UT Conference Explores Ethics Around the Breakneck Advancement of AI

At times UT-Austin‘s Good Systems Symposium felt a bit like listening to a Drake album, with conversations heavily focused on trust issues, broken promises, and community engagement (maybe not so Drake on that last one). Artificial intelli­gence experts took turns calling out the human side of a technology that, on its surface, appears to be very much not human. The emphasis on the natural world extended all the way to the decor – a string-of-pearls plant centered and two Swiss cheese plants framed the panelists.

As its technology hurtles toward uncharted territory, AI is in a precarious place – and increasingly moving beyond labs and research centers into the public eye. Autono­mous vehicles abound in Austin (hello again, Waymo) and ChatGPT is barreling ahead with designs on revolutionizing the search engine experience and beyond. Meanwhile, more than 20,000 people have signed an open letter to AI labs that poses the question, “Should we risk loss of control of our civilization?” and calls for “a pause on giant AI experiments.” The signatories include prominent tech figures such as Elon Musk and Steve Wozniak, but also many academics, including several from UT-Austin.

Chad Jenkins, a robotics professor at the University of Michigan, pointed out that while technology is moving at a fast pace in the present moment, some of the models being used aren’t particularly new. “The first neural network was McCul­loch and Pitts in 1944. Rosenblatt at Cornell developed the first object recognition system using neural networks in 1958,” he said. “This has taken decades to get to this point.” Jenkins, who delivered the conference’s keynote address, agrees with how the letter emphasizes the benefits and risks of the technology. However, he opted not to sign it. “I believe that if somebody really wants to develop AI technology, a moratorium is not going to stop them,” he said. Instead, he sees education as the key to building positive AI tools.

Jenkins, who said that computing and AI literacy should extend to K-12 education, would have likely been delighted to hear the response of the event’s youngest attendee – a boy about 15 years younger than anyone else in the lecture hall. He told the Chronicle that the most interesting thing he learned was the difference between “symbiotic robots that work with humans and autonomous robots.” If the rest of Genera­tion Alpha is already considering the implications of how humans and robots interact (and can casually use “symbiotic” in a sentence), we might be in good hands.

Indeed, that human-robot relationship is an area that UT Good Systems researchers are exploring. Ken Fleischmann, a professor in UT’s School of Information, described a project focused on designing tools that will make work safer for deaf welders. The researchers built a hand tool decked out with sensors and Internet of Things technology (humans and robots tested the tool) and also created a basic augmented reality training program for welders.

The impact of AI and robotics on labor is another area of focus for Jenkins. “I worry that if we only think about AI and robotics to do things more efficiently … that will lead to replacement of jobs, and that will lead to people not having opportunity,” he said. “But if we’re thinking more broadly, and we’re investing in areas such as education, infrastructure, health care, these longer-term investments that will have a really positive impact, then there becomes opportunity to reinstate people into the economy to address really critical needs.”

Health care is the under-the-radar field that Jenkins and others see as one of the most potentially beneficial applications of AI, particularly as the U.S. population ages. However, given health care’s history of racial discrimination, it raises the issue of whether or not humans can feed unbiased data into AI systems. Jenkins said diverse representation will be important in the advancement of the technology.

Meme Styles, a data activist, founded Measure to provide data support to Black- and brown-led organizations. During a panel on using AI to advance racial justice, she said that “racist algorithms” harmed her 21-year-old son and exacerbated his schizophrenia. She added that she hopes a less-biased algorithm can be developed, but that it’s important to consider the questions posed to find the data that eventually informs AI technology. “What is the historical and cultural context of the problem that you’re seeking to solve with your design and your solution? What communities have provided their input?”

UT’s Good Systems program seeks to address some of those sorts of questions – in part by making the Austin community more central in the process. Marc Coudert, the city of Austin’s Climate Resilience and Adaptation Program manager, said six years ago the Rockefeller Foundation organized a workshop bringing together city and county leaders to discuss climate impacts, which led to collaboration with community groups such as Go Austin/Vamos Austin. When he began speaking with community members he realized that many had taken their own initiative to start addressing challenges brought on by climate change and that there is a long history of “well-founded mistrust” of the city. “If we’re taking data, how does it connect to people’s lives?” he asked. “When we’re talking about people’s lives, how do we get them the data so they can tell us how to make a decision?” Because, in the end, it will be up to humans to ensure that AI is a tool for good.

This article appears in April 14 • 2023.