Recently there have been some excellent articles and talks on AI, including:
- The Physicist Who Sees Crime Networks
- Science has outgrown the human mind and its limited capacities
- Mobile Is Eating the World, 2016 (recommended video)
Each of these articles deals with the advances in AI with respect to data analysis. The articles herald the amazing conclusions we have only recently been able to draw from the data. The second article goes on to talk about how the way humans think is changing to better use the new AI tools. The third article talks about how AI tools have totally changed in the last ten years, moving away from rules toward something that approximates human "learning". I could have picked 10-20 other articles in the last month that herald the same trends as these three articles.
The depressing conclusion I draw is that science and social science is now going to be almost entirely shaped by the available data sets and the AI tools available to researchers. Also, probably for the first time in history private companies now have greater resources to do AI research than universities and more importantly are using the resources for this research. However, this corporate research is more likely to find the next behavioral insight than a breakthrough such as chaos theory or quantum mechanics. Furthermore, the best academics in Physics, math and computer science are being lured away by the corporates to do AI research. The outlook for breakthrough research that changes the future of mankind looks much less likely.
All of this would be distressing if not for a bigger concern. As we continue to advance science we are coming to realize that almost all human behavior can be explained by biology and physics. A representative article showing this trend is Swirling Bacteria Linked to the Physics of Phase Transitions. (For some unknown reason I love these articles where the apparently chaotic is explained.) I call this the "think like an ant theory". To understand a problem, think like an ant.
A popular emerging field is the combination of biology and business, where complexity science brings a cohesiveness to two such disparate fields. From this increasing understanding and reliance on biology and complexity science, one comes to realize that these simple models and methods can probably explain nearly everything. For example, to understand cities, first understand hunter-gatherer tribes. To understand entrepreneurship, first understand exploration and exploitation (what we now call Lean Startup), the fundamental behaviors of every living thing. This trend toward simplicity, however, is not supported by the popularity of AI and data sets.
If we look back in human history, until the twentieth century research was not heavily reliant on data collection and data sets. Therefore, most of the great thinking in human history was done by methods that were far simpler than AI. I do not think we need to return to the 16th or 17th century methods at the expense of AI, but I do think we need to realize that AI is not a panacea and traditional thinking still has much to contribute.
One could probably argue that the problem with research since the last century has been an increasing effort for it to look data-based and scientific. AI will only make such approaches easier and wider ranging. The original style of thinking of Aristotle, Descartes and Hayek should not be forgotten or held in disdain.
Brain Pickings today writing about Bertrand Russell:
"But beneath such a surface impression is enormous depth of insight and a timeless, increasingly timely clarion call for nuance in distinguishing between the sort of knowledge driven by a greed for power and the higher-order wisdom that makes and keeps us human."