Glossary

On this page, you will find an overview of important terms and concepts to help you better understand Alan.

AI Assistant Alan

Alan is an AI assistant that uses artificial intelligence (AI) to assist users with various questions and tasks. The assistant is designed to understand and respond to natural language, enabling intuitive interaction with people. Alan analyzes the entered prompts and generates responses based on them. For this, Alan utilizes Large Language Models (LLMs) and other advanced machine learning algorithms.

Large Language Model (LLM)

A Large Language Model (LLM) is a form of generative artificial intelligence specialized in understanding and generating natural language. It is a sophisticated artificial neural network trained through machine learning by analyzing large amounts of text. During this training process, the model has learned how words and sentences relate to each other and how texts are structured. The trained model can generate responses appropriate to a given context.

LLM Context

The context of a Large Language Model (LLM) refers to the information available to the model at a given point in time. It includes the system prompt, the entire chat history, sources such as uploaded files or excerpts from knowledge bases, the user's current prompt, as well as the (partial) response currently being generated by the model. When generating a response, the LLM completes this current context. Thus, the context significantly influences the model's response behavior and quality. However, the maximum size of the context is limited, which also restricts the amount of information available to the model at once and the length of the response it can generate.

Prompt

A prompt refers to an input given to a Large Language Model (LLM) that initiates a response from the model. The model analyzes the current context, including the prompt, and then generates an appropriate response based on the patterns and structures learned during its training. The prompt is crucial for the quality and relevance of the response provided by the model.

Tips for crafting prompts can be found here.

System Prompt

A system prompt is a special type of prompt used to control the response behavior of the Large Language Models (LLMs). It is located at the beginning of the context and is not defined through user prompts in the chat input. System prompts are used to set general instructions, background information, or rules that the model should consider throughout the interaction.

To adjust the system prompt in Alan, you can use experts.

Hallucinations

Hallucinations of a Large Language Model (LLM) refer to situations where the model generates information that is incorrect or even misleading. These hallucinations occur because the model has learned patterns and structures of language but is not always able to verify the accuracy or source of the information. Hallucinations are a known issue with LLMs and highlight the limitations of this technology. It is important to be aware of these limitations and take appropriate measures to minimize their impact.

Tips for minimizing hallucinations can be found here.