Please ensure Javascript is enabled for purposes ofwebsite accessibility

Artificial Intelligence, Part 3: OK, it's bad. But it's also good, right?


In this file photo from May 2, 2019, a computer monitors some of Cadet Cheyenne Quilter's reactions as she works with a virtual reality character named "Ellie" at the U.S. Military Academy at West Point, N.Y. (AP Photo/Seth Wenig, File)
In this file photo from May 2, 2019, a computer monitors some of Cadet Cheyenne Quilter's reactions as she works with a virtual reality character named "Ellie" at the U.S. Military Academy at West Point, N.Y. (AP Photo/Seth Wenig, File)
Facebook Share IconTwitter Share IconEmail Share Icon

Let’s rewind a little bit to earlier in this week’s AI coverage – the detecting cancer part.

Through all the “robots are taking over” rhetoric surrounding AI, there are scientists taking major steps in health care innovation, and humanity reaps the benefits in that arena.

Take a radiologist, for example. These are experts tasked with finding a small malignancy in an image – but not just one image, hundreds. They spend hours daily in a dark room searching and scanning, and based on what they find, they have to prepare themselves to tell life-altering news to patients.

AI is showing an ability to spot cancer at least as well as human radiologists. Researchers in Boston say they’ve developed AI that can detect early signs of lung cancer years before doctors could see it on a CT scan.

In one study, the tool – called Sybil – accurately predicted whether a person would develop lung cancer in the next year 86% to 94% of the time by looking for signs of where cancer is likely to turn up.

Similar predictive technology is being worked on in neurology.

Cynthia Rudin at Duke University is working with neurologists to try to predict whether a patient in the intensive care unit of a hospital will have a seizure by using a scoring system. Once doctors have a good idea of who needs to be monitored for seizures, they can prioritize who gets an EEG monitor (an electroencephalogram detects abnormalities in brain waves and the electrical activity of the brain).

AI diagnostics is still in its early stages though, and more clinical trials need to be conducted before any widespread adoption.

For now, health officials say AI can assist in the mundane tasks in health care – filling out forms, analyzing patient data and identifying patterns.

Just hammering out the administrative work could be transformative in an industry that faces staffing shortages and burnout.

A 2016 study by the American Medical Association found for every hour a physician spends with a patient, they spend an additional two hours on administrative work. AI is capable of filling in paperwork in a fraction of the time it would take a physician.

“I think that’s where the assist comes in real-time. Rather than a busy primary care physician or other specialists having to comb through reams of medical records that are digitized, the computer could be able to do that and then present something in a timely way, and then make a suggestion,” said Dr. Bruce Rollman, a professor of medicine at the University of Pittsburgh School of Medicine and a primary care physician.

Last month, Microsoft announced it’s expanding its partnership with Epic Systems, which created the MyChart software many hospitals use. It stores people’s health information so they can directly access it online, and generates automated responses to questions patients may have.

Microsoft also unveiled a new clinical notes application called DAX Express, which automatically drafts a clinical note within seconds after a patient’s visit.

There’s the potential for scale because you’ll never have enough doctors, you’ll never have enough therapists,” Rollman said. “And, of course, computers don’t get tired. So they have that positive potential to deliver population health.”

Speaking of paperwork, AI can alleviate the same sorts of burdensome tasks for educators, who are also facing staff shortages and burnout after the COVID-19 pandemic. Melissa Loble at Instructure said teachers are already using AI programs to grade subjects like coding and algebra – where there’s a correct or incorrect answer with no nuance.

It can also be used by teachers and professors to build remedial content.

“Let’s say you teach an algebra class and there’s a basic math topic that’s not taught in your class but you’ve got some students that just don’t have that grasp of the beginning topic,” Loble said. “You can actually point AI to help you build some simple content that you can then include in your instruction. You’re not having to write it from scratch or edit it.”

While AI is capable of also taking over the administrative work in the criminal justice sector, researchers are taking it a step further to try to use it to predict crime.

Researchers at the University of Chicago developed an algorithm to forecast crime by learning patterns in time and geographic locations from public data on violent and property crimes. It predicted future crimes one week in advance with approximately 90% accuracy.

AI-powered tools can also help judges assess risk. COMPAS is a tool that uses machine-learning algorithms to predict the likelihood of recidivism by analyzing factors such as criminal history, socioeconomic background and mental health.

In fact, the U.S. Sentencing Commission already uses AI to develop and implement sentencing guidelines to help judges determine fair and consistent punishments.

Another area AI can prove transformative is studying climate change.

Jim Bellingham, executive director of the Johns Hopkins Institute for Assured Autonomy, said AI is crucial to assist in analyzing the impacts of climate change and trying to mitigate them – efforts that have faced scrutiny over cost-benefit analysis.

Plus, climate data sets can be tremendously large. Climate scientists are faced with analyzing data on the entire history of the entire planet.

This is arguably the toughest problem humans have ever taken on in a scientific sense, in terms of the complexity of the entire planet and its ecosystem,” Bellingham said. “Direct linkage is something that we need to understand when we're figuring out what the costs are. Because there's a cost-benefit, a trade-off to everything we do. And as a consequence, we need to understand whether or not the mitigation steps that we're taking are really going to have the impact we want.”

AI can connect data sets from satellites and observations with model predictions, which are the key in drawing conclusions and recommending policy changes. The technology can also assist in the design of clean energy, like constructing the largest windmill possible but using the most lightweight materials still able to withstand extreme weather conditions.

Furthermore, it’s used to optimize energy grids by managing the flows between homes and businesses and the power grid itself, reducing energy waste.

AI’s integration into farming is found to result in higher productivity and yield. The technology can give farmers real-time insights into their fields, plant health, soil quality, temperature, irrigation and pesticides, using drones and robots.

And transportation is a sector with AI already deeply embedded. It’s in GPS technology that tells drivers the best route to a destination based on road traffic. Uber and Lyft drivers use AI to anticipate supply and demand, estimate arrival times and optimize decision-making.

Stephen Smith is a research professor of robotics at Carnegie Mellon University. He worked on a team that developed an adaptive traffic signal network called Surtrac, which improves commutes during peak traffic times in Pittsburgh. AI systems collect real-time data on where cars are on the road to change the traffic light signals based on where traffic is the worst.

The city partnered with his team after a pilot program showed it improved traffic flow and reduced average travel times by 25%.

Not only does this reduce the number of headaches for commuters, but cars spend up to 40% less time idling, reducing that much more carbon emissions from the air.

“There are technologies out there that will adapt the traffic signals to the traffic, but they tend to be applied to more suburban corridors where you have a main drag and then maybe side streets,” Smith said. “The problem we’re solving is one where you have multiple competing, dominant flows that change throughout the day, so you can’t really decide in advance where your dominant flow is. The system has to recognize that in real-time.”

In the disability community, different AI-generated platforms can translate text into sign language, identify visual disorders in children to prevent blindness and translate emotion into sounds to help blind people “see” the emotion of the person they're talking to.

The list goes on with the things AI can achieve to society’s benefit.

The question is, how can we reign it in to only achieve things that benefit society, and prevent it from doing things that detriment society?

Loading ...