided unless they are based on real documented content. This approach results in higher accuracy rates and cost savings, making AI21’s TSM an effective solution for common business tasks.

One of the main benefits of TSM is that, unlike base models or Foundation Models (FM), they are specifically trained to perform unique and specialized tasks, such as creating varied versions of text while maintaining a consistent tone or condensing lengthy texts into concise summaries. Examples of mentioned TSM include AI21 Contextual Answers, AI21 Summarize, and AI21 Paraphrase.

The process to start using these models is relatively simple and well-documented. Companies need to set up an AWS account and a SageMaker environment, subscribe to TSM on AWS Marketplace, and then deploy SageMaker endpoints. Additionally, adjusting service quota limitations is required to use the necessary GPUs. Estimated costs for running these models vary. For example, using the Contextual Answers model on a ml.g5.48xlarge GPU for approximately 15 minutes costs $20.41, while using the Summarize and Paraphrase model on a ml.g4dn.12xlarge GPU for 10 minutes each costs around $4.94.

In addition to being easier to implement and manage, AI21’s TSM offer significant benefits in terms of accuracy and costs. Compared to FMs, which require more elaborate prompt engineering and may generate inaccurate or irrelevant answers, TSM provide more precise answers within their specific scope. This makes them especially suitable for applications in sectors like customer service, legal, finance, and education.

The use of AI21’s TSM on AWS proves to be an effective tool for improving and optimizing various business operations, reducing the need for extensive prompt engineering and offering tangible benefits in terms of performance and costs.

via: MiMub in Spanish

Scroll to Top
×