IBM Research defines generative AI as deep-learning models that can generate high-quality text, images and other content based on the data they were trained on. AI models rely on an input, or a “prompt,” from a human and respond to the prompt with generated forms of media based on what they know from deep-learning and training data. They can be used to streamline processes, reduce time spent on mundane tasks, enhance ideation, and give structure to creative deliverables.

It’s important to note that generated responses are largely dependent on the data used to train the specific AI model being used, and therefore, humans should consider possible biases, inaccuracies or hallucinations that could be introduced by the AI based on its training.

Developers of AI technology are pursuing artificial general intelligence (AGI), which is a recreation of human cognitive ability through software. AGI would be capable of generating a solution when prompted with an unknown problem to solve.

• Source: IBM Research - What is generative AI?
• Source: Definition of AGI
• Source: What are AI hallucinations?

Generative AI capabilities are increasingly available in tools used to produce images, text, audio, video and code each day, illustrating the technology’s potential to foundationally transform the way marketers and communicators do their work. 

As a Carnegie R1 research university, we embrace technology that advances human discovery and progress, but we firmly believe that our people are central to those outcomes. We believe that humans are uniquely capable of true emotion, imagination and experience, and we view their role in creating media as critical to its authenticity. No matter how powerful an AI tool is, it will never be inspired by a lived Clemson Experience. For this reason, humans at Clemson University will always be the best stewards of the Clemson brand and how it is portrayed in marketing and communications contexts.

We view generative AI as a tool to complement our unique human abilities, not to serve as the origin of new ideas or the primary contributor to completed work. Generative AI is a great resource to help streamline processes and tasks, aggregate and analyze information, and ideate in the early stages of creative processes. However, we believe that human involvement should always be a step between what AI generates and what is used, published or posted.

• Directing and guiding the AI through multiple phases of generation, ensuring that what is generated is informed more by the prompter and less by the tool itself.
• Interacting with AI tools to further train them to produce outputs that align with Clemson branding, tone, pacing and standards.
• Generating overarching strategic plans for specific marketing materials and campaigns with the assistance of AI.
• Thoroughly editing, revising and modifying visuals or text generated by AI.
• Building upon or integrating generated media into original work rather than using it wholesale.
• Using reliable, authoritative sources to fact-check and verify information generated by AI to avoid deception or misinformation.
• Reference resources created and maintained by human experts in their field to confirm AI-generated facts and figures.
• Assessing the likelihood that media generated from AI could have been trained on copyrighted
  media and applying it thoughtfully and responsibly.

• Disclosing the use of AI responsibly to maintain a level of transparency in our use of AI. Furthermore, we believe that humans remain accountable for what is generated by AI from their prompts. The use of generative AI for institutional purposes comes with great responsibility. The guidance that follows is in alignment with Clemson University’s Artificial Intelligence Guidelines and builds upon it to provide marketers and creator recommendations and best practices for using these tools in their own processes.

Data Privacy and Security:

Before entering University-owned data into an AI tool, verify that the University has an executed contract, and that the data entered will not be used to train the AI model. Failure to do so is effectively disclosure of the data. For more information, see Clemson’s Data Classification Policy.

Authorship:

Normally, the author or the University owns the copyright to created work. If AI significantly assists in generating that work, the copyright ownership becomes less clear. You can mitigate this risk through substantial editing, modifying and interpreting the output from AI, therefore retaining authorship of it. In this case, the author should also disclose their use of AI as a resource.

Plagiarism:

AI, to an extent, is collating and regurgitating information based on its training, so there is not always clarity on the source of information. Consider prompting AI to cite its sources. A good failsafe is to search parts of what is generated to check if it is verbatim pulled from another source or pasting the generated content into Grammarly to check for plagiarism. For more information, see Clemson’s Research Misconduct Policy and Academic Regulations.

Content Quality:

Search engines are increasingly capable of identifying web content generated by AI. In recent years, Google’s Core Algorithm Updates and Quality Rater Guidelines have focused on prioritizing human-written and human-oriented content. As a result, websites with large volumes of AI-generated content are penalized, resulting in decreased visibility in search and decreased authority online. Clemson University communicators and web managers should author content themselves following E-E-A-T guidelines and should not programmatically generate complete web pages. (Content filling event and news widgets will continue to pull from existing databases programmatically).

Inherent Limitations

AI tools are powerful because they allow users to tap into a massive amount of information and imagery they were trained on, but even this is limited: Remember that if content has been digitized or is not on the internet, it doesn’t exist to AI. Exploring texts not available online or enjoying artwork that has not been turned into a jpeg and shared millions of times may be a better alternative to enhance or prompt creativity than an AI prompt or a search engine.

In all University communications, we seek to build trust and inspire confidence with our audience. For this reason, it is critical to be transparent and disclose the use of AI when it applies. Consider providing a footnote or concluding statement that AI was used in concert with humans to help generate and develop this content.

Adobe is an industry leader in ethical AI development. As an enterprise customer of Adobe, Clemson faculty, staff and students have access to a suite of products that are enhanced by Adobe’s AI technology. Guided by Adobe’s AI Ethics principles of accountability, responsibility and transparency, they have created standardized processes from design to development and deployment, including training, testing and a review process overseen by a diverse AI Ethics Review Board. They have also focused on developing AI that minimizes exposure to harmful and offensive content, embraces diversity and respects the creative community. Through the Content Authenticity Initiative, they are driving global industry standards for transparency.

AI tools available through Adobe Enterprise:

• Adobe Firefly.
• Integrated AI tech in Creative Cloud applications, including Photoshop, Illustrator and Acrobat.

Microsoft is committed to making sure AI systems are developed responsibly and in ways that warrant people’s trust. Through their investment in Open AI, the creators of ChatGPT, Microsoft can offer a wide range of AI products from their public version of Copilot to integrations with Microsoft’s specialized products. Clemson’s Microsoft 365 tenant provides students, faculty and staff access to Microsoft Copilot with commercial data protection when signing in with your Clemson account. Some Copilot advanced features and integrations may require additional paid licenses.

Access Microsoft Copilot:

• Microsoft Copilot
• Copilot experiences explained

The following AI tools are widely adopted by marketing and communications professionals at Clemson and beyond. This list provides a basis to understand popular AI tools and should not be interpreted as an endorsement from Clemson University.

• ChatGPT
• Dall-E
• Midjourney
• Google Gemini
• Grammarly
• Github Copilot
• Scite (Through Clemson Library)

This guidance document is a starting point for a shared understanding of the University’s perspective on AI and its acceptable uses in marketing and communication. If you have questions that aren’t addressed here, please visit Clemson University’s Artificial Intelligence Guidelines.

Clemson University’s Guidance on AI will be reviewed and updated as technology and our understanding of it evolves and grows.