General Purpose Generative AI language models

By: Aiden Kim

In recent years, artificial intelligence has taken the world by storm. Generative AI language models, especially in the form of ChatGPT, are widespread in modern society, with occupations across the world scrambling to find what use cases, if any, generative AI language models can bring to modern society.

Well how does an generative AI language model work?

This is a highly simplified overview of the methods generative AI language models work, but in general, AI language models work by having a huge dataset, often totaling over tens of thousands, if not even millions or billions, of datapoints represented within large files.

These datapoints are then sent to a model to be trained, in a process that is commonly referred to as dataset training. In simpler terms, AI models are taught to string words together by representing them as numbers, thus, over successive generations of careful adjustment, learning how to communicate with complex lingua.

Afterwards, the now trained AI model, often referred to as simply trained model weights, can be applied to generate responses as appropriate to theoretically any subject. In this state, an AI is essentially operating in a separate mode called “inferencing,” which uses the already existing model weights to generate an appropriate response.