Introduction to generative AI

Common terms 

Terms you might hear in relation to generative AI:

• Model - an AI system. You might have already used some examples, such as ChatGPT, Google’s Gemini or Microsoft’s Copilot. You may even have used image generators that turn your ideas into pictures. 
• Data - the information, like books, websites, or pictures, that is used to train the AI system.
• Algorithms - the instructions that tell the AI system how to process the data.

Training generative AI 

When training the AI system, it learns patterns in the massive amounts of data it's given.  For example, a language model (like ChatGPT) looks at huge amounts of text to understand how words and sentences are usually put together.

It's like a human learning by reading, practicing, and getting feedback to improve. This is also how generative AI learns and improves. 

How generative AI works 

Once an AI model is trained, it's ready to generate new content. When you give it an instruction or a request, the model uses the patterns it learned during its training to make educated guesses about what should come next to create a response that makes sense based on the context of your prompt. 

For example, think about how you anticipate words when you read or listen. If someone starts a sentence with "The cat sat on the...", you'd likely predict "mat" before "banana".

It's important to remember that generative AI doesn't "understand" information in the same way a human does. Instead, it's incredibly good at recognising and applying patterns to create responses that usually make sense and are relevant.

Just like humans, however, these models can sometimes make mistakes. It's important to always check the output for sense and accuracy before you use it. 

Generative AI is powerful because it can help with many tasks, for example: writing assistance, designing images, coding, or even composing music and video. As it continues to improve, it’s becoming an increasingly useful tool for creativity and productivity.