GenAI allows for the idea of “Zero” shot learning, basically the ability to build a solution without any training data.
This made it easier to get started with building AI/ML solutions.
However, data is still the most important aspect of any AI/ML system.
This workshop hopes to provide some insight into why AI/ML is a data centric problem and that there is no silver bullet that solves any problem.
An incident is declared. We want to be able to quickly surface relevant information to help the responders.
Question: how might you approach this problem?
Here are two example framing of this problem:
- Ideal solution: A fully automatic Root Cause Analysis Agent
This is a hard problem to solve that will take a lot resources to explore and build.
- Realistic solution: Find and recommend relevant existing resources (e.g. dashboards)
This is a much easier problem, we are simply making recommendation instead of doing diagnosis.
We will explore 3 different approaches today:
- Regex
- Traditional Deep Learning
- GPT4