KONE safeguards AI applications with Amazon Bedrock Guardrails (AIM302)
Responsible AI with Amazon Bedrock Guardrails
Overview
Sham Shashank, Principal Product Manager at AWS, led the session on Amazon Bedrock Guardrails, a service for responsible AI development.
He was joined by Tero Hin, VP and Head of Strategic Partnerships at K Corporation, a global leader in elevators and escalators, and Tala Chata, a Senior Generative AI Specialist Solutions Architect at AWS.
Challenges in Generative AI
As generative AI models become more powerful, there are several challenges that need to be addressed:
Undesirable or irrelevant content
Toxic or harmful content
Privacy protection
Prompt attacks
Hallucination (generating inaccurate or fabricated information)
Organizations have additional requirements, such as customization of safeguards, alignment with responsible AI policies, and consistency across models and applications.
Introduction to Amazon Bedrock Guardrails
Amazon Bedrock Guardrails is a service that provides safeguards at every layer of a generative AI application, including the application logic, prompts, and the model itself.
It offers the following key features:
Content filtering: Detects and filters out harmful content.
Prompt attack filtering: Protects against prompt injections and other attacks.
Deny topics: Allows you to define topics that should be avoided in the application.
PII filtering: Detects and masks or blocks personally identifiable information.
Word filtering: Allows you to define specific words to be filtered out.
Hallucination detection: Provides contextual grounding and relevance checks to identify and address hallucination.
Automated reasoning check: Identifies and explains factual claims in responses, ensuring accuracy.
Bedrock Guardrails Architecture
Bedrock Guardrails can be integrated into your generative AI applications using the Invoke Model or Converse APIs.
The Apply Guardrail API allows you to apply the safeguards to any model, whether it's a Bedrock-managed model or a self-hosted or third-party model.
K Corporation's Journey with Bedrock Guardrails
Tero Hin, from K Corporation, shared their journey in leveraging Bedrock Guardrails to build a Technician Assistant application.
K Corporation, a global leader in elevators and escalators, faces challenges in maintaining diverse equipment from multiple manufacturers and addressing the aging workforce in the field service industry.
They partnered with AWS to build a generative AI-powered Technician Assistant that provides quick access to relevant documentation, historical cases, and expert advice to help field technicians solve issues more efficiently.
K Corporation leveraged Bedrock Guardrails, particularly the relevance check feature, to ensure the responses provided to technicians are contextually relevant and accurate, reducing the time and cost compared to their in-house solution.
Conclusion
The session highlighted the importance of responsible AI development and how Amazon Bedrock Guardrails can help organizations build safe and trustworthy generative AI applications.
Attendees were encouraged to explore the Bedrock Guardrails documentation, code samples, and upcoming workshops to learn more and start implementing responsible AI practices in their own applications.
These cookies are used to collect information about how you interact with this website and allow us to remember you. We use this information to improve and customize your browsing experience, as well as for analytics.
If you decline, your information won’t be tracked when you visit this website. A single cookie will be used in your browser to remember your preference.