Calbright faculty member, and chair of the faculty senate, Michael Stewart, is a recognized leader in the field of machine learning and Artificial Intelligence (AI). He works in the industry, teaches courses on AI development, and is one of just six faculty members to sit on the California Community College Chancellor’s AI Conversations Working Group, which helps determine how AI is integrated into the CCC system.
But speaking to his colleagues at Calbright’s Distinguished Faculty Lecture Series last week, the inaugural lecture in a series that honors faculty who have made exceptional contributions to their field, Stewart admitted that AI is a relatively recent addition to his lifelong love of computers.
“I had a computer technology class in the 11th grade, and I used the computer to draw a circle,” he said. “And as soon as I saw that circle, I was hooked.”
But he didn’t do anything with AI until 2013, when he created an app for the iPhone as part of his Masters thesis in Instructional Design. The app was designed to let people struggling with addiction attend recovery meetings virtually, and he needed to have it automatically create avatars of its users so that they could remain anonymous.
But it wasn’t until a few years later, when he was invited by fellow Calbright faculty member Elizabeth Biddlcome to help her at a workshop she was giving at DefCon (a hacker convention), that he opened his eyes to the importance of this new technology. “The name of the workshop was Machine Learning for Nubes. An hour into the workshop, I realized I was the nube,” he said.
“They were teaching people how to get machines to hack themselves,” he told the crowd, “and at that point I knew I was behind the curve in technology. I knew I needed to get on this train and sit in the front before it passed me by. From that point on it was all about machine learning for me, which is a subset of AI. AI is more than machine learning, but machine learning is the engine that powers AI.”
Now he’s on the cutting edge, and he and Calbright faculty member Elizabeth Biddlecome, who interviewed Stewart as part of the presentation, said that this is proof that the technology is open to anyone who wants to put time and effort in. With the right training, anyone can use it at a professional level.
Caution Ahead
Especially important, Stewart said, is understanding what AI can do well, what it can’t, and how to interact with it.
It will never replace teacher interactions with students, for example. “AI lives in the past,” he said, “and it lies. It also has no emotional intelligence. That will crush students if we let it.”
Similarly, Stewart said, overreliance on AI can have negative effects. “For me personally, the smartphone has made me less smart,” he noted. “If I lose my phone I can’t call my kids, I don’t even know their numbers! I can’t call them! That’s AI using me as an assistant. Technology is great, but we can’t rely on it to think for us.”
And, very importantly, never input sensitive information into a chatbot, like ChatGPT or Claude, that isn’t your own closed system. “All our input is training all of these products.”
“A Tool, Not A Crutch”
The key to using AI well, Stewart said, is to understand what you’re trying to ask it to do and being able to evaluate its output. “We should use it as a tool, not a crutch,” he said. “I think that’s my catchphrase.”
People can use AI effectively to massively increase their productivity in areas that they’re domain experts in, and to automate rote tasks. A composer can effectively use AI to make good music, while a computer programmer can use AI to make good code … but a computer programmer won’t be able to use AI to make better than mediocre music, and a composer won’t be able to use AI to make better than mediocre code.
As a teacher of Information Technology, Stewart uses AI to create a dazzling array of ways for his students to learn, like using automation to turn course materials into podcasts that students can listen to. It works in part because the AI technology is so amazing, but also because Stewart, as a subject matter expert, can ensure that the information the AI presents is accurate.
Stewart concluded his talk by giving a workshop on the importance of effective prompting. “We never want to accept the first response that comes out of an LLM (Large Language Model),” he said. “We want to feed it back and keep making it better. Because it’s going to give us its easiest answer the first time. Stick a prompt in multiple times, you will get something different multiple times. As users, our job is to make sure it keeps getting better.”
