//DAG// The term of the hour was machine learning, though in getting its brief moment on the stage, the term had shifted from its traditional meaning: no longer was it a technique to build a machine that mimicked human behavior but a series of methods that could tell humans how they behaved.
The setting was a research conference, a place where we retreat to a change of scenery and good food to talk about the future of computer science and software engineering. The stories of machine learning came on a day when a parade of young researchers took the podium to tell us of their creations. They started their stories by describing a problem of management, energy control, medicine, or process control. They would then shift the scene by saying, "and then I applied machine learning to this problem."
In their research, they used machine-learning algorithms as a tool, much as others might use a spreadsheet, search method, or statistical procedure. They were in search of human patterns. Once they identified the pattern, they left the machine behind and moved to the study of the human, which is where the real problem was.
//EDD// David and his fellow conference attendees could have learned the same lesson—that new machine learning was more involved in human behavior than with technology's—from the space program of his own era.
When asked to recall their most remarkable memory of their time in space, many astronauts describe a view of Earth. The vibrant blues and greens and white swirling clouds formed an inviting portrait of home.
No doubt contributing to the beauty of our planet in the astronauts' eyes was a sense of achievement, having surpassed the atmosphere with enough fuel to return. It was a technical triumph but also a social science one: good engineering management had led thousands of civil servants, armed services personnel, and government contractors to work together to send the astronauts far away and return them safely home. The challenges of space flight are both technical and social—from recruiting and training the top candidates to procurement and systems integration, not to mention the need for economic and political viability.
One of my earliest and most influential professors once described modern history in a graphic: we move across time (the x-axis) through a series of "ups and downs" (on the y-axis). We have an innovative period and move up (the line spikes); then we go through a time of sorting and organizing (the line drops slightly). After trying something new, it takes us time to recognize the many applications of the innovation and to understand its meaning. It's hard in the moment of technological—or social or political—advancement to see the full spectrum of its impact. Once we establish that we can accomplish the technical, we must work through the social. We can build a rocket to take us to Mars, we know. It's the psychological, political, and economic that holds us back.
//DAG// Back on Earth at the conference, more than a few of the old guard recognized a subtle shift in the nature of computer science. They rose, one by one, to ask why computing needed to embrace the methods of social science.
Most acknowledged that computing has been a remarkably fluid field. Computer science has expanded to embrace ideas from circuit theory, mathematics, logic, management theory, and graphic design. Today, it seems to be moving rapidly into the social sciences, not to build machines that simulate our actions but to hold up a mirror to our lives.