I have two purposes in mind writing this blog post:

  • Bring to your attention a new Santa Fe Institute paper on technological progress. I consider it quite a remarkable paper.
  • Provide guidance as to applying the findings reported in this paper.

Bela Nagy et al have recently published a working paper entitled Statistical Basis for Predicting Technological Progress. Numerous intriguing observations are made in this paper, including the following comparison between Moore’s Law and Wright’s law:

We discover a previously unobserved regularity that production tends to increase exponentially. A combination of exponential decrease in cost and an exponential increase in production would make Moore’s law and Wright’s law indistinguishable, as originally pointed out by Sahal. We show for the first time that these regularities are observed in data to such a degree that the performance of these two laws is nearly tied[1].

No doubt, quite a remarkable discovery and conclusion. “Time is the teacher” (i.e. Moore’s law) is for some purposes as good a predictor as “We learn by doing” (i.e. Wright’s law). The fact that Wright’s law takes into account (whereas Moore’s law does not) the number of units produced during the period of time under consideration does not really matter.

The two laws IMHO are indeed indistinguishable not in a broad sense but in quite a narrow sense. The figure of merit used for measuring technological progress in this paper is ”inflation adjusted cost of one ‘unit’”.  While this choice has various pragmatic merits such as straightforward aggregation of data across different kinds of technologies, it does not capture the performance “dimension” of a technology. For example, this figure of merit can’t be used to predict future performance of a computer system that is implemented using a certain technology. Technological progress over time, no doubt, will lead to improved performance of the computer system once it is re-implemented using the later generation of this technology. But, the results reported  in the paper are not applicable to making such predictions.

In summary, the paper validates the explanation of Wright’s law in terms of Moore’s law that had originally been proposed by Sahal in A Theory of Progress Functions. This validation applies to cost as the figure of merit. It does not apply to the performance “dimension” of technology.


[1] The authors describe the two laws in the context of the paper as follows: “Moore’s law here refers to the generalized statement that the cost of a given technology decreases exponentially with time. Moore’s law postulates that technological progress is inexorable, i.e. it depends on time rather than controllable factors such as research and development… Wright’s law, in contrast, postulates that cost decreases at a rate that depends on cumulative production…”

avatar

Israel Gat

Israel Gat is Director of Cutter Consortium's Agile Product & Project Management practice and a Fellow of the Lean Systems Society. He is recognized as the architect of the Agile transformation at BMC Software. Under his leadership, BMC software development increased Scrum users from zero to 1,000 in four years. Dr. Gat's executive career spans top technology companies, including IBM, Microsoft, Digital, and EMC.

Discussion

  4 Responses to “A Note on “Statistical Basis for Predicting Technological Progress””

  1. avatar

    This paper is a testament to the power of data.

    Only by assembling the best and biggest historical datasets could the authors find their “previously unobserved regularity,” much as Clayton Christensen did in The Innovator’s Dilemma.

    • avatar

      No doubt – incredible compilation of data. The authors use cost and production data from 62 technologies!

      While their choosing ‘unit cost’ as the figure of merit enabled the authors to aggregate data across those 62 technologies, it comes at a cost (pun intended…) Unit cost is, well, cost. It is not a measure of performance of the technology. Hence, their remarkable results, unless augmented by corresponding performance data, are valid only for this particular figure of merit.

      A classical applications of this kind of study is cost v. performance analysis for a family of systems that will be implemented using a specific technology. Such analysis would not be possible based on this paper alone.

      All in all, I am very impressed with the paper: it could advance the state of the art in predicting technological progress. Having said that, I am advising my readers to make certain they apply the results of this study in accord with the postulates of this paper.

      - Israel

  2. avatar

    Andrew McAfee wrote on a related topic last year with some charts on computer asset prices and US IT investments (see http://andrewmcafee.org/2011/01/jevons-computation-efficicency-hardware-investmen/). A couple of charts make it clear what is going on here.

 Leave a Reply

(required)

(required)

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>