05/25/2023 | News release | Distributed by Public on 05/25/2023 13:42
Public artificial intelligence programs such as ChatGPT are being embraced by faculty across New Jersey Institute of Technology as the latest classroom tool, just like the introductions of videoconferencing, laptops, computer-aided drafting and pocket calculators that came before.
AI itself is an established technology that emerged from research labs in the 1950s - here's a Newark College of Engineering chemical engineering dissertation using AI techniques in 1974. AI also makes periodic dalliances into consumer life, such as computer chess games in the 1980s and IBM's Jeopardy-dominating Watson in the 2010s.
But not until the present has AI, and specifically large language models, been so accessible with nothing but an ordinary web browser and no special skills required. Asked to compare itself to previous systems, "ChatGPT differs from established AI tools in its ability to generate human-like responses to natural language inputs with a high degree of coherence and relevance," ChatGPT stated.
Now, "What we need to do is learn how - it's going to be around, it's not going anywhere - so how do we best help our students learn to use it wisely? How do we as educators use it and integrate it into what we are doing," said NJIT Dean of Students, Marybeth Boger, at a university town hall event. "If we make it into this big monster, I can tell you, [students] are going to use it in all the wrong ways. How are we going to help students learn to use it in a wise manner? That's the goal."
"Any technology tool must be used for the right reasons and objectives," Interim Provost Atam Dhawan added. "If it is used appropriately and responsibly, it could help in learning and research positively as an incredible resource. But we need to be careful about the use of misleading or inaccurate information and biases in the process."
They agreed that it's sensible for the NJIT Information Systems and Technology Office to take early steps in addressing such concerns. Justin Krawiec and Nicole Bosca, who specialize in digital and online learning, are leading the charge. "We have started to create guidelines and recommendations for how instructors can use ChatGPT in the classroom," Krawiec said. The document explains how to reference AI usage in syllabi, adjust evaluation strategies, use different assessment tools, actively involve AI in student lessons and the status of detection software.
She and Bosca recently presented on AI to the Faculty Senate. Their message was that AI is already widespread as a helpful robotic brain on campus, such as in plagiarism detection software, in design software like Nvidia Omniverse and in automated help desk software, all of which are popular here.
"AI has been here, it is here and will be here. It's just getting more advanced and we're seeing that in a number of different places. We're looking into ways that we can leverage that technology and prepare students for the workforce that they're going to be in, rather than limiting their access to it," Krawiec said. "We're training students for many jobs that might not even exist at this point."
Programming superhuman monsters
The recommendations include perspectives that some professors throughout NJIT's academic colleges are already implementing to help students learn and to keep pedagogy fresh and interesting.
"The code it was writing was as good as what I write," said Keith Williams, university lecturer, in the informatics department of Ying Wu College of Computing. "It makes me a programming superhuman monster. It allows me to program so fast and explode my productivity. I can do a full week's worth of work in a day. It's like comparing somebody using a shovel to somebody who has a backhoe, and you need to dig a hole.
"I got this idea where I was going to make a simulation of working in a web development company. I started using ChatGPT to write all the content, to explain it to students … They're going to have to listen to a client interview and extract the information to do the assignment," he said. The concept would have been tedious to build before ChatGPT existed, serving as a robotic assistant.
Williams thereby teaches students about documenting their code and writing client acceptance criteria, since he believes the actual task of coding will be mostly automated in the near future, except for specialists who develop the AI itself. He said most software developers will spend their time on bigger-picture tasks such as application integrations, rather than the details of coding.
Williams wants to test his ideas for AI-led classrooms in Africa. "I am forming a foundation, a non-profit, to help transition teachers into using AI effectively in their classrooms, to build tools for them and to build a community to discuss the issues. Because we're going to have to make this transition, fast. We're not going to be able to wait five years." His organization is called MyWebClass.org and he intends to build pilot projects with educational colleagues in Zambia.
In the College of Science and Liberal Arts, Assistant Professor Megan O'Neill directs the first-year writing program. She uses ChatGPT to help her teach the concept of genre. Students can pick a topic and instruct the software to write about it in different ways, whether it's a haiku, non-fiction or Shakespearean tragedy.
AI is similarly catching on in Hillier College of Architecture and Design. "We are just starting to experiment with various AI tools, particularly in the context of design - visual and storytelling," Associate Professor Andrzej Zarzycki said. "Over the last couple of months, I have spent a significant amount of time experimenting and trying to fold these tools into curricular ideas. In the context of design and architecture, it mostly focuses on the 2D illustrations and visuals. They often can be combined with text based tools like ChatGPT to help students with developing narratives in support of graphic novels.
"There is also interest in hybrid use of AI tools in what is called outpainting and inpainting, where AI tools expand or fill parts of the original image with content that understands the overall context and additionally follows designer's prompts. Beyond visuals, there is an interest in synthesizing soundtracks, motion capture and facial animation of 3D characters to name a couple of directions. Some of these are being tried by a number of colleagues within HCAD to develop strategies for AI-driven and creativity-focused curricula," Zarzycki added.
Students may feel hesitancy, "As visual authoring AI tools may be seen as taking over the agency in image creation. This is why it is important to develop curricular strategies to use AI as enabling tools and not the endgame of design," he continued.
Newark College of Engineering faculty are using ChatGPT to show students how computer-generated ideas compare to human ones. Assistant Professor Joshua Young uses ChatGPT in the study of chemical and materials engineering. "You can go into 'GPT and say, 'Pretend you're a data scientist and write me a random forest model that predicts Y from features X, and it will give you Python code'," Young found.
Young has his students compare their own work to the automated code, and then look for strengths and weaknesses. Students know that code itself is only a small part of data science, compared to context such as deciding what to measure and determining the input parameters. ChatGPT doesn't know that.
Meanwhile, in Martin Tuchman School of Management, there's MGMT-416: AI for Business Decisions. Students were already learning how to use AI for customer service and data analysis, before ChatGPT came along in fall 2022. Now, they're seeing if machines can develop answers that are as good as human-developed versions, explained senior university lecturer Wayne Fox. Automated chatbots are already common on corporate websites, but the addition of large language models may truly convince customers that they're talking to a live person - though perhaps a person who lacks the best judgment, Fox observed.
Examining AI itself
Fox's course is just one where artificial intelligence itself is the star. In the College of Science and Liberal Arts, there's HSS-404: Philosophy Seminar in AI Ethics, HIST-390: Historical Problems Through Film: Artificial Intelligence in the Modern World, STS-365: Artificial Intelligence and Ethics, and STS-373: AI and the Human Mind. Additional CSLA courses that cover AI topics include PSY-361: Found of Cyberpsychology, STS-351: Minds and Machines, and STS-492: Technology and the Future of Work.
These courses are all in addition to the offerings at Ying Wu College of Computing, which has CS-370: Introduction to Artificial Intelligence for undergraduates along with an entire master's degree in AI software development that is now available. The college is also home to the Center for Artificial Intelligence Research led by Distinguished Professor Grace Wang, where experts study AI development under the hood.
Junior computer science major Carlos Pineda serves on the university's Highlander Integrity Council, and in that role he has seen both sides of the ChatGPT debate. He recognizes its value as a teaching tool and its detriment as a shortcut that some of his peers might exploit.
"It will regurgitate what might possibly help you. It's a tool. It is no different than having a textbook next to you, except it's not a valid textbook. It's whatever is on the Internet," Pineda said. He tested it with a computer science problem, "And it literally just gave me Stack Overflow code," referring to the popular coding assistance website. "It's not special, it's just a faster way of searching for an answer."
Pineda wants to work in cybersecurity and said he does anticipate AI being important to that field, because it can automate labor-intensive processes, in turn helping to balance the scales against hackers.
O'Neill, the writing professor who's also an advisor to the NJIT Institute for Teaching Excellence, added further perspective. "What I really want to impart to both students and faculty is that, yes, these large language models, they are amazing but they still have real limitations," she said. "The idea that education is over or all white-collar jobs are going to be gone, is a reactionary response. What ChatGPT really tells us as educators is that it's a really nice moment to stop, look at your materials, and ask yourself: If a computer can create what I'm asking students to do, then maybe I need to be asking students to do new things."