Large Language Models

LLMs (Large Language Models) are a form of Generative AI. Generative AI is able to generate to produce data. They require a form of prompt to generate from and so can be through of as transformers. Examples of generative AI include:

• Chatbots. These create text from a text prompt. Examples include: ChatGPT, Copilot, Gemini, and LLaMa.
• Text-to-image. These generate images from a text prompt. Examples include: Stable Diffuson, Midjourney, and DALL-E.
• Text-to-video. These generate videos from a text prompt. Sora is an example.

More recent and advanced LLM based tools have become very popular but how do they function?

In essence LLM based chatbots are very advanced auto-fill tools. To properly function a model needs to be created, called a Large Language Model. This LLM allows the tool to predict what the next word in a sentence would be and therefore create sentences and paragraphs based on a user’s input.

This is why LLM based chatbots struggle with answering mathematical questions. They can only give correct mathematical answers to questions they have in their LLM. Common questions such as 1 + 1 will be fine as it is in many text books, however unique questions not found in text books will trip it up.

LLMs are permeating the internet, our computers, and society. Many find them useful to get answers but is it ethically correct and how damaging is it to the environment?