Revisiting the Human-Machine Symbiosis

By William Regli, Defense Advanced Research Projects Agency

“The hope is that, in not too many years, human brains and computing machines will be coupled together very tightly, and that the resulting partnership will think as no human brain has ever thought and process data in a way not approached by the information-handling machines we know today.”

– J. C. R. Licklider, “Man-Computer Symbiosis,” 1960

Fifty-six years ago, J. C. R. Licklider outlined a prescient vision for computing machines coupled with human brains and, together, thinking thoughts previously unattainable by human beings thinking on their own. This vision influenced a generation of scientists and engineers and is largely the basis for our experience of computing today. Yet, I don’t feel a partnership with my current machines, and I often find myself bending my brain, and subjugating my will, to adapt to them. Shouldn’t it be vice versa? Did I miss the symbiosis?

Certainly our computing machines are considerably more user-friendly than they were in 1960. The fields of ergonomics and classical human factors have made great strides in creating interaction technology that better partners humans and machines. This is manifest in the mainstream computer science field of human-computer interaction (HCI) and, more recently, human-centered computing. We see the results of this progress in devices such as F-35’s Gen III Helmet Mounted Display System, Microsoft’s Kinect, and the multitude of new input technologies made possible by low-cost sensors and plentiful computing power. But symbiosis means something considerably more significant than just interaction technology.

The potential of the human-machine symbiosis is easily visible in the arts, where computing technologies have enabled the creation of previously unrealizable forms of expression. Computing technology has empowered a new legion of artists working in mediums such as immersive and augmented reality games, animated feature films, and music composition and performance. Where previously one needed a recording studio, one now has GarageBand; where one once needed a darkroom, one now has Photoshop; and where one needed celluloid film, one now has Blender and RenderMan. In these areas we are beginning to see humans and machines as complete partners in artistic creation. When we harness human beings and our wet matter (after all, the brain is just an alternative computing platform) we not only make impossible problems tractable, but we also create radically different approaches. We conceive of new creations, literally thinking thoughts that were previously unthinkable. Our objective then should be to redouble our work toward systems that augment our human nature and give us the means of seeing and thinking differently.

Amazing Opportunities Abound

The emerging area of human-based computation has begun to reveal some amazing opportunities in this direction. Luis von Ahn has shown how our human predilection for games and puzzles can be harvested to perform amazing feats, such as labeling images on the Internet via the ESP game and “OCR’ing” vast books via CAPTCHAs. In these cases, we are changing the relationship among the user, computation, and the overall task being solved.

In my own experience, as a DARPA performer on a mobile edge-networking program focused on the needs of dismounted soldiers, my team used the operational training doctrine from the Army Field Manual to build a content-centric networking paradigm. We rethought the networking paradigm based around the mission dynamics of the user. Our approach was contrasted with other performers who wished to use machine learning, sentiment analysis, and other data-analytics techniques to “learn” the content types of the user. The learning and big data approach was mathematically elegant and algorithmically sophisticated. But it solved an artificial problem. In contrast, by focusing on the human users and their information needs and mission goals–all of which were explicitly known at the outset–many of the algorithmic difficulties simply vanished, and the result was a radically different type of networking capability for our warfighters.

This is really what is at the heart of the symbiosis: to understand the proper and, ideally, optimal role of the human at the center of the human-machine system. The problems become those solved by the human and the computer together, rather than just those solved by the computer alone. Perhaps this is what has been missing: We often are asking the machine to solve a problem on its own rather than as part of a human-machine team. For example, to some traditional engineers a computer is merely a better calculator; MATLAB has replaced slide rules, but the human-driven tasks are essentially the same. This is the reality to many outside of computer science, where a database is still only an Excel spreadsheet and a word processor is merely a replacement for a typewriter. To fully realize the information revolution and bring about the economic bounty of “the Second Machine Age” (noted by Erik Brynjolfsson and Andrew McAfee) we need to refocus computer scientists on reformulating the problems of society to be tackled by a human-machine partnership. (I wonder if this is not one of the roots of computer science’s diversity problem: we treat the machines as separate from us, and they separate us from our humanity rather than augmenting it.)

From this perspective, there are at least two implications for computing education. First, with the exception of the fraction of students that take a course on HCI, cognitive psychology, or (perhaps) certain flavors of software engineering, it is quite possible for a computer science undergrad to obtain a bachelor’s degree without ever being exposed to the needs of a human user, let alone a user who’s in a complex or messy science or engineering domain. Use-inspired and user-centered thinking has to become more common in our undergraduate computing curricula, and we should seek out these complex or messy science and engineering problems to tackle with our techniques.

Second, in almost all traditional computer science education, the computing machine is usually viewed as an isolated box. It is the place with the processor, data, and memory; the mathematical pseudo-code for an algorithm; the software and hardware “cyberphysical” system that delivers the bank transactions, flight controls, video game or factory plans, and more. How can we teach students to understand the role of the human at the center of the human-machine system? We want to train students to envision how to fundamentally change the very nature of problems so that they can be tackled by human-machine teams.

I think the implications are similarly radical for computer science research; we spend considerable intellectual effort to advance what we assume is fundamental computer science, yet we might be missing a much bigger opportunity because we are not asking the right question.

Asking the right question was key to the industrial innovations of Henry Ford and his team at the River Rouge automotive plant. Rather than simply scaling up existing labor practices, they systematically rethought the relationship between human and machine labor as being part of a shared system of production. The challenge of our time is to train a new generation of scientists and engineers who can rigorously explore the potential opportunities for shared human-machine labor–not to replace human labor but to augment it in all forms.

In an editorial published in 1949, Albert Einstein ruminated on the implications of science on society, concluding “We shall require a substantially new manner of thinking if mankind is to survive.” To my mind, the human-machine partnership is central to this new type of thinking. Given today’s age of rapid technological and societal change, the only way we’ll think “faster” and different is with our computational partners with whom we are codependent for survival.

About the Author

William Regli is the deputy director of the Defense Sciences Office at DARPA. He started at DARPA in 2014 after 17 years on the faculty of Drexel University. Regli has published more than 250 technical articles, including those in leading venues for research in computer graphics, artificial intelligence, robotics, wireless networking, tissue engineering, and engineering design and manufacturing.

He is a senior member of the Association of Computing Machinery, the Institute of Electrical and Electronics Engineers, and the Association for the Advancement of Artificial Intelligence.