Open manisnesan opened 1 year ago
It's not a knowledge answering engine but a reasoning engine.
Limitations
Point of Failure Areas
Problem: LLMs can generate responses only based on the training data.
Solution:
Limitation: Scaling
Option 1 using external data: Read the webpage and ask the question about the document. This helps the model to reason about the document.
Lessons Learned
Commoditized ML
In-house ML
Objectives of ML (until now)
What does it mean for ML Engineers?
Conclusions
Focus on
Presenter: Tanmay Chopra
Infra Thorns
Output linked Thorns
Start Strong
Challenges in Prod
Start Simple
Prompting -> Few Shot Prompting -> Retrieval + Prompting (Langchain, LlamaIndex) -> Iterative Refinement (CoT, Decomposition)
Reliability: Add Structure
Why tooling is important
Different types of tools
Challenges:
Spotlight on OpenAPI tools
Related
Tweet thread - LLMs in Production hosted by MLOps Community
Considerations
Iteration Cycle