August 06, 2018

“Big Data and the Future of Research” is the topic of the 2018 Sigma Xi Annual Meeting and Student Research Conference this October in Silicon Valley, the global center for technology, venture capital, innovation, and social media. Big data has enabled major advances in science, as illustrated by many of the articles in the September–October special issue of American Scientist. On October 26, symposia will start with a keynote talk by Steve Ritz, who is building the 3.2 gigapixel camera for the Large Synoptic Survey Telescope. Jeff Dean, head of Artificial Intelligence (AI) at Google, will deliver the other keynote talk about using big data to solve many of humanity’s most challenging problems.   

Big data is changing the nature of scientific research. Digital representations of large scientific data sets permit the identification of subtle patterns. Finding correlations that predict customer selections can be economically valuable even if the underlying causes are obscure. But such correlations are not enough for science; correlations do not prove causation. Understanding casual connections is essential—although  the use of umbrellas correlates with rain, it does not follow that banning umbrellas will reduce the amount of rain. Some claim that correlations are scientifically valuable in themselves, but the choices of what data to collect and how to analyze it inevitably affect the implications that can be drawn. Sources of incompleteness and bias always need to be identified and avoided in order to draw robust conclusions. 

Big data, and the use of AI to analyze it, raise new opportunities and new challenges. I’ve been impressed with how rapidly new tools such as convolutional neural networks can lead to new scientific achievements. The availability of new computing technology such as powerful graphic processing units has made non-linear and non-parametric analysis of big data not only possible but also relatively inexpensive. However, unequal access to big data can exacerbate inequality, because well-funded organizations are better able to collect and analyze large data sets. Data on human subjects inevitably involve privacy and ethics issues. New modes of analysis and visualization are needed—and are being developed. 
 
At a conference at Asilomar in Pacific Grove, California, in 2017 many AI leaders adopted principles, including one that states, “Superintelligence should only be developed in the service of widely shared ethical ideals, and for the benefit of humanity rather than one state or organization.” In June 2018, the California Assembly passed a resolution expressing the support of the Legislature for these principles as guiding values for AI development. The Annual Meeting is the next opportunity to discuss how researchers can respond to changes occurring due to big data. 

Joel Primack
Sigma Xi President