Technical Footnotes/Explanation on Sustainability:



Nuclear Safety: The main safety concern here is the availability of nuclear materials and technology to erratic subnational groups of all kinds; however, there is also a possibility of larger than expected damage from accidents which could occur under third world safety regimes, when there is pressure to hold down costs. These concerns really are a matter of sheer survival. Plausible scenarios for what might happen in the wake of certain events are not reassuring.


Stuck in a Nash Equilibrium: When each player acts to maximize personal advantage, without considering the bigger picture, all players may suffer. “Prisoner’s dilemma” is the classic example. In economics, people commonly say “A Nash solution is often not a Pareto optimum.” That means that there is a “win-win” possibility, an outcome which would benefit all players, if we could figure out a way to get there. Some old politicians used to say: “If it’s a bigger pie, everyone can have a bigger slice.”  

In the specific case of energy, for example, I try hard to explain how a more forward-looking, sustainable strategy can be tuned to ensure benefits to all major players, oil users and suppliers both – but in practical situations this can be difficult. When time is limited, oil consumers may want to hear about their part of the story.  And there are those who think their job depends on shooting or projecting a false image and asking questions later if ever.

Von Neumann, the father of game theory, would probably have viewed the work of Thomas Schelling and of Howard Raiffa as really essential in creating the kinds of insights needed to get us to a Pareto optimum. More formal mathematical theories of dynamic games vary in their ability to do justice to such insights; many in game theory now believe that one must pay more central attention to the role of learning, for which we now have much more powerful tools. The specialist might note that most of the tools developed to approximate the Bellman equation carry over easily to the Isaacs equation – but the real challenge concerns  visions, designs and strategies for interaction which get beyond the Nash equilibrium to something more like a global optimum.

In the world of society and economics, as in the world of neural networks, change is much easier when an incremental path can be found. Incremental adaptation schemes based on derivative or value signals propagated through a large system are essential to even the mouse level of intelligence, I would claim. But it is also important at times to have a more global “imagination” system (to provide coherent analysis of global scenarios not yet experienced in past history) and a “creativity” system (to escape from “local minimum” or “nonconvexity” problems, related to what Soros calls “reflexivity” in economic systems).


Checking Atomic Elements: Could Any of Them Run Out? Though I am not an expert on that topic, I have asked a colleague for a few quick words to be a starting point for those who wish to look further. He writes, informally:


“The work by Graedel is probably the most complete in the area of metal cycles and the overall consumption of metals by various regions.  I attach the papers on Copper, Zinc and Silver, which are all thought to be a possibility for scarcity.
Also a really good source for minerals is the usgs, for example their 2006 surveys are linked below, and there are others.  I include the platinum and palladium metal group also which they survey.  You can find historical data going back on these which you could use to assess the accuracy of their earlier projections versus the current data.
Personally I am not so worried about metals, they are readily fungible, the more difficult one is water which is not and which tends to cross borders with an unreasonable abandon, and which we have figured out remarkable engineering feats to stop doing so, with enormous potential political consequences.”


Fiscal and Political Sustainability: This is a large topic which I hope to post more on in the future. For the moment, I should note that my Harvard PhD thesis was not just about backpropagation (as discussed in the page on neural networks). It was really four pieces in one – a piece on backpropagation or reverse differentiation; a small piece just for me (chapter 5); a piece on new ways to fit causal models better to time-series data, to get better forecasts and better understanding than with classical statistical methods; and – most of all – the use of these methods to create an operational version of a model of “war and peace” developed by my thesis adviser, Prof. Karl W. Deutsch. (The thesis has been reprinted in its entirety as part of the book The Roots of Backpropagation: From Ordered Derivatives to Neural Networks and Political Forecasting, by me, Wiley, 1994.) Deutsch was a fascinating person, with a very visible and powerful intuition which he used every day of his life – even though he fervently disbelieved in the very idea! I don’t think he even knew what a neural network was, but his book, the Nerves of Government, provides a fascinating account of how one might try to understand political and social phenomena as a kind of neural network – but he also warned that we shouldn’t take the idea of collective intelligence in the wrong way, and he recommended that we also read Janis’s book Groupthink. He saw my thesis as the final vindication of the ideas he presented in his book Nationalism and Social Communications. He was at that time president of the International Political Science Association, and one of the key thinkers behind the creation and strengthening of the European Union.

          For the time being, I will post just one of my papers related to these themes:


P. Werbos & J. Titus (1978) An empirical test of new forecasting methods derived from a theory of intelligence: The prediction of conflict in Latin America, IEEE Trans SMC, September.

          This work taught me many, many lessons. One was that a lot of earlier empirical work on conflict – at the level of macro data – was grossly misleading, because it was dominated by two data points (typically World War I and World War II). Some variables can be predicted relatively robustly, like some of the key determinants of conflict. But conflict itself is much harder. Certain kinds of data on content in communications network data might do much better, but no one on earth has been working the combination of methods and disciplines that would be necessary to do so. For the past many years, I have relied more on other approaches to understand these things – as in chapter 5 of the thesis, and in applying the kind of conceptual understanding that other partsof this web page provide.