Chatting with large language models has been all the rage the past year and a half, but ChatGPT and its brethren have their limits. First, their data is only as current as when they were trained. Secondly, they struggle to interact with your data and your systems. Third, AI systems can be hard to ground against hallucinations and safely extend.

In this talk we'll look at Semantic Kernel, Microsoft's new open-source AI Orchestration framework, and see how it supports retrieval augmentation generation and AI Orchestration workflows to address these concerns.

We'll cover the basics of creating and invoking the kernel and attaching it to large language models on a variety of platforms. We'll cover the idea of plugins and planners and see how plugins can be created, selectively integrated, and unit tested. We'll also see how text embeddings and memory can further enhance the chat experience.

By the time we're done you'll see how semantic kernel can integrate new capabilities and connections into your enterprise chat capabilities.