Postcards From the Edge
Reflections on the implications of computable reality
In 1938 Alan Turing created the idea of a machine with an infinite storage tape, a read/write head and starting and stopping instructions and showed that this logical machine could execute just about any algorithm every conceived of. This, to me, was the big bang of what I call Computable Reality — and computable reality has been expanding rapidly ever since.
The Law of Computability determines what is computable or not. I define the law of computability as:
Level of knowledge of the phenomenon of interest * Level of digitization of the phenomenon of interest = Computability of the phenomenon of interest.
In this model there are three levels of knowledge: categorization, correlation and causation — each increasing in their completeness.
Consider the self driving car. There was a high level of knowledge of both the car and its driving environment. The car was largely computable — in design, testing, manufacturing, etc., there’s a robust digital twin. However, the car’s driving environment was not digitized enough to be computable. Chris Urmson, the first leader of the Google self driving car project, described how Google had to add a LIDAR to paint and gather 1.3 million pieces of data per second to add to Google Maps, onboard sensors, GPS and other gear in order to get the car’s error down from 3 feet to a tolerable size. In the language of the law of computability, by digitizing the car’s driving environment, self-driving vehicles became possible.
Once something is computable, it fundamentally changes the task, its economics, the economies of scale and scope and many other factors. From this lens of computability, the transformer architecture that is making the large language models possible has taken language from categorization to correlation — enough to enable prediction of proper dialog. Progressive Insurance drove transformation of the property and casualty auto business because it could compute risk better, faster and with more precision. If you can compute a valuable task, and your competition can’t — you win.
Facebook is computing individual and social cognition. LinkedIn is computing job search and job changing behavior. Amazon is computing buying behavior. The competitive battle of this century will be for firms to surf the leading edge of computability for tasks that are valuable and economically defensible.
Computability, writ large, has implications for firms, markets, the law, societies, art, culture, individual learning, social policy, etc., etc. This blog will explore my research, musing, thoughts, and screeds about this topic and I hope to create a wider conversation about the economic, legal and moral implications of this ever expanding universe.