This guidance for marketing and communications support is aligned with the overall AI guidance from OIT:

OIT’s Artificial Intelligence Webpage

Generative AI Guidance Knowledge Base Article

***

Large language models (LLMs), such as ChatGPT, are artificial intelligence (AI) models that have been trained on massive amounts of text data to learn to generate human-like responses to prompts. These models are based on deep neural networks and are capable of understanding and producing complex natural language text, including language patterns, nuances, and even emotions. 

LLMs have the potential to transform the way businesses interact with their customers. Because LLMs can be trained to generate personalized and targeted marketing content, such as email campaigns, social media posts, and website copy, they have the potential to materially influence public perception and consumption.

As a proof-point, the above two paragraphs were originally drafted by an LLM, with small adjustments made for clarity and focus.

The field of AI and opinions about how to react to AI in academia are evolving rapidly. This webpage refers primarily to the writing assistant and media generation LLMs. But AI models are also being used in other ways, such as organizational effectiveness. They can also be used to automate customer service interactions, providing instant and accurate responses to customer inquiries, complaints, and feedback. 

AI models will continue to grow in capabilities, and are already incorporated into software and search engines.

Blank Space (small)
(text and background only visible when logged in)

The Process

We describe the LLM generation process as a two-step activity: An end-user writes up a framed description of the content they want generated (the ‘prompt’), and the AI model generates the requested content (the ‘output’).

Communications Advantages

LLMs have the potential to streamline and automate procedural, boilerplate marketing, and communications tasks.

Efficiencies

AI can help create initial drafts of content (e.g., emails, blog posts, cover letters, summaries) and generate stock photography-equivalent assets, allowing for time savings in everyday and procedural tasks.

Idea Generation

AI can generate topic ideas, formulate initial and counter arguments, and provide context from an explicitly defined perspective or persona.

Editing and Revisions

AI can provide edits and revisions based on specifically targeted prompts and a quoted writing sample.

Guidelines

Generative Student Photography

As with stock photography assets perceived to be real students, we do not recommend generating photorealistic student photography.

Generic Disclaimer Wording

Attach a disclaimer to stories, pages, and content that heavily use LLMs in their generation. The following is the standard language that should accompany all AI-generated copy.

Content on this page was generated (wholly, or in part) using a Large Language Model tool. All AI-generated content is reviewed, edited, and revised to publication, and follows the Institute’s Editorial Style Guide.

Trademark and Legal Considerations

See the Brand Identity Standards policy and the Use of Name and Marks section of the Institute brand standards. Georgia Tech trademarks should not be used in third-party platforms without prior written approval from authorized representatives of the Institute.

Use of Name and Marks

Challenges of AI

While LLMs can be impressive in their capabilities, caution must be exercised when using them. So, keep the following top of mind when considering the possible implications involved in the use of AI-generated content.

Inherent Biases

An LLM’s output is determined by its source content. This means there is the likelihood of AI models retaining biases from the stereotypes and misinformation present in human writing on the internet. As such, all AI-generated content should be reviewed for inherent or structural biases.

Because of this deficiency, special care should also be taken to ensure the creation of prompts that minimize bias; this necessitates the explicit definition of a range of perspectives and personas. Exercise extreme caution in avoiding expressions, content, and imagery that may misattribute American cultural norms.

Inequity

AI models restricted through paid access may exclude certain audiences, as availability is limited to those who can afford to pay for them. This creates a potential disadvantage for individuals or groups who may not have the resources to access or utilize these powerful tools.

Inaccurate Content

AI-generated content may contain factual errors, incomplete quotes, and erroneous findings. Therefore, output should be reviewed to confirm factual evidence, quotes, or findings — and adjustments made where necessary to correct output inaccuracies. Likewise, image responses should be visually reviewed for any inaccuracies or absurdities. 

We recommend following the updated adage: “Don’t believe everything you read on the internet and what an AI bot generates based on the internet.”

Intellectual Property

It is not clear who owns AI-generated content or the user-created prompts. For now, AI models repurpose content under “fair use.” The U.S. Patent Office provided initial findings on AI copyright status in February 2023.

This ongoing conversation may impact the use of AI now, in the future, and retroactively. Caution should be taken when generating content, such as imagery, that may infringe on existent intellectual property.

Resources

AP AI Stylebook

The AP provides a primer on situating Generative AI into stories. To accurately explain the technologies to audiences, communicators must understand both the vocabulary associated with AI technologies and the concepts used in their creation. The following guidance also can serve communicators seeking to use AI models in their coverage or news production systems.

AP AI Stylebook

AI Research at Georgia Tech

Georgia Tech’s AI research is dedicated to advancing the field of artificial intelligence through cutting-edge research, innovative projects, and collaborations with industry partners.

Tech AI, the AI Hub for Georgia Tech
This document structure and framing is heavily inspired by the AI and Teaching at Duke webpage from Duke’s Learning Innovation.