Microsoft believes artificial intelligence, or AI, can be a force for good in the world. To prove the point, last September the company started its AI for Humanitarian Action initiative — a $40 million, five-year program to use AI to address some of the world’s biggest humanitarian challenges.
The initiative recently bore fruit through a collaboration between Microsoft data scientists and researchers at Seattle Children’s Research Institute. The research resulted in a study, published in Pediatrics, that has helped shed light on the link between maternal tobacco use and Sudden Infant Death Syndrome (SIDS), for which there is no known cause.
Researchers already knew there was a relationship between smoking and Sudden Unexpected Infant Death (SUID) — an umbrella term that includes SIDS as well as accidental asphyxiation or suffocation in bed — but weren’t sure how significant a role tobacco use played, according to Dr. Tatiana Anderson, a postdoctoral neuroscience researcher at Seattle Children’s Research Institute, and the lead author of the study linking maternal smoking with SUID.
The study found smoking a single cigarette per day while pregnant doubled the risk of a woman’s child dying of SUID. The odds of SUID increase .07 with each additional cigarette a woman smokes. Researchers were most stunned to find that the children of women who stopped smoking before they were even pregnant still saw an increased risk of SUID. Smokers who plan to become pregnant should quit smoking before becoming pregnant to ensure optimal health for the baby, Anderson said.
Anderson said the research shows that 800 infant deaths in the United States could be prevented each year if no women smoked during pregnancy.
The study wouldn’t have been possible without Microsoft’s AI poring over vast quantities of data culled from the Centers for Disease Control. The study looked at every live birth in the U.S. between 2007 and 2011, which represents more than 20 million live births and more than 19,000 cases of SUID, Anderson said.
“What the machine-learning algorithms will do is try to take the signal from the noise,” said Juan Lavista, senior director of data science at Microsoft. The algorithms try to isolate which factors associated with SIDS are important and remove the ones that aren’t. “In this particular case, we’re talking about smoking. Once you control for all the other factors that matter about SIDS, what is the impact of … smoking once you remove the rest? That’s what the machine-learning algorithms … do.”
The algorithms are able to sort through data far faster than human beings could, Lavista added. “It would take a human, to do it by hand, a significant number of years,” he said. The AI “doesn’t do everything, but it identifies those features that are really relevant, in an automated way. And that provides a lot of power for us.”
Microsoft’s team will continue to study SIDS with Seattle Children’s researchers. Other research initiatives also are on the horizon, focusing on the initiative’s four pillars: supporting disaster recovery, addressing the needs of children, protecting displaced people, and promoting human rights. “I think we can do a lot of amazing work with this type of technology,” Lavista said.
Although AI can do amazing things, the human element is still a crucial part of the research, Lavista said.
“Data science and … machine learning can only go so far in the sense that we can uncover and find these clues in the data set, but an extremely important aspect of this is having the medical expertise and medical knowledge of, what exactly is the data saying?” he said. “What are these clues? On top of this, it’s very important to understand how the data was collected in the first place, and basically, what other questions do we need to ask the data? This is why it’s so important to have both aspects working on these types of projects.”
Anderson said she thinks the collaboration between Microsoft and Seattle Children’s is one that could be expanded to answer many more questions. “I think that’s where Microsoft’s … program really wants to go, is to help use these amazing data science skills to help answer these questions,” she said. “Because we’re just getting more and more data, and computing power just gets more powerful every day, so I think a lot of advancements can be made if groups like Microsoft and Seattle Children’s join together on a single mission.”