Anticipating Rare Events of Major Significance: Eavesdropping on the Adult Table
Being retired, healthy and affluent, I could just play around like a Big Kid, but one should sometimes still join the adult table. This article reports on a meