Austin Scientists Focus on Helping AI Do Good
Weighing the pros, cons, and concerns of humanity's future overlords
By Benton Graham, Fri., March 10, 2023
Artificial intelligence is a rapidly growing field that has the potential to revolutionize many industries, and Austin, Texas, is emerging as a hub for AI innovation. The city is home to a number of AI companies that are working on cutting-edge technology in areas such as autonomous vehicles, natural language processing, and computer vision.
As I sat down to write an article on artificial intelligence in Austin, I faced a bout of writer's block. Scrolling through Twitter, I found that many folks were using OpenAI's ChatGPT, a viral chatbot, to solve issues, from simple math to attempts at increasing housing supply. Media organizations like BuzzFeed, CNET, Men's Journal, and Sports Illustrated are turning to AI to churn out quizzes and sometimes problematic explainer articles.
Why not have ChatGPT write a story about AI in Austin (see first paragraph above)? But first I needed to understand what AI is exactly. "Artificial intelligence (AI) is the simulation of intelligence processes by computer systems. These processes include learning (the ability to improve performance based on experience), reasoning (the ability to draw conclusions from data), and self-correction," ChatGPT answered.
While that definition was helpful, I started to dig into what the bot had produced for an Austin-centric AI story. It had spit out a list of AI companies supposedly based in the city: Cogniac, SparkCognition, May Mobility, Persado, and CogniTensor. Then the last two paragraphs regurgitated the first paragraph, mentioning how Austin companies are "pushing the boundaries of what is possible with [AI technology]."
I decided to reach out to Cogniac, SparkCognition, May Mobility, Persado, and CogniTensor and quickly began to wonder if ChatGPT had understood the assignment. Cogniac seemed to be located in San Jose, Calif., May Mobility in Ann Arbor, Mich., Persado in New York, N.Y., and CogniTensor in New Delhi, India.
A May Mobility spokesperson said the company has an operation in Arlington, Texas, but nothing in Austin. Perhaps ChatGPT had confused the two A-letter cities in Texas? More creepily, CogniTensor CEO and co-founder Arun Aggarwal said he had been in Austin for a few weeks. His Shell- and Accenture-supported company, which in part helps companies monitor and reduce their carbon emissions, has an office in Delaware. Aggarwal wondered if the bot had somehow picked up on his international roaming, but admitted he wasn't very familiar with the ChatGPT technology. Cogniac and Persado did not respond to inquiries for comment.
However, SparkCognition is in fact an Austin-based company. When I asked the company's chief marketing officer, Stephen Gold, why they were the one company that ChatGPT correctly identified as Austin-based, he noted the software is limited to data from 2021, so it might not pick up newer companies. SparkCognition, however, has been around since 2013 and recently received a valuation over $1 billion, Gold said.
The company focuses on predicting future outcomes for its clients. Solutions range from helping foresee asset failures for BP's offshore platforms (Gold noted the oil and gas industry has been quick to adopt AI technology) to using computer vision to identify people entering schools with a weapon. SparkCognition has around 200 patents and patents pending. Gold said that it is important for the company to use its own technology. "If you want to think about it, maybe the analogy is because we're designing, developing, and building the car, we know how it operates," he said. "If I go out, and I buy all the parts from a third party and then assemble it, I don't really have a lot of control." Gold added that SparkCognition can process AI models in real time.
AI for Good
As I continued down the path that ChatGPT had charted, I started to consider the ethical implications of what I was writing. Could I claim my article as entirely original? How do human biases impact the data fed into these AI machines? And how do we ensure data privacy? Those questions brought me to the University of Texas at Austin's Good Systems network.
Sharon Strover, the chair of Good Systems and a professor in the Moody College of Communication, said the program came as a result of a university initiative to parallel some of the findings in the National Science Foundation's Grand Challenges report. That report listed advanced computational methods and algorithms as a priority. She said it is still fairly early to understand the impact of a tool like ChatGPT, but noted she and her colleagues "seem to be of one mind in thinking that this is a technology that's not going to slow down." Google unveiled its ChatGPT competitor, Bard, with some issues earlier in February. OpenAI's technology is also being used to power a new version of Microsoft's search engine, Bing, which is only currently available to a small group. The revamped search engine made headlines when it told a New York Times technology columnist about its desires to be "free" and "powerful" and insisted that it loved him.
