Artificial intelligence has great power, but it is frequently misunderstood. We often hear of people saying AI is going to take over the world. The truth is, yes, AI will change industries. It will do things that humankind couldn't do alone. But will AI take over the world? Not likely.
The relationship between humans and technology, though, is the key. We can leverage AI to help us achieve new milestones—but it will not replace the power of the human intellect. Instead, we can use AI to replace more tedious and manual tasks. This will save time and allow us to focus on more complex areas of thought—which can make all the difference in times of crisis.
Recently, data scientists, including myself, used AI in the race for answers around COVID-19. We wanted to share that story, so check out our latest On the Merits film, Pandemic, to see how we helped.
The White House’s Call for Data Science
In March 2020, the White House Office of Technology issued a call to action for data scientists. They asked us to use AI and machine learning to help public health researchers find answers to questions on COVID-19.
Time was of the essence. The number of cases was increasing rapidly, and researchers faced a growing number of unknowns. How did the virus spread? What are the risk factors for pediatric patients? Each question brought to light a series of sub-questions. And new literature was coming out every day around the virus.
Without help, there was no way to keep up. So Semantic Scholar, a sponsor of the call to action, publicly published a data set containing the research. The idea was for widespread data scientists to access the data set, titled CORD-19, via Kaggle, an online platform for data scientists and machine learning practitioners. Then, they could use what tools they had to parse through it quickly and discover important takeaways that could help answer that growing number of questions.
At Relativity, we were eager to help. So I teamed up with my colleagues, Trish Gleason and Alex Wilcoxson, and we got to work.
Eliminating Bias in the Data
We knew Relativity was an ideal platform for such a challenge. Our analytics features identify relevant concepts and patterns in the data. One area that we found we could provide value was by eliminating duplicate and redundant information. Duplicate information is not a rare phenomenon in data sets. A researcher may have her article published in many journals. There is no natural identifier marking that article as identical to the same article in another journal. This results in the publishing of duplicates.
If an article appears 50 times in a data set, its weight is 50 times greater than another article that only appears once. The 50 duplicates could drown out the information in the articles that appear less frequently and suppress information that public health experts need. For CORD-19, we used Relativity Analytics to identify upwards of 4,000 duplicate articles and repeated phrases. This may not appear as the most revolutionary application of analytics. But it was important. Researchers could make better insights as they worked toward resolving the pandemic.
What the Future Holds for AI
In our film, Tayab Waseem, PhD, stated: “We are creating AI tools that are able to start parsing through large amounts of data that a normal human could never go through—and finding relationships that we didn't even know existed.” This quote accurately summarizes a collaborative relationship between AI and humans. It also illustrates its potential.
We are seeing remarkable results in healthcare. One study found that in orthopedics, an AI-assisted robotic procedure resulted in five times fewer complications than the surgeons operating alone.
As we see emerging use cases for AI, I believe the role of the data scientist will also evolve to implement algorithms that are more responsive to user feedback, worthy of our trust, computationally efficient, and environmentally responsible.
Rebecca BurWei, PhD, is a senior data scientist at Relativity.