Can Greatness be Planned?
Published in IEEE Spectrum Magazine, Jan 2016
Ken Stanley and Joel Lehman have recently published a book entitled “Why Greatness Cannot be Planned,” in which they claim that setting objectives for a project can be self-defeating.
They have been motivated to reach this conclusion through experience with an image morphing website, www.picbreeder.org, where users can “breed” new offspring images by successive selections among mutations of input images. By this selective image breeding users starting with random blobs have produced a number of images that strikingly resemble real objects or faces. However, such realistic images are never generated by setting out with an objective to produce a particular image. For example, an image that looks very much like a car might be discovered, but that is not done by selecting the most “car-like” image in each generation. Instead, it might be found by further breeding of another user’s image that happened to look a little like a car, in spite of that user’s search for something else altogether.
Stanley and Lehman go on to argue that, as a general principle in engineering and life itself, greatness cannot be achieved by making incremental progress towards pre-selected objectives. While I am frankly skeptical of this principle, it is one well worth examining, since all of our engineering culture and environment is based on the setting and achievement of objectives. In education and practice engineers are problem solvers, and in the obtaining of funding and support it is always necessary to make proposals based on objectives that are proposed to be achieved.
It also seems to me that most, if not all, of those achievements that we consider the greatest of engineering have been made through the seeking of established objectives. Marconi set out to invent radio, and the inventors of the transistor sought to invent a solid state amplifier that could replace vacuum tubes in the telephone network. Jack Kilby set out to implement the first integrated circuit, and then Gordon Moore established an overriding objective that has guided further development of large scale integration for more than half a century. The Internet was created under contracts and program guidance of ARPA.
However, the arguments that Stanley and Lehman make are more nuanced than their declarative title might indicate. I am especially interested in their description of greatness as being achieved through a series of stepping stones, or branch points, none of which – until the end – resembles the final product. This has certainly been true in the image evolution in Picbreeder, but also is probably the case in great inventions if we look at a longer time frame. For example, at the time that ARPA began the Internet development, the stepping stones were already in place – integrated circuits had enabled computers and previous studies had proposed packet switched computer networks. The end product was in sight.
In my own experience as a research manager I have often felt that the least obviously successful projects were those for which more stepping stones would have been needed for achievement of their ambitious goals. In such cases it may be indeed true that setting objectives is self-defeating.
On the other hand, Stanley and Lehman argue that instead of setting objectives we should seek novelty and look to collect stepping stones that may be useful to others. Perhaps the researchers in those projects that I regarded as unsuccessful did that.
I’m still thinking about all this.