When artificial-intelligence chatbot ChatGPT released in November 2022, it quickly became the fastest growing program in history, thrusting Artificial Intelligence into the spotlight of technological debates. Researchers in the social sciences are focusing their attention on AI research and development and its impacts on human society. This includes the University of Michigan, which has several student initiatives and faculty labs dedicated to understanding and administering AI.
AI is a technical name for a computer algorithm which simulates how humans think, talk and write, but it has no idea what it is saying or why. Though they may not take over the world, AI does have the capacity to change it.
The Michigan Student Artificial Intelligence Lab is a student organization which focuses heavily on education and research surrounding artificial intelligence technology. In an interview with The Michigan Daily, LSA sophomore Asad Khan, who leads education sessions at the lab, spoke on the ways his organization keeps students engaged with AI.
“We offer a variety of different initiatives,” Khan said. “We educate students on artificial intelligence (and) we get them exposed to some research papers. For example, they replicated a research paper, a project on deep convolutional neural networks … It’s just a good way to (get involved with) applicable code as well, by applying the (technical) knowledge and just getting a general overview of the concepts.”
Khan also spoke about ChatGPT, the ubiquitous consumer-facing AI model. ChatGPT is a language model, which is a way for computers to put together sentences and phrases using statistics. Khan added how AI can be used in a variety of ways, such as identifying handwriting and photos, in addition to being used for sports analytics, autonomous vehicles and more. Though it’s become increasingly popular, Khan said ChatGPT is only one of several kinds of intelligence model.
“(ChatGPT is) a large language model, but definitely AI can be a lot more than that, for example, classifying handwritten digits,” Khan said. “That’s like computer vision, which is basically classifying images.”
While many people see these new capabilities to be the sign of an exciting new future shaped by AI, others see the new progress as an ethical and governance concern. As AI increasingly interfaces with the public, researchers and students at the University are focusing on responsibly administering the technology in policy. Khan spoke about why safety is important in AI considerations, and why it may be undervalued.
“Basically, you have this market that really emphasizes the progression of AI and people are afraid of this because they’re valuing progression and growth over safety,” Khan said. “So a lot of times if you value progression and growth and stuff that will increase your sales and your profits, you’re not really going to be worried about safety.”
U-M alum Jakub Kraus was a student leader at the Michigan AI Safety Initiative, a student organization that focuses on safety and policy issues related to AI, last year. In an interview with The Daily, Kraus elaborated on why governance concerns are particularly important in a modern AI environment.
“Now, (AI progressing) is getting faster and faster, but the rate at which humans can respond to this change and build correct social structures and laws and … technical and engineering solutions for safety; I don’t see that progressing as fast,” Kraus said.
To address these gaps in AI research, Kraus said AI governance has become a focus of some U-M research initiatives, where some professors are focusing on specific ways to make AI models produce honest, appropriate content.
“From the AI safety point of view, I think there are some professors working on responsible AI and fairness and trustworthiness,” Kraus said. “I’m excited about that … so there’s a lot of focus on risks that are right in front of us.”
Nikola Banovic, professor of computer science at the University, specializes in human-computer interaction, a field at the intersection of computer science, design and social science. Banovic spoke with The Daily on how people connect with computers, and how they use computers to connect with each other.
“My research is actually in human-computer interaction,” Banovic said. “Just like any other HCI researcher, I’m interested in understanding people, understanding their goals, understanding their needs, understanding their desires, understanding different tasks to be performed to kind of reach those goals or satisfy those needs.”
Banovic said he is particularly interested in ethical concerns with AI, which is one of the core issues in HCI. He studies how human systems can interact properly with digital systems. One of the issues he studies is trust, or how AI is perceived as honest, consistent and reliable.
“I think it’s important to note that trust itself is complex,” Banovic said. “And when you’re looking at definitions of what makes an AI trustworthy, a lot of those definitions come from computer science … (People will ask) ‘Is it accurate? How competent is the AI in any task that it’s supposed to do? Is it secure? Does it preserve privacy?’ So it’s more like, how (does it protect) from those external threats? “
Banovic emphasized both the importance and difficulty of creating policy to address AI safety concerns. He said though regulatory policy is the most important societal aspect of AI, implementing regulation is difficult because there are many different necessary stakeholders who need to collaborate to create effective policy.
“The most important aspects are going to be regulation and policy,” Banovic said. “There’s absolutely no question about it. We are already seeing it. We’re seeing a lot of proposals both in the U.S. (and in the European Union) … The challenge here is that there are so many people from many different disciplines that need to come together to be able to actually craft regulations and policies.“
Daily Staff Reporter Amer Goel can be reached at amergoel@umich.edu.