Data and AI Training

Home | Power BI | Excel | Python | SQL | Generative AI | Visualising Data | Analysing Data

GenAI Model Description

This section describes how generative AI work under the covers. You don’t need to know this to use LLMs effectively but it helps to understand the reasons for their strengths and weaknesses. Most of the AI tools we use in this course are based on large language model (LLM).

Large Language Models

A large language model (LLM) is a type of Generative AI program designed to understand, generate, and interact with human language. It is trained on vast amounts of text data. An LLM uses statistical analysis and language modelling techniques to repeatedly predict the next word in a sentence to build a response to a prompt.

It does not always choose the most probable next word. This is controlled by a setting named “temperature” If the temperature becomes warmer, the LLM is less likely to choose the most probable word and the output becomes more creative.

An LLM can answer questions, write essays, and create code. Here are a couple of useful terms

• prompt - input to the LLM, usually a sentence or paragraph
• completion - the output (or response) from the LLM

For example, here are three different completions of the next three words of a prompt “A tasty breakfast”

• could vary depending
• may include oatmeal
• includes scrambled eggs

Another example: given the prompt “She walked through”, possible continuations, with made-up probabilities, are:

1. fire (10%)
2. hell with a smile (5%)
3. the park (3%)
4. the fair (2%)