Generative AI

Generative AI does not work by magic. It generates new, original content from existing content (sometimes referred to as data) like text and images and even sound. Computers generate this new content through unsupervised and semi-supervised machine learning algorithms. Algorithms provide instructions. The algorithm is simply the method by which the AI system does its work like predicting output values from input data.

You may have already discovered through the use of popular tools like ChatGPT that generative AI is good at identifying patterns in data. The type of generative AI model behind a tool like ChatGPT is known as a transformer model. Transformer-based models can take information such as product reviews or customer complaints and create textual content from it.

Textual content can be anything that’s written, like an article, a book, or text on a website. It can also be programmatically-generated text like closed captions and transcriptions.

For example, a generative AI model can take a transcribed customer complaint, find patterns in it such as sentiment, and based on this pattern, emit an appropriate response to the customer. This is known as a Q&A chat application.

It can also take medical transcriptions and find patterns such as medical specialties and then categorize (or classify) a batch of records by medical specialty. This is something that would take oodles of time by hand.