The essential agent terminology, glossary and definitions

Diving into the world of agents can be daunting, with all the agent terms you'll encounter. I'm here to help you understand the essential agent terms in the conversational AI landscape. Whether you're just starting or looking to deepen your knowledge, this guide will cover the agent terms you need to know to navigate the world of agents with ease. Let's unravel the jargon together and make you fluent in the language of agents!

Agents

Agents are advanced AI-powered systems capable of performing tasks or interacting with users in the real world. These could range from customer service agents to virtual personal assistants.

Artificial Intelligence

Artificial Intelligence is the simulation of human intelligence processes by machines, particularly computer systems. It encompasses a variety of technologies that allow machines to perceive, learn, reason, and assist in decision-making.

AI-based agents

Agents that utilize AI technologies to engage in human-like conversations and assist users in completing tasks by understanding context and intent from their inputs.

Algorithm

A set of rules or procedures designed to solve a problem or accomplish a specific task. Algorithms are fundamental to the workings of computers and AI systems.

Assistants API

The Assistants API is OpenAI's interface for developers to create conversational AI models. It provides the tools necessary for building, training, and deploying AI assistants that can engage in human-like dialogue, perform tasks, and provide information in various domains.

Agent builder

An agent builder is a tool or platform that enables users, often without extensive programming knowledge, to create and customize agents for various applications.

Agent

A agent is a computer program designed to simulate human-like conversation based on user inputs via websites and messaging apps. Agents can assist with customer service, information retrieval, and many other tasks.

Agent Deployment

The process of integrating an agent into live environments where users can interact with it, such as websites, messaging platforms, or mobile apps.

Agent Flow

The predefined path that a conversation with an agent may follow, determined by the bot's design to guide interactions. It is used in more rigid, rule-based agents that follow a previously designed decision tree.

Chat widget

A chat widget is a ready-to-use, customizable chat window that an be added to a website. A chat widget shows your agent and enables users to converse with it.

Conversational UI

Conversational UI refers to user interfaces designed to engage users through conversation, typically employing text or voice-based interactions. This approach offers a more natural and intuitive way for users to interact with technology, moving away from graphical user interfaces to a more human-centric mode of interaction.

Popular examples of conversational UI include agents, voice assistants like Siri or Alexa, and messaging apps.

Custom GPT

Custom GPTs enable users to design a ChatGPT variant with bespoke instructions, added knowledge, and select actions for distinct applications, such as learning, entertainment, or work. They are crafted to assist with specific scenarios and can be created without programming expertise, making them widely accessible for customization and sharing.

Deep Learning

A subset of machine learning involving neural networks with many layers. It's capable of learning from vast amounts of data, enabling more complex understanding and predictions than simpler machine learning models.

Entities

Specific pieces of information extracted from user input, such as names, locations, dates, and other data points, that are relevant to processing the user's request.

Entity Extraction

The process of identifying and classifying entities within user input, allowing the bot to understand and act upon key information.

Fallback Response

A default answer given by a AI agent when it cannot understand the query. This response is designed to prevent dead ends in conversation by asking for clarification or offering help in another form.

Function Calling

Function Calling refers to an AI model's capability to call and execute predefined functions or external APIs to perform certain tasks. This ability allows AI to go beyond simple conversation and engage in practical task execution, such as data retrieval, computation, and automation of complex workflows.

Generative AI

Generative AI describes artificial intelligence systems that can create text, audio, video, simulations and other content. These AI systems learn patterns and examples from existing data and use that knowledge to create new and original outputs. Some examples include GPT and DALL-E.

Hallucinations

AI hallucination refers to the instances where a model produces factually incorrect, irrelevant, or nonsensical results. This can happen for several reasons, including lack of context, limitations in training data, or architecture.

Human Fallback

Also called Human Escalation - The mechanism by which an agent transfers a conversation to a human agent when it is unable to process or understand the agent user request adequately.

Intent

The goal or purpose behind a user's input to an agent, such as making a reservation or requesting information.

Intent Detection

The process of identifying the user's intent from their message. Advanced AI agents use natural language processing (NLP) to accurately determine intent, enabling more precise responses.

Large Language Model (LLM)

An LLM is a machine learning model trained on vast amounts of data and uses supervised learning to produce the next token in a given context to produce meaningful, contextual responses to user inputs. The word "large" indicates the use of extensive parameters by the language model.

Machine Learning

A type of AI that enables software applications to become more accurate in predicting outcomes without being explicitly programmed to do so.

Natural Language Processing (NLP)

Natural Language Processing (NLP) is branch of AI that enables computers to understand, interpret, and generate human language in a way that is both valuable and meaningful. It lets agents understand the context and meaning of a message and what the user wants.

Omnichannel

Refers to a seamless and consistent communication experience across multiple platforms and devices, allowing users to engage with services and send messages on different conversational channels, eg. messaging platforms like Facebook Messenger.

Scraping

Web scraping, also known as web harvesting or web data extraction, is the process of using automation to extract content and data from a website. Unlike the manual process of copying and pasting information, web scraping uses an automated method to traverse the web and collect the data you need. This process is often used in collecting information for creating custom GPT agents.

Training Dataset

A collection of data used to train machine learning algorithms. The dataset includes input examples that the model learns from, improving its accuracy over time and allows agents gain their natural language understanding.

Utterance

Utterance is any input or phrase that a user communicates to an agent. Agent utterances are analyzed to understand the user's intent and to provide relevant responses.

Virtual Agent

Similar to an agent, a virtual agent is designed to interact with users, often with a more advanced level of AI, to handle complex tasks and provide personalized assistance.