Types of Generative AI

The sections below list common types of generative AI, with brief descriptions and some illustrative examples.

Text: (text to text, audio to text, image to text)

ChatGPT 3 or Chat GPT4, Bard, and Bing are among the most popular large language models (LLM) producing AI tools. They work by answering a prompt the user creates. Prompt creation, also known as prompt engineering, is an integral part of getting the best performance from a language model AI tool and faculty and students will need to practice to understand how to create the most useful prompts. Some examples from Bard and ChatGPT3 are found below:

Prompt used for Bard (Google’s generative AI tool) and ChatGPT3 (Open AI): Describe the mascot for Boston University in 50 words. 

Bard’s Response:

Sure, here is a description of the mascot for Boston University in 50 words:

Rhett the Boston Terrier is the mascot of Boston University. He is a black and white dog with a short snout and big ears. He is often seen wearing a scarlet and white BU jersey and a hockey helmet. Rhett is known for his playful and energetic personality. He loves to cheer on the BU Terriers at sporting events and he is always willing to pose for pictures with fans.

ChatGPT3’s Response:

The mascot for Boston University is Rhett the Boston Terrier. Rhett is an energetic and spirited canine figure, characterized by its distinctive black and white coat, and represents BU’s tenacity and passion in sports and academics. Rhett embodies the university’s pride and determination, captivating the BU community’s spirit.

Visual Art: (text to image, text to 3D image)

Some generative AI tools are designed to turn a written description into an artistic image. DALL-E (Open AI), Firefly (Adobe), and Bing Image creator (Microsoft) are examples of image creating AI tools commonly used. Below is an example of an image created using generative AI.

Prompt used for Bing: Create a science fiction type female chatbot chemistry tutor that looks friendly. Use shades of purple and blue, add chemical structures. (Microsoft Bing Creative 2023) The New Bing – Learn More

Bing’s response:

Audio: (text to audio, audio to text, audio to audio)

The conversion of speech to text or vice versa is another form of generative AI that is gaining popularity. A common use of this tool is the creation of closed captions for videos or even live meetings in platforms such as Zoom. Transcripts, notes, summaries can all be created from speech recognition. Podcastle and Speechify are examples of  tools that can also generate text to speech, creating human sounding voices to be used for recordings, podcasts, customer service calls and more.

Video: (text to video)

AI generated video combines animated visuals from generative adversarial networks (GANS) and AI generated audio to create video content. These tools can be used in the education world in a multitude of ways to enhance and support student engagement and learning. Synthesia AI Video Maker, GilaCloud, InVideo, and Lumen 5 are some of the industry standards for AI video creation.

Coding: (text to code)

Web and programming developers utilize AI coding assistants to assist in developing and writing code faster and more accurately. Prompts are used in a similar fashion to other AI tools to develop code snippets, edit, detect errors, and auto complete common lines of code.  GitHub CoPilot, Amazon CodeWhisperer, and Tabnine are examples of AI coding specific platforms.

Prompt: Write code for R statistical software that can be used to calculate ANOVA of a large data set (ChatGPT 3.5 created 8/17/23)

Last updated: August 18, 2023