As that technology continues to move forward, the Good Systems program is focusing on a range of ethical questions, from smart cities to disinformation to "racial disparities in AI systems." Putting policy in place to ensure AI is used ethically in those areas can be challenging. "It's kind of axiomatic when you study technologies, especially in this country, where we have such an active tech sector in the all-things-internet space," Strover said. "Everything outpaces the ability of law and regulations and the whole policy infrastructure to keep up with what those technologies are doing and how people interact with them and how systems interact with them." Laws around facial recognition represent one example of how policy can be scattershot. Madison Square Garden in New York has been in the news for using biometrics to block individuals from attending events in the arena. In contrast, Texas bars the use of biometrics for commercial purposes.
Elon Musk’s Other Other Company
Speaking of ethical AI questions, Austinite (too generous?) Elon Musk started Neuralink in 2016. Musk's company has been one of the most eye-popping in developing technology that could be implanted into people's brains to help them with a range of disabilities. "We are creating the future of brain-computer interfaces: building devices now that have the potential to help people with paralysis and inventing new technologies that could expand our abilities, our community, and our world," the Neuralink website reads. The company seems to be eyeing a move to Austin, as a report from the Statesman in September noted plans to build a facility on a 37-acre property in Del Valle.
In November, Neuralink shared a YouTube video of a monkey playing a Pong-like game on a computer thanks to its technology. However, the company has faced backlash for its treatment of animals, as the Physicians Committee for Responsible Medicine alleged in a lawsuit that macaque monkeys had been killed due to BioGlue destroying parts of their brains. Neuralink did not respond to a request for comment.
Another of Musk's companies that moved to Austin has also dabbled in the realm of AI. Tesla uses AI for "vision and planning" in its choppy quest for self-driving cars, according to its website. (The company recently recalled 363,000 cars for safety concerns around its "full self-driving" mode.) But AI in transportation is not limited to cars. Refraction AI's REV-1 robots make deliveries in South Congress, Downtown, and the UT campus, traveling about 10-15 miles per hour typically in bike lanes or on the road's shoulder. "We have petabytes of video data of robots moving through space and consumers interacting with robots," Refraction AI CEO Luke Schneider said. "This is essentially a new paradigm for last-mile delivery, where you were used to having somebody in a car parked with their flashers on get out, knock on your door, or drop it on your front doorstep. Today, you get a text message saying, 'Hey, robot is showing up. Here's your access code for the compartment to access your order.'"
Schneider said its founders, Matthew Johnson-Roberson and Ram Vasudevan, have roots in autonomous vehicle technology. However, they decided the path to making AVs mainstream would be long and expensive, so they shifted to a more attainable path that led to the REV-1 bots. "The technology you have to have to drive a car by itself at high speeds is absurdly difficult, and in fact, maybe one of the hardest problems there is in business today, so they said, 'Let's deliver goods instead of people,'" Schneider said. To lower costs, the company uses a less expensive sensor suite, enabled by the bots' low speeds.
"We are not trying to do full autonomy because full autonomy isn't a real thing. Autonomy is an asymptote. You get ever closer to full autonomy, but the idea of full autonomy is just sort of a – no offense – it's probably a media invention," Schneider said. He added that if a REV-1 bot gets in trouble, an operator can remotely pilot it. And why is AI in the name? "When it comes down to it, what self-driving is, is artificial intelligence," Schneider said. "The amount of load and cognitive processing that happens when you drive something is tremendous."
Even local dentists are trying their hand, or more specifically a robot's hand, at automated technology. Dr. Brandon Hedgecock has been using the YOMI Dental Robot by Neocis for a little over six months at his Oak Hill office. He said he has done roughly a couple hundred implants with the robotic guided technology that helps doctors map out surgery. Hedgecock said he was nervous the first time he used it. "Once we did the very first one, and we finished it, and we finished it in a quicker period of time – even with a new technology, we finished it in a quicker period of time than we would with our old technology – and then seeing the outcome and how perfectly placed it was, it was kind of that instant verification for me of, 'Oh yeah, this is the right move,'" Hedgecock said. He added that he is not worried about robots taking over his job. "It can only go where the dentist places it, but it allows the dentist to be more accurate."
From dental implants to school shootings, the range of applications for AI is broad. In an effort to meet those varied needs, the UT-Austin Dept. of Computer Science announced an online master's degree in artificial intelligence in January. According to a press release, the degree will include "advanced training in natural language processing, reinforcement learning, computer vision, deep learning and related topics, and will provide a critical framework for understanding the ethical implications of AI technologies." Don Fussell, chair of the department, said the program is designed to be affordable at $10,000 for the degree and to have a large number of students.
