AI ALPHABET SOUP
This three-letter acronym stands for "Large Language Model"
What is LLM?
This is what you call the text input you give to an AI model
What is a prompt?
The process of teaching an AI model using data
What is training?
This AI technique involves rewarding good outputs and penalizing bad ones
What is reinforcement learning?
When an AI makes up false information confidently, it's called this
What is hallucination?
When you customize a pre-trained model for your specific business needs
What is fine-tuning?
These billions of internal settings determine how an AI model behaves and responds. Generally, a model with more of these will have a greater capacity to learn complex patterns in data and achieve better performance
What are parameters?
The maximum number of words/tokens an AI can process at once is called this
What is context window (or context length)?
This technique lets you give AI temporary knowledge without retraining the entire model. Common in scenarios where there is proprietary data.
What is Retrieval Augmented Generation (RAG)?
This type of AI can perform tasks independently without constant human guidance
Agentic AI
An AI that can process both text and images is called this
Multimodal
The practice of testing AI outputs for accuracy, bias, and safety before deployment
What is model evaluation (or evals)