How the Next Generation is Building Artificial Intelligence

Artificial Intelligence

Teen scientists use machine learning and neural networks to detect and diagnose diseases, track space debris, design drones and justify conclusions at Intel ISEF 2017.

While sentient computer beings like HAL from the classic 2001: A Space Odyssey or Samantha from the 2013 film Her may still be on the distant horizon, some forms of artificial intelligence (AI) are already improving lives.

At the 2017 Intel International Science and Engineering Fair (ISEF) – where nearly 1,800 high school students gathered to present original research and compete for more than $4 million in prizes – the next generation of scientists used machine learning and artificial neural networks to find solutions to some of today’s most vexing problems.

“AI is critical to our future,” said Christopher Kang, a budding computer scientist from Richland, Washington, who won an ISEF award in the robotics and intelligent machines category.

Chris Kang
Chris Kang developed an app that can detect skin cancer.

“Humans have a limit as to how much data we can analyze,” he said. “AI is extremely powerful in analyzing huge volumes of data and correlating it. It can be used to understand the already existing research we have and assimilate it. It can then analyze additional data and turn it into actionable insights.”

AI as Physician’s Assistant

After his father’s brush with skin cancer, Kang created an artificial neural network capable of identifying moles and skin lesions that are potentially cancerous.

The teen created an app that lets users take an image of a questionable mole or lesion. The app is then able to determine if the skin anomaly appears cancerous, with a level of accuracy commensurate with a dermatologist.

“Patients can get enough information to know whether or not they need to see a dermatologist,” Kang said.

“Early detection is crucial because once skin cancer metastasizes, the probability of a five-year survival rate drops to a fifth of what it would be before,” he said.

AI for Diagnosis

Gaurav Behera of Rochester, Minnesota used a similar approach to develop an artificial neural network that identifies sickle cell anemia from a microscopic image of a blood smear.

“Sickle cell anemia is mostly present in the developing world, and access to education and diagnosis is often lacking or very expensive to find,” said Behera, whose great aunt died of the disease in India.

ISEF Artificial Intelligence
Gaurav Behera is using AI to advance sickle cell research.

“What this application does is bring down the cost of diagnosing sickle cell anemia and make the process more efficient.”

Currently, the diagnosis involves a tedious process of counting each of the 300 to 400 cells in a blood smear, where only 10 or so might be misshapen sickle cells. Behera’s app takes human error out of the equation.

“Since I’m having it centralized in one location, it can update itself constantly,” he said. “So the neural network can learn and become more precise over time.”

AI for Predictions

While the most common use of machine learning to date has focused on classification tasks, neural networks also can be used to track data and make predictions.

Amber Yang of Winter Park, Florida won a top prize at ISEF for developing an artificial neural network capable of predicting the location of space debris, which can be a hazard for spacecraft. She created an algorithm that can train the network to recognize specific debris clouds in space and, based on past trajectories, predict future locations.

Amber Yang
Amber Yang discovered that AI can find space debris.

Meanwhile, Vincent Moeykens of Windsor, Vermont created a neural network to analyze data and predict stock prices.

Using historical stock data from Apple, Amazon, Google, GoPro and Tesla, the teen entered the first three quarters of a year’s data, allowing the network to predict the fourth quarter.

The model proved to be accurate with less than a 0.15 percent difference between predicted and actual stock value more than 50 percent of the time.

When additional variables were input, Moeykens said, “the neural network met the constraints by training itself and improving its accuracy.”

Machine vs. Human

Robert van Zyl of Peachtree City, Georgia won an award in engineering mechanics for his research on whether machine learning could outperform human designers in engineering design.

He pitted his machine learning algorithms against humans in the design of complex power trains for racing drones.

“Racing drones were chosen because of the high stakes of the competition and the high complexity of the engineering problem,” said van Zyl.

The budding engineer provided both sides with the characteristics and performance data of multiple brushless motors and propellers in various combinations, and tasked them with selecting the best performance design for three variables: maximum thrust, efficiency and dynamic punch.

In the end, the machine won out.

“Despite some good performances by humans, the AI algorithms beat the best humans in two of the three contests,” said van Zyl. “The machine learning algorithms were particularly impressive in finding the optimal powertrain design to optimize punch, arguably the most important dimension in drone racing.”

AI for Good

“A problem with neural networks is you can’t look inside them, you can’t see what they’re thinking,” said Robbie Barrat of Shenandoah, West Virginia. “They’re like a black box.”

Barrat’s research seeks to make the inner workings of artificial neural networks more transparent.

Robbie Barrat
Robbie Barrat is using neural networks to justify themselves.

He developed two separate machine learning models that enable neural networks to return justifications with their conclusions in both prediction and classification problems.

“The justifications not only strengthen the conclusions provided by the network, but also provide insight into the ‘thought process’ of the neural networks,” he said.

For anybody still worried about a worst-case sci-fi scenario, Barrat said he believes such transparency can help ensure that, in the future, AI continues to serve humanity and not the other way around.

“Honestly, AI is still in its infancy,” said Barrat, noting that it took until the 1990s for computing power to be powerful enough to make neural networks practical. “But it’s here to stay. Because neural networks are very powerful tools and, given enough computational power, they can do anything.”

iQ newsletter signup

The post How the Next Generation is Building Artificial Intelligence appeared first on iQ by Intel.

Share This Article

Related Topics

Education Science

Read This Next

Read Full Story