Jon Udell: Best Practices for Working with Large Language Models
Sunday, April 28, 2024
… you don’t need a low-level understanding of the neural networks at the core of large language models in order to align with the grain of that architecture. Although I can’t explain how LLMs work – arguably nobody can – I’m able to use them effectively, and I’ve begun to codify a set of guiding principles.
Here’s my list:
- Think out loud
- Never trust, always verify
- Use a team of assistants
- Ask for choral explanations
- Outsource pattern recognition
- Automate transformations
- Learn by doing
Good rules here. I’ll add a couple of guidelines:
-
Don’t pretend the AI is intelligent or conscious. It’s fancy autocomplete. When you talk to the AI, you’re talking to a version of yourself.
-
Also: The AI doesn’t have feelings. Take advantage of that. The AI is never going to get bored or discouraged. (But don’t be cruel to the AI—it’s a bad idea to be cruel, even to an inanimate object that cannot suffer.)