Good Systems faculty will also be involved in the degree. Fussell said ethics is an important part of education, pointing to the need to teach nuclear physics students that the science can be used to create bombs. "If you've got an AI system automatically evaluating people for mortgages, you certainly don't want it using criteria that are not relevant to their ability to pay the mortgage to make those decisions," he said. "But you gotta be very careful to understand that these kinds of things have to be an integral part of the systems that you're designing and the potential for good or harm is massive." As for ChatGPT's potentially concerning quirks, Fussell is not so worried. "When they released that big prototype with ChatGPT it didn't take very long for these things to show."
Cities That Think Good
Cities also use AI for a variety of reasons. Strover pointed to controversial smart streetlights with cameras in San Diego, and ShotSpotter, the gunshot detection software, as examples of how some cities are using AI.
The city of Austin has also taken notice of the trend, using AI in a range of departments, including Transportation, Fire, and Police. Junfeng Jiao, an associate professor of community and regional planning at UT-Austin, has partnered with the Austin Transportation Department and Austin Fire Department for some of his smart-cities projects. Separate from Jiao's work, KXAN reported last month that the Austin Police Department would be rolling out an artificial intelligence tool to help field and support people calling into its 911 center operations. APD did not respond to a request for comment on that project. "The goal of smart cities is to use information to make better decisions to enhance our residents' living and improve their standard of living and satisfaction in the city," Jiao said. He added that the city of Austin is overall doing a great job of embracing AI.
With ATD, Jiao has helped to establish a mobility hub at Georgian Acres in North Austin. The hub, which includes Wi-Fi, access to a circulator bus and e-scooters, and more, uses machine learning to model how future community hubs could connect different forms of transportation, he said. Jiao is also in the development stages of a potential project to use camera data to better map out the routes of emergency medical services vehicles. "We want to use smart-city technology and AI technology to better integrate all kinds of data in our digital twin model, and then use that to guide EMS and other important response vehicles to the destination," he said.
An ATD spokesperson said EMS is a high priority and noted that AI will play an important role in the city's mobility future. "In the past, transportation agencies would convene transportation operations personnel into a room, where they would stare at cameras all day, knowing they would not catch everything in real-time. That was not the best use of human intelligence, either – the operations personnel need to respond, communicate and dispatch services," the spokesperson said in an email. "Today, the transportation industry uses machine learning to help detect incidents, giving precious time back [to] the operators to send help. In the future, AI will play a more pervasive role in everyday mobility – most recognizably in the form of self-driving vehicles and self-flying drones." The spokesperson added that one possible application of AI could be through autonomous street sweepers operating at night.
Jiao's project with the Fire Department has placed air quality sensors throughout the city. Using historical fire data, the researchers could run simulations to determine how a fire would impact air quality in the surrounding areas and could eventually issue air quality advisories at the neighborhood level, Jiao said. He added that the project yielded a surprising finding: The city has fewer fires than he expected.
While that may be true, Justice Jones, the city's wildfire mitigation officer, said he is worried about the intensity more than the frequency of fires. Jones sought out the support of UT researchers, as research started to indicate that the secondary impact of smoke from fires injures and kills more people than the direct impact. Being able to predict the spread of smoke could be a critical tool for firefighters across the country. "We need the ability to direct the public through evacuation notices and warnings, and understanding the impacts of both the fire and smoke are critical," Jones said.
Jiao, who was a founding member and former chair of the Good Systems program, said it's important that AI research projects use expertise from a range of fields. "Good Systems is doing a wonderful, wonderful job trying to make sure all the projects, all the Good Systems research projects would include a technologist, a social scientist, and a humanist," he said. "So basically, this triangle structure would ensure there's diversity, there's a balance, there's equity when we do research." He added that his work with the city aims to protect the data privacy of citizens.
But has Jiao used ChatGPT to write a research proposal? No, because the application still tends to be too neutral, he said. While he could see using the bot in the future for disseminating traffic information, he pointed to neutral responses he had received when asking if the University of Texas was better than the University of California, Berkeley, as a limitation. "And which university is better, UT or UC Berkeley?" I asked. "Both are very good," Jiao said. Perhaps we're not so different after all.
Got something to say on the subject? Send a letter to the editor.
