From robot-assisted surgery to email spam filters, there is much we gain from artificial intelligence. However, people are often wary of systems that they do not fully understand. Dr. Ashis Banerjee, an associate professor of industrial and systems engineering and mechanical engineering at the University of Washington, has made it his mission to educate people about the ways in which automation can contribute to the betterment of humankind.
The words “automation” and “artificial intelligence” conjure up scenes from science fiction movies, complete with talking robots and sentient mega computers. Humanoids wear lifeless expressions as they stare ahead, performing programmed tasks while quietly plotting the destruction of humanity.
Truth be told, automation has actually been a boon to humans for quite some time now, even if we do not always recognize it. Space exploration, deep ocean research, bomb defusion, weather forecasting, and microrobots that may someday even treat cancer tumors at the cellular level are just a few of the numerous areas where automation has aided humanity.
Automated machines have the unique ability to intervene in dangerous situations where humans cannot. During the height of the pandemic, a robot dog named Spot helped nurses at Brigham and Women’s Hospital triage patients from afar. Spot is capable of “measuring skin temperature, breathing rate, pulse rate, and blood oxygen saturation in healthy patients, from a distance of 2 meters.”
Another model of Spot rolled into the Chornobyl Nuclear Power Plant Exclusion Zone, the site of a catastrophic nuclear disaster nearly 40 years ago, to create a 3D map of the radiation distribution. The area is still inaccessible and uninhabitable to humans.
AI is everywhere, from navigating congested cities using Google Maps to smart home devices to credit card fraud detection to your next Netflix movie recommendation. And yet, the idea of something nonhuman capable of learning and adapting unsettles us.
Dr. Ashis Banerjee, an associate professor of industrial & systems engineering and mechanical engineering at the University of Washington, wants to help people understand how automation can continue uplifting humanity.
Dr. Banerjee was drawn to studying engineering for precisely that reason. He wanted to apply mathematical and scientific principles to find solutions to real-world problems. While he did not initially focus on automation, he saw its potential to provide novel answers.
“That’s the kind of thing engineers are trained to solve. You should take in the scientific ideas you illustrate with some proficiency and then show how those principles might be applied to the construction of actual systems, equipment, instruments and so on.”
One of his current projects addresses a problem that has plagued consumers and businesses alike in recent years — supply chain issues. Most of us have been impacted somehow by the wide-scale disruption of previously established supply chains during the pandemic.
A pressing example is the national baby formula shortage, which is a direct result of supply chain issues and several formula recalls.
According to the White House, “entire industries that shrank dramatically during the pandemic” are now trying to reopen.
“Some businesses report that they have been unable to hire quickly enough to keep pace with their rising need for workers, leading to an all-time record 8.3 million job openings in April. Others do not have enough of their products in inventory to avoid running out of stock. The situation has been especially difficult for businesses with complex supply chains, as their production is vulnerable to disruption due to shortages of inputs from other businesses.”
According to Dr. Banerjee, many original equipment manufacturers source their parts from small and medium-scale enterprises. The chain that results from these interdependencies is the proverbial “supply chain.” Manufacturers often have to wait over a year for parts to be delivered from one of the thousands of suppliers worldwide. When only a few vendors stock essential components, production delays cause whole orders to back up.
Dr. Banerjee and his team built a prediction system to reveal the movement of components across networks in real time and to forecast future trajectories. With these forecasts in hand, businesses may be able to better anticipate and adapt to any fluctuations.
“What we looked into is whether we can do a better job at predicting the lead times of parts that tend to have very long lead times, which is very common, especially in the aerospace sector where you order out parts one year, sometimes even two years out into the future. And in one or two years, the whole world can change dramatically, as we all know.”
It is not always reliable to plan operations based on “whatever promised delivery times or dates for those parts might have been since they constantly get revised.” Those revisions cause “a huge degree of uncertainty over which parts will be available to actually do the final assembly of large-scale components.”
“So being able to do some degree of prediction way out into the future, or at least a couple of months ahead into the future of when the parts would arrive, rather than just being reliant on what the latest promised dates were, can potentially bring a lot of value to the inbound supply chain.”
Still, one of the roadblocks to further developing such a prediction model is “that it is very difficult to get enough of data or its difficult to convince enough companies to share their data.”
However, based on limited data sets through industry partners and macroeconomic trends, Dr. Banerjee and his team can predict six to eight weeks into the future when parts will be delivered instead of relying on supplier-estimated data.
