Approaching AI prompting
Talking to AI
Antonio
AI is being added everywhere. Big companies are racing to get their LLM (Large Language Model) in front of you.
AI Use
Large Language
This is the vast amounts of text data to learn statistical relationships and mimic human speech patterns
Model
A computational "model" or representation of many mnay (many) statistics.
- Try and understand natural language
- Try and generate a response to a text prompt
- Respond in natural languge
Large Language Models
What's so hard about that??
SO much
Let's look at how difficult language can be in general.
An eggcorn is the alteration of a word or phrase through the mishearing or reinterpretation of one or more of its elements, creating a new phrase having a different meaning from the original but which still makes sense and is plausible when used in the same context.
Egg Corns
Text
Examples
Understanding?
Humans have trouble communicating. Between vocabulary, slang, idioms, etc..
How can machines do it???
They Can't!!
Instead they guess based on LARGE (super large) amounts of examples. And because people are kind of predictably they can appear to understand.
Prompt
We prompt LLM models. We don't script them.
This means there is not an exact output.
Result will vary with each prompt as the LLM tries to give you what you asked for.
Doesn't always work...
Google's Gemini Model
- Zero-Shot Prompting
- Few-Shot Prompting
- Chain-of-Thought Prompting
- Role-Playing Prompts
LLM Prompting
- GPT: Strong in complex reasoning, creative tasks - From: OpenAI
- Claude: Excellent at following precise instructions - Anthropic
- Gemini: Strong in multimodal interactions - From: Google
- Mistral: Good performance with fewer parameters - From:
- Llama: - From: Meta
Popular LLMs
LLMs are everywhere.
In the wild
Desk
By Antonio Ortega
