CIO Roundtable: Bans aren’t the answer: Understanding the implications of AI writing/ChatGPT in Higher Ed

How should you react to ChatGPT?

This generative AI tool has the ability to generate essays, reports, and even poetry that can easily pass for student-written text.

In a very short time, ChatGPT has emerged with opportunities and risks

• What does this mean for academic integrity in higher ed classrooms?
• Is this similar to the "calculators in the classroom" debate?
• How useful are tools to identify AI-written text such as “Classifier” by OpenAI?

Banning technology is not the answer

Join our experts for practical guidance on this emerging issue:

• Hear directly from leading experts and peers on the best approach to managing AI in the classroom.
• Go beyond reactivism and focus on understanding what AI writing tools such as ChatGPT, as well as identifiers such as Classifier, can and cannot do.
• Explore the benefits of generative AI for higher ed learning, such as driving inclusiveness, technology gaps for underserved communities, and enhancements to student experience.

Why is ChatGPT special?

We’ve had Siri for years. Why the hype?

Unlike previous iterations of the GPT model (the most recent being 3.0), ChatGPT (3.5) is specifically designed to serve a chatbot function. Compared to similar technology, ChatGPT is adept at engaging in dialogue with its users and can respond to feedback, seek clarification, and iterate on its answers based on how users respond to them. Try a complicated, multistep request with Siri or Alexa and you’ll see why ChatGPT is getting as much attention as it is. Other services, like Jasper or ServiceNow’s Virtual Agent, have more specific use cases (content creation and technical support workflows respectively) and are generally focused less on the conversational aspect.

GPT 3.0, the model that undergirds ChatGPT, has 175 billion parameters (according to one source, though the tool itself claims that the number is not public), making it an extremely versatile tool. Users can have it write poems about fudge, hypothesize what a world where commercial airliners were replaced by dirigibles would look like, or even write marketing copy for a report on the future of generative AI — all of which make it attractive for curious internet users.

Like its sister model, Dall-E 2, ChatGPT is generative AI, and unlike your standard chatbot, which excels at executing basic workflows, ChatGPT can create new content based on inputs. There have been other examples of this type of AI in the past, but few are as accessible or impressive as ChatGPT.

“GPT, or Generative Pretrained Transformer, is a type of language model that uses deep learning to generate text. It is trained on a large amount of data and can generate text that is similar to human-written text in terms of style, coherence, and grammatical correctness. GPT can be used for a variety of tasks, such as language translation, summarization, and text generation. It is particularly useful for generating long pieces of text, such as articles or books, because it can maintain coherence and consistency over extended periods of time.” ChatGPT, discussing itself

Recommendations

How can you best incorporate ChatGPT and other generative AI products into your service catalog?

1. ChatGPT and other generative AI solutions are tools — nothing more. Like any piece of software, computer hardware, or any other kind of machinery, there are things that this technology is especially good at and others that it is not especially useful for. Understand your business processes and highlight opportunities to reduce friction, increase the quality of the service experience, and drive efficiency.
2. Though it may be appealing to dive right into the AI end game, start with augmentation. Generative AI is an incredible technology, but it’s not self-sufficient. It still needs guidance and feedback from human curators. As you review business processes, look for those that generative AI can supplement, perhaps by producing first drafts of press releases to be reviewed by a human editor, or by writing code that will be reviewed by a senior developer before it is pushed to production. As you augment your production capabilities, you should also expect to incur the cost of augmenting your supervisory capabilities.
3. Talk to your lawyer. Chatbots that manage workflows aren’t complicated, but a bot that will interact with users and customers and produce content could expose you to legal risk, perhaps by lifting content from other sources, or — in the case of generative AI — producing artificial nude pictures of real people, which could be illegal. With new frontiers come new potential problems, so it’s a good idea to ask these questions.

What’s the next step for IT?

Refine your use case for generative AI.

Review capability map for high-value processes

1. Identify your capabilities that support key business processes.
2. Build a heatmap, highlighting areas where a generative AI solution could be effective (repeatable, generative, valuable — framework).

Conduct a basic cost-benefit analysis for the tech

1. Of the most appealing processes/capabilities, select one or a small number that are most suitable for generative AI intervention.
2. Outline a high-level business case — how would generative AI reduce manual work? What would its output look like? What would the expected benefit be? What are the drawbacks? How much would it cost?

Explore the vendor landscape

1. Identify the table stakes — what does your AI solution need to do? What solutions does it need to integrate with?
2. Outline a high-level business case — how would generative AI reduce manual work? What would its output look like? What would the expected benefit be? What are the drawbacks? How much would it cost?

Why Should I Attend?

• Connect with a global network of IT leaders in the Higher Education industry.
• Review options available on the market. Many tools exist but are not "products." ChatGPT is available as a research preview, while other tools, like Jasper, Rytr, and the various ITSM chatbots, can all be purchased for enterprise use.
• Compare your requirements to your options, select, and begin piloting.