Alabama Alumni Magazine spring 2023 issue cover art

The University of Alabama is on the forefront of how artificial intelligence is changing the way we live in the world. By Edwin Stanton

Admit it— when you think about artificial intelligence, your first thoughts are of Arnold Schwarzenegger, Skynet and T-800 cyborgs.

That’s understandable. The Terminator movies— about self-aware machines determined to wipe out the human race—are wildly popular and have a huge impact on how we think of artificial intelligence.

But AI is not science fiction doom-and-gloom.

In simple terms, AI is the simulation of human intelligence processes by machines. The two types of AI are weak and strong. Weak AI are systems that are trained for a particular task. When you ask Siri a question on your iPhone or tell Alexa to play your favorite song, that’s weak AI in action.

Strong AI is more complex. It deals in systems with cognitive functions that mimic the human brain. This mirrors what you see in the movies: The Terminator, The Matrix, Blade Runner or 2001: A Space Odyssey. 

There may come a day when Hollywood’s version of AI technology is achieved. Right now, AI is about optimizing machine learning and task performance.

“AI is software that’s performing a task that could formerly only be done by a human,” said Camgian software engineer Ethan Mines, ’21. “If that’s the case it really makes it a moving target. A lot of people say AI is whatever AI hasn’t done yet. There is a lot of truth in that.”

The University of Alabama’s College of Engineering works with AI in a number of ways. The Alabama Astrobotics team has gained a reputation for success.

The team has claimed numerous national championships for designs in the NASA Robotic Mining competition.

UA has also partnered with Camgian, a Mississippi-based company that develops digital technologies and platforms through AI. Camgian’s Center for AI and Machine Learning operates in Tuscaloosa, which allows for a perfect partnership with UA’s College of Engineering.

“We’ve been working with Camgian for about three or four years,” said Dr. Kenneth Ricks, head of the Electrical and Computer Engineering Department at The University of Alabama.

“They’ve got some (Dept. of Defense) projects they wanted help with, and our role with that was to apply machine learning and AI to various types of object recognition, which is a very popular application of AI and machine learning.”

It’s a symbiotic relationship between Camgian and UA. Camgian gets help on projects from UA and hires UA students, while the money UA gets from Camgian clients goes toward funding graduate students and programs.

UA and Camgian worked together on numerous projects, including a few for the U.S. Department of Defense. One job involves object recognition, which Ricks mentioned. Machine learning was a part of that.

In machine learning, large amounts of data are uploaded to an offline network and the AI system is trained, through algorithms and convolutions, to make predictions or recognize patterns.

“You are giving a computer a set of inputs and you want a desired output, but you don’t want to explicitly tell the computer how to get the right output,” said Camgian software engineer Vyas Padmanabhan ’22. “You’re giving it a chance to figure it out itself.”

When an AI system fails or doesn’t have a high accuracy rating, the parameters are tweaked until the system’s predictions are more accurate.

“Some of it is guesswork,” Mines said. “Say you are training a model

to classify images of whether they have a cat or a dog. That consists of the model initially guessing the image has a dog in it, and you telling it ‘no it’s actually a cat.’ As it processes these large numbers of random numerical parameters it tweaks and adjusts for each correction you give it. You nudge it toward giving a correct answer next time.”

Grad students at UA did this by designing an AI system to recognize objects using LiDAR (Light Detection and Ranging), which uses much less data than actual images. The LiDAR emits several laser points, which can penetrate fog, rain, smoke and dust; works in the dark; and measures the echo from an object.

“It collects terabytes of data in the field, thousands of images. These data sets are 40-50,000 images that have to be manually labeled,” Ricks said. “You tell the computer what the image is. You feed all that to the network, which learns how to interpret various scenarios compared to what it’s seen in its training set, to then make an educated guess to whether it’s a train or a truck.”

The AI system is deployed into the world after intense offline training. This is just one of hundreds of AI systems in the world. AI is so common today that it’s hardly noticeable, even around the house. Smart locks and smart thermostats that can be

turned on with your smartphone use AI. The robot vacuum that navigates your home incorporates AI. Even refrigerators have AI; they can come up with a recipe based on the food you have.

Like other technology, AI continues to improve as more innovations are made.

“Ten years ago, you didn’t have radar on your car that gave you a signal when you were going to hit something,” Ricks said. “That’s why it costs so much money to fix your car now. Look at all the driver assist that’s come along in the last 10 years.”

The day of fully autonomous cars will happen eventually, Ricks said. People’s full trust in these cars is the key issue.

“I read an article recently that said children born this year will never have to learn to drive. It will be pretty cool, but I don’t know if it will happen that quickly.”

Innovations with AI are everywhere. It’s commonplace in the tech industry, but it’s also moved into other areas of daily life. There are AI deep-learning programs that can write essays, poems or even songs.

The world is changing so fast it’s hard to predict what AI will look like in the future. It’s not going away, but it will be something new and unique.

“I’d be wary of predicting the future,” Padmanabhan said. “You see those old newspaper clippings saying, ‘This is what we think the world is going to look like in 20 years.’ They were way off. It’s obviously going to be something different than what we have thought of.”

Roll Tide and AI

Alabama Astrobotics team instrumental in NASA lunar missions

There is no rest for the University of Alabama Astrobotics team.

When you win seven straight national championships with record-setting results— slacking off isn’t an option—the University has a reputation to uphold.

The UA Astrobotics team goes for overall national title No. 10 and eighth straight in May at the Robotic Mining Challenge.

The goal for the team is to create a robot that can mine materials on the moon, specifically ice, which is embedded a few feet below the surface. At the competition, the ice is simulated by gravel.

NASA needs robots to support life on its Artemis missions, which aims to send astronauts back to the moon. Artemis I, which served as a flight test, launched in November 2022, orbited the moon and returned to Earth. Artemis II will launch in 2024. Artemis III, which plans to land a crew on the moon, is set for launch in 2025. The robot UA designed will be instrumental in mining for ice and other materials to provide water, hydrogen and oxygen for the astronauts once they land.

The team has been creative in its designs with the use of artificial intelligence systems, using LiDAR (Light Detection and Radar), which doesn’t involve human interaction.

“We have solved some of the problems with autonomy,” said Dr. Kenneth Ricks, head of UA’s Electrical and Computer Engineering Department and faculty advisor for Alabama Astrobotics. “We are using LiDAR. It requires less processing, which requires less power and it can be smaller and require less space. All those things apply to that application.”

Another plus for UA’s rover design: it’s optimal for handling the hazardous moon conditions.

“The dirt is a hazard. It’s compact and hard to dig through,” Ricks said. “With the top layer, it’s easy to get bogged down. It’s also electrostatically charged, so it wreaks havoc on electronics, and it sticks to everything it touches.

“Machine learning and AI will play a huge role in mapping the environment and identifying obstacles.”

The competition robot needs to be built using specific weight and size guidelines. That’s no problem for the UA team. It has set the standard for lunabotics with its innovative designs.

Alabama has improved upon its robot designs every year. The 2022 championship team built ‘Goby,’ a 176-pound robot that collected a record 5.2 kilograms (just over 11 pounds) of simulated ice. The second-place team collected just over one kilogram.

“A mining rover is not a delicate little flower. It is a workhorse,” Ricks said. “It has to be unbelievably rugged and durable and has to be able to do its own health management.”

The team has some tweaks for the new rover it will enter in the 2023 competition. If they are anything like the previous models, championship No. 10 is well within UA’s grasp.