Introduction to Generative Artificial Intelligence

A.I. Terms Worth Knowing

For the past week, we've been throwing a lot of terms out there that you might not be familiar with, so before we go live on tonight's training I wanted to make sure we were on the same page and that my using jargon isn't going to confuse you.

You don't need to be an expert on any of these terms, but since they are fairly often used when discussing generative artificial intelligence, it seemed like a good time to provide a broad overview.

Want to print this list out so you have it as a handy reference? I've created a cheat sheet for you that fits on a single piece of paper (if you print it front and back…):

• Click here to download a cheat sheet/glossary of terms

General Terms

• Generative A.I. (Artificial Intelligence): Computer programs that can create new content, like text, images, or music.
• LLM (Large Language Model): A type of A.I. trained on vast amounts of text data to understand and generate human-like text.
• API (Application Programming Interface): A general term referring to a remote way to access a software or service and is how individual applications make use of different LLMs.
• GPT (Generative Pre-Trained Transformer): A type of LLM that can understand and generate human language that seems natural by trying to continue text in a way that sounds like how humans speak and write.
• NLP (Natural Language Processing): The field of A.I. focused on understanding and generating human language.
• Fine-tuning: Adapting a pre-trained A.I. model for a specific task or style.
• Text-to-Text Generation: A.I. creating written content based on a text prompt.
• Text-to-Image Generation: A.I. creating images based on a descriptive text prompt.
• Chatbot: An A.I. program designed to simulate conversation with human users.
• NSFW (Not Safe For Work): This is not an A.I. specific term but when referencing A.I. it is about the safeguards some models have in place to keep sensitive materials from being generated. For example, some models (especially the biggest private models) may not allow you to generate erotic scenes (or even risque scenes) or any content related to violence or crime.

LLM Models

• Public Model: An LLM created open-source, publicizing it's training data, and allowing you to run the model locally on your desktop (if you have the proper equipment) or on your own server. Best for privacy-minded users or users that want to prompt or generate “NSFW” concepts such as romance or thrillers.
• Private Model: An LLM created in closed-source, where the training data may or may not be made available publicly, and the model is less likely to be installable on a local server and can not be compiled directly by any third parties.
• Open Router: Open Router is a service that provides a common API that applications can use to access many different LLMs, both private and public models, and provides some additional features. This makes it easy for a 3rd party application to allow users to choose from many different models, even if they don't live in a country supported by a specific LLM or don't have the proper equipment to run their own local copy of a public model LLM. (Not really relevant for tonight's training, but useful to know what it is if you hear it mentioned or want to use software that supports it.)
• Hugging Face: Hugging Face is a service that provides a place to access and download many different public LLMs (and other types of A.I.) and to provide an API for accessing their own public interfaces for those LLMs for users who don't have the equipment to run them locally. They do have some free tiers for any user to use their own hosted interface or their API, but also offer paid tiers based on actual server time consumed. (Not really relevant for tonight's training, but useful to know what is if you hear it mentioned or want to use a service that supports it.)

Writing Specific Terms

• Prompt: The initial text or instructions given to an A.I. to guide its output.
• Token: A unit of text (like a word or part of a word) that A.I. models process. Tokens are the most common unit of measurement for the cost of running an A.I.
• Context Window: The amount of memory an LLM has available to it, often expressed in a number of tokens. Basically, the model’s attention span.
• Hallucinations: The process of an LLM generating text that seems believable and as natural as possible, but is actually made up and based on events, places, or things that don’t actually exist.

Common A.I. Activities

• Content Summarization: An activity where an A.I. condenses longer texts into shorter versions.
• Style Transfer: Applying the writing style of one text (or illustrative style of one image) to another using A.I.
• Sentiment Analysis: An activity where A.I. determines the emotional tone of a piece of writing.
• Grammar and Style Check: An activity where A.I. reviews and suggests improvements to a piece of writing.
• Plagiarism Detection: A tool that specializes in assessing the likelihood that a piece of text was plagiarized and identifying copied content.
• A.I. Writing Assistant or Virtual Assistant: Software that helps with various aspects of the writing process using A.I.

I will try to avoid using too much jargon on tonight's training, but as these are common terms that are often used, I expect that I'll let some slip here and there and want to make sure you know what I am talking about.

If you haven't already, you should register so you can attend live and not only watch our demo and tutorial but also be able to ask your questions and get direct answers.