The Do's and Don'ts: GenAI Applied to Infrastructure

This 30-minute presentation by Roxane Fischer from HashiCorp delves into the application of Large Language Models (LLMs) and Generative AI in infrastructure management. Explore how LLMs function, their training methodologies, and probabilistic nature through comparative examples of Python and Terraform code generation. Learn about the challenges in Infrastructure as Code (IaC) generation, including context limitations that can produce suboptimal configurations and security risks from models trained on public data that might propagate vulnerabilities. Understand the critical distinction between generative AI (which creates code/content) and synthesis AI (which analyzes existing information like logs to identify issues). Discover the future potential of contextual AI that integrates comprehensive environmental context, including configurations, service interdependencies, and security policies for more effective infrastructure management.