Guest Editor Letter

February 2023

Best Way to Tackle Generative AI: Focus on the Humans

By: Badri Roysam, University of Houston

Dear ECEDHA Members, Industry Partners, and Colleagues,

Hope your Spring semester is off to a good start, and some much-needed calm has returned to your department. As always, it is now time to start tackling the world’s biggest challenges.

One “weird mixture of opportunity and challenge” that came to the fore recently is Generative Artificial Intelligence (G-AI), as exemplified by ChatGPT and DALL-E, and we discussed them in our December issue of the ECEDHA Source. Much has been said, and a lot has been personally experienced by most of us, on this topic. There are authentic reasons to be excited about this technology, and equally authentic reasons to be concerned about it. We have already witnessed disruptive changes wrought by ChatGPT in many industry sectors. Notably, realtors are now routinely using it to generate their home sales flyers.

The big question is – how do we deal with these new technologies in our classrooms and research labs? Importantly, how do these technologies impact the employability and career prospects of our graduates? I posed this question to my department’s thoughtful and helpful Industry Advisory Board (https://www.ee.uh.edu/people/industry-advisory-board ). We had a spirited discussion, and our conversations wandered far and wide. In the end, they converged on a surprisingly simple conclusion.

“The best way to deal with technologies like Generative AI is to focus on the humans.”

Please allow me to elaborate. Given that we are in the business of educating people, both ourselves and our students, it is natural for us to focus on faculty, staff, and students. We start by recognizing that G-AI is, in the end, a tool, and an imperfect one at that as we have seen by now. Next, the inner workings of this tool are no longer a mystery, and there are now many well-written articles and video lectures available. In a nutshell, it is a sentence completion tool that is trained on a massive internet dataset and is aided by multiple peripheral software “filtering” systems that provide the human interface. Given this knowledge, and of course, in 20/20 hindsight, it is clear that ChatGPT cannot do Laplace transforms or Physics. Unlike traditional AI systems like Wolfram Alpha, ChatGPT clearly lacks any mathematical structure – it does not have any axioms or theorems built in. In other words, it is basically a sophisticated language processor, no more, no less. This line of reasoning defines our first strategy:

Strategy #1: Learn the inner workings of Generative AI, and teach that to our students

Once students have a basic understanding of how Generative AI works, they will no longer see it as a mysterious black box, and find appropriate uses for it. Importantly, they will learn to distrust inappropriate uses. For instance, they will learn that ChatGPT can generate a grammatically clean summary of a topic, but any scientific/mathematical details cannot be trusted. To make this real, we need to develop the basic rules of responsibility.

Strategy #2: Hold the humans (faculty, students, staff) responsible for their products, whether or not they use Generative AI

This is true for almost all tools. Whether a student uses a calculator, MATLAB, Wolfram Alpha, or a Generative AI (current or future), they must understand that they are ultimately responsible for their products (theses, papers, talks, patents, …etc.). If a bridge or a space-faring rocket is designed by an engineer, they must be responsible for their correct functioning, regardless of what tools were used in the design phase.

Strategy #3: Teach critical thinking and “sanity checking” skills

Given that engineering is the application of mathematics and the sciences to solving practical problems, this is an especially opportune time to teach critical thinking and sanity-checking skills to our students. They should learn the old-fashioned art of running a quick and approximate back-of-the-envelope calculation designed to make sure that a machine-generated solution is reasonable, and importantly, does not violate physical laws. Beyond physical laws, we must also consider societal norms and rules.

Strategy #4: Double down on teaching Ethics, Intellectual Property Laws, Integrity, Liability, Attribution of Sources, Emotional Intelligence, Citizenship, Reputation Building, etc.

The senior design experience may, perhaps, be the best place to describe these most basic human values and skills that, when imbibed effectively, result in world-class engineers whose products excel in our complex world, and avoid reputational risks. All of these factors live in the minds of students and we need reliable methods for making sure they are present. This brings us to our fifth strategy.

Strategy #5: Double down on viva voce – the old-fashioned, but still effective evaluation of a student’s knowledge base by oral examination/discussion.

Just for fun, I looked up Wikipedia, and here its what it had to say:

“The oral exam (also oral test or viva voce; Rigorosum in German-speaking nations) is a practice in many schools and disciplines in which an examiner poses questions to the student in spoken form. The student has to answer the question in such a way as to demonstrate sufficient knowledge of the subject to pass the exam. The oral exam also helps reduce (although it does not eliminate) the risk of granting a degree to a candidate who has had the thesis or dissertation ghostwritten by an expert.”

I welcome your thoughts and suggestions on this topic.

We would love to feature more ECE conversations in future issues of this newsletter. Here are some of the topics of current interest:

The many impacts of AI on ECE education and research;
Strategies and Experiences in meeting the semiconductor re-shoring challenge;
The growing footprint of healthcare-related research in ECE departments;
Growing the workforce for the future electronics-intensive defense industry;
The new ECE Chair experience, the ups, and the downs;
Strategies and best practices for developing future ECE faculty;

As always, we welcome short articles from ECE department heads & faculty on topics of broad interest to our community. If you have any thoughts on such a topic of general, let us plan to meet and talk in New Mexico.

Wishing you a wonderful 2023,

Peace,

Badri Roysam, D.Sc., Fellow IEEE, Fellow AIMBE
Hugh Roy and Lillie Cranz Cullen University Professor
Chair, Electrical & Computer Engineering Department
University of Houston
Houston, Texas 77204-4005
Phone: 713-743-1773
Email: broysam@uh.edu