AI is Here to Stay: A Q&A with Jon Oakes

How can students, faculty and the public engage with ChatGPT in productive ways? Learn more about the new AI platform on April 20 at the library.

Learn the tips, tricks and traps with the SJSU King Library

In the dark King Library Experiential Virtual Reality (KLEVR) Lab on the fourth floor of San José State University’s Dr. Martin Luther King, Jr. Library, the only light is emanating from cracks in the curtain and the glow of the computer screen. Jon Oakes, the technology coordinator for the SJSU King Library, clicks to expand the ChatGPT window. On the left side of his screen is a stack of virtual chats, some experiments while other chat messages are task-oriented. 

What is ChatGPT? Ask ChatGPT and it will respond that it is a conversational artificial intelligence model developed by OpenAI. It is a state-of-the-art language model that has been trained on a massive amount of text data from the Internet and has the ability to generate human-like responses to a wide range of questions. ChatGPT is designed to provide answers to questions, summarize information and generate creative text. It has applications in customer service, virtual assistance, and many other areas where quick and accurate information is needed.

However, according to Oakes, “it makes connections among ideas the way our brain does, or similarly to how our memory works — but it doesn’t think or process all the time. It’s only instantiated based on a prompt that you give it. The prompt basically frames the question and then applies it to the model and creates these vectors of math through the model with a trail of tokens that are words and letters that actually mean something relevant to the prompt it was given. It doesn’t actually think for itself, but it can respond in a way that is very convincing that it does.”

Oakes quickly types in ChatGPT while speaking his prompt out loud. He waits for the response generated from GPT-3 and extra layers of fine-tuning and programming by OpenAI. It’s pretty spot on. However, Oakes says that although AI is a great tool to explore information, the next step for higher ed is to teach students to use it appropriately to enhance their education experience. 

“It can be used to be a term paper machine, which is easy to defeat,” says Oakes. “Five minutes with faculty members, I can make them very comfortable in knowing that no student will likely get away with an AI-generated term paper if they know what to look for.” 

That is Oakes’ goal for planning an AI symposium on April 20, hosted by the SJSU King Library in collaboration with the Digital Humanities Center. Register for the symposium to learn more the tips, tricks, and traps of AI. 

“I want a friendly and non-techy look at the AI tools available for use, how they can be used, and how they can be abused and some ideas of what the future holds,” he said. “This is going to happen very, very fast and be very transformative — [it will be] good and bad over the next few years and there is no stopping it. There is no getting off this ride and you are going to have to hold on and try to manage it. If you do, you are going to be way ahead of the game. 

“The SJSU King Library plans to train our librarians and staff to understand these issues and be a resource for our campus to use these tools and learn strategies to incorporate them into their research and dissuade students from using them improperly.

“I don’t have all the answers but the important theme is that we fear and loathe what we don’t understand, so get in here and understand it,” urged Oakes.

Jon Oakes, technology coordinator for the SJSU King Library, weighs in on the ChatGPT debate. Photo by Maleek Diaz, Political Science ’25, SJSU King Library Marketing Team.

Learn more about the transformative potential of AI in the following Q&A with Jon Oakes. His answers have been edited for clarity and length.

What are the criticisms of ChatGPT? 

Jon Oakes (JO): ChatGPT is a proprietary system developed by OpenAI. We don’t really know exactly what went into it, what was kept out, what’s being filtered or promoted in responses. In the effort to make the model safe, OpenAI has also inserted bias, which can have unintended consequences. Also, ChatGPT is NOT a free service for long. It’s going to cost real money and be a barrier for some students to use it.  That will create a very stark ‘digital divide’ problem the university will need to address.

How can faculty use AI in the classroom? 

JO: Assuming ChatGPT continues to be available for faculty to use, it’s a fantastic summary generator. You can cut/paste large bodies of text and get pretty accurate summaries and present them in different styles or arguments from different points of view. Nothing is perfect about it, but as a starting point for discussion, it’s amazing. 

How can students get fooled by AI? 

JO: ChatGPT will try to give you what you ask for without too much questioning. It’s also a convincing liar. If you ask it for peer-reviewed articles as sources, it’ll give you many of them.  Some might be real, some might be fake and you won’t know until you do the research manually. ChatGPT will create summaries and quotes from experts that don’t exist with apparent glee and confidence. And while it’s gotten better, its math skills and ability to engage in physical sciences where actual math is needed is atrocious — but it’ll give you a very convincing [and completely wrong] answer. That’s as of Feb. 2023. It might be very different in just a few months! It’s improving fast!

How will AI affect librarianship? 

JO: AI will not replace any human experts at the top of their skill set, but we all engage in tasks within our skill domain that are basic. Librarians currently spend time introducing students to basic research skills and providing very low value information, but they also engage in very deep and nuanced research with meticulous sourcing and attention to detail. Hopefully, they will be able to offload some of the work that doesn’t maximize their skills and be able to have more time to engage in deeper research and help. Also, GPT-3 doesn’t know anything it wasn’t trained on and lots of important information still is only between the pages of books in an archive or scanned online and not in GPT’s memory. Again, this gap will close over time and not be an issue for many years to come.

Register for the King Library AI symposium on April 20th from 1-4 p.m. to be a part of the discussion.