In order to deepen these predictive models, more organizations will need to be open to these novel types of technologies.
“Not enough companies or organizations are seeing the value of this kind of predictive automation in their supply chain, even though they feel the pain. I wish more organizations would be willing to share the data in some capacity so that more accurate and more broadly applicable models could be built, which I think can help broader industry a lot.”
With the worldwide pandemic’s aftermath potentially ushering in a new era of constant upheaval, it is more crucial than ever for businesses to adopt innovative strategies for dealing with uncertainty.
The beauty of AI is that it continually learns from new data and can sift through large amounts of inputs quickly to recognize trends at a pace humans cannot. Using these dynamic prediction models could prevent a national scarcity of a crucial product, like baby formula, and allow for the efficient and economical replacement of stock in the face of rising demand.
When there is a disturbance in a region where warehouses are situated or when supplies are lost due to bad weather, supply chain automation can foresee the cascading effects. It can assess supply chain risk and performance from a fresh and instructive angle by developing and tracking what-if scenarios. The potential benefits are substantial.
The benefits of automation reach beyond utilitarian production. Dr. Banerjee is also investigating how using automation can improve the health of people who perform repetitive motions in the workplace. Injuries and illnesses of the musculoskeletal system are regular outcomes of this kind of labor.
“We wanted to have an ergonomic risk assessment of humans who are working in these warehouses or even in certain manufacturing sectors where they are doing this on a daily basis — activities that are particularly harmful to all their joints which lead to musculoskeletal disorders. Maybe the effects are seen years later, but once it happens, it becomes a real pain point. So those skilled workers now have to take extended leaves during which all their associated health care costs are really expensive and sometimes, they are not well enough to carry on their jobs, so they have to take on early retirement.”
Dr. Banerjee and his team are “developing this automated camera-based addition system which can basically provide real-time access to the risks of performing these individual activities.” Carrying out certain activities together over a long period of time builds cumulative stress and strain on the body.
“It would be helpful if we could build a real-time monitoring system; it gives us the possibility to do two things. Either you can provide simple enough recommendations on how these particular processes can be done a little bit differently so that they are going to be less harmful to workers, or you can also suggest that you can shift or rotate your workers so that the same set of people are not doing this stressful process. But more importantly, it can lead to a better design of workplaces, including the use of collaborative robots, so that you don’t have anyone at any time executing these ergonomically challenging jobs.”
Many people in America fear that automation will take over these necessary jobs. Dr. Banerjee feels that, in the long run, integrating automation into the workplace will help us to better utilize and appreciate the true intellectual potential of the individuals who currently fill those jobs. For instance, if automation advances far enough, repetitive office tasks are ideal for robots.
Dr. Banerjee believes that we should look at automation “not purely from the organization perspective of producing more things faster and higher quality, but it can also enhance the lives of humans who are working on those processes.” Then we can take advantage of the cognitive capabilities of humans “rather than asking them to do the same mundane jobs.”
“Most people are aware that AI seems to be making a huge impact. It is important to realize that AI and automation or robotics are not designed to remove people from the workforce. Automation’s goal is to transform human life, but even if we don’t reach that lofty goal, the plan is to at least change human life for the better, regardless of what domain we might be talking about. It’s going to impact people’s jobs and all, definitely, but the idea is to make jobs better, more productive, less hazardous or more fun to do.”
Because technology has its own unique difficulties, human knowledge and expertise will still be necessary, even while relying on automation. We can be happier and more productive if we make the most of both worlds, capitalizing on advantages while compensating for each other’s shortcomings.
For instance, there are many borderline or edge cases where an automated system may not know what to do. In those cases, there is a need for human intervention.
“There are quality inspection paths, which, even though we have a lot of vision-based systems that do these regular monitoring of parts as they move through processes, will require human supervision.”
Any autonomous system will also need upkeep in the form of mechanical repairs or software updates.
While “sometimes automation technology is developed by people sitting in air-conditioned rooms, it ultimately impacts the people working in much more harsh conditions on the floors.”
For this reason, we must dedicate ourselves to “retraining and re-educating” these employees to ensure they have the skills to take advantage of and excel at these new job opportunities.
Dr. Banerjee acknowledges the psychological reservations that make some individuals reluctant to embrace AI in their personal lives. This hurdle must be crossed before the broad use of automation can be possible. As a potential solution, Dr. Banerjee and his team are experimenting with adaptive models that consider the end user’s risk aversion.
One prime example of its application is existing automation in our cars and even entirely self-driven vehicles.
Dr. Banerjee says, “although we are not there yet, we are in conditionally automated vehicles when it comes to regular driving activities, like the usage of warning systems that alert you whether you’re going off lane or you’re potentially colliding with another vehicle.” These systems have all been automated.
They may not work as well during certain environmental conditions, like snowy conditions or a “torrential downpour.” However, they still work reasonably well. Dr. Banerjee says that people are still reluctant to take advantage of them, no matter their demonstrated competence.
According to studies carried out by various companies, there is not a tremendous universal acceptance of these technologies. This may be due to sporadic reports, software incidents and accidents or individual mindsets of drivers. So while “the same person might trust the lane direction system, they may not trust something else.”
“A recent work we did asked the question: can we actually customize this automation technology? It’s the same technology as a lane tracking system, nothing beyond what is already out there, but can we build an additional layer on top of it to customize the system recommendations depending on the preference and the risk tolerance level of the end users? And we managed to show, using some machine learning models, that it is possible to capture, at least to some degree, what the user preferences are, and how we can adapt the system to changing user preferences.”
The system would react to how comfortable the driver is while driving an automated vehicle. The driver may begin by being very risk averse, but over time, may start to trust the system more. In turn, the system would also become more risk-tolerant to a reasonable extent. So in the case of a lane-tracking system, it may go from being conservative that it is only okay to change lanes with a 150-feet gap between cars to something more moderate.
This adaptation process aligns with Dr. Banerjee’s philosophy that to get the general public to accept automation, you have to start slowly. You need to prove some positive outcomes and then expand the scope of what recommendations an automated system can provide.
Recently, an interesting use of artificial intelligence made headlines when a man won an art fair with AI-generated art. In order to produce the art, the winner had to input a particular set of keywords to create his submission. This led to another wave of reluctance to automation, specifically within the creative community. Dr. Banerjee believes that is not the ideal way that AI should be used.
“I think even in creative fields automation can be useful in the same way as in domains like engineering, operations, transportation or consumer services. If we are talking about artists, they may, for instance, already know what they want to create. They might have written or drawn some sections of it. And then they want either new inspirational ideas on what can be done to embellish the work, or they just want something that can potentially finish the more mundane parts of their creative effort. So, AI potentially has some value for that kind of thing, which might be accepted.”
This type of use returns to the principle Dr. Banerjee holds, that automation should exist to enhance and improve human life rather than replace their contributions. Dr. Banerjee became a teacher to carry this message forward.
“I care a lot about mentoring and inspiring the next generation of students to do something that’s both exciting and interesting to them, but also having a real impact doing something for the social good.”
He initially started off focusing on only research but found that the inquisitive nature of students and the general environment of a classroom deepened his passion for engineering.
“When I was just doing research, I mainly interacted with my peers who already had a wealth of technical knowledge. I enjoyed collaborating with them, but I was not seeing the kind of spark in hearing about something for the first time and then feeling, oh wow that is a cool thing for me to get into and understand more deeply. So the mentoring or education aspect was missing in my work. And that’s one of the big reasons why I chose to come back to academia.”
Dr. Banerjee also wants students to know that they do not have to be trained in computer science to excel in the field of automation since engineering disciplines are now so interconnected. He entered the field of automation in a nontraditional way, receiving a B.Tech. in Manufacturing Science and Engineering at the Indian Institute of Technology Kharagpur and going on to receive an M.S. and a Ph.D. in Mechanical Engineering at the University of Maryland.
Eventually, he went on to receive several awards, including the Top Engineer of the Decade Award in 2021 from the International Association of Top Professionals, a 2019 Amazon Research Award, and the 2009 George Harhalakis Outstanding Systems Engineering Graduate Student Award from the Institute for Systems Research at the University of Maryland.
“I did not do a computer science degree for my undergrad, so that’s not what my formal training was, but I am working here on both theory and applications of AI in so many different domains. So, I just want people to realize that you know, it’s always possible to get into this field even if you, for whatever reason, do not succeed in getting into a computer science major.”
Above all, what is important is a strong grasp of the fundamentals, a desire to learn continuously, an openness to new opportunities, and a steadfast commitment to making an impact in the world.
For any inquiries, you can contact Dr. Banerjee at firstname.lastname@example.org
Moumita Basuroychowdhury is a Contributing Reporter at The National Digest. After earning an economics degree at Cornell University, she moved to NYC to pursue her MFA in creative writing. She enjoys reporting on science, business and culture news. You can reach her at email@example.com.