Introduction to Generative AI: How Generative AI Works

How Generative AI works

This guide provides an introduction to various aspects of Generative AI for students to consider as they study at Victoria University.

Generative AI and Large Language Models are subsets of Artificial Intelligence.  This is how Generative AI relates to the wider Artificial Intelligence environment.

Generative AI Models

Generative AI models are unsupervised computer programmes (that means they can learn on their own) that generate new output in response to a prompt or input.

Some examples of these models are ChatGPT, Claude, Microsoft Copilot and Google Gemini.

The research (DEC, 2024; Hirabayashi et al., 2024) tells us that more than 80% of students surveyed indicate regular use of one or more of these tools.  Students are using these tools for academic research, seeking answers to everyday questions, paraphrasing, grammar checking, or writing drafts.

 

 

Digital Education Council. (2024). Digital Education Council global AI student survey 2024. Digital Education Council.

Hirabayashi, S., Jain, R., Jurković, N., & Wu, G. (2024). Harvard undergraduate survey on Generative AI. Harvard University Undergraduate Association.

What the different technologies do

NAME

JOB

HOW IT LEARNS

EXAMPLES

Artificial intelligence

To mimic human intelligence, thinking and learning.

All of the below

All of the below

Machine learning

to make predictions

learning from data with human intervention - supervised

recommendations and speech to text

Deep learning

to make predictions

learning from data with reduced human intervention – semi-supervised

predictive text, autonomous vehicles, spam filtering

Generative AI

pattern recognition and creative content generation in response to a prompt

Independent learning from data - unsupervised

generate text, music, images

LLM

to understand language patterns and make predictions in response to a prompt

Independent learning from data - unsupervised

spelling & grammar checkers

Generative AI model

To generate new output in response to a prompt

GenAI + LLM

OpenAI, Quillbot, ChatPDF, Ref-N-Write, Trinka

 

What do we mean by "supervised", "unsupervised" and "semi-supervised"?

  • Supervised learning models consume data that has been prelabeled by humans
  • Unsupervised learning models discover patterns in data that hasn't been previously labeled
  • Semi-supervised learning models involve an iterative process that works with both labeled and labelled data
  • Reinforcement learning models use algorithms that can tune models in response to feedback about performance after development

 

Lawton, G. (2024). The different types of machine learning explained. TechTarget. https://www.techtarget.com/searchEnterpriseAI/tip/Types-of-learning-in-machine-learning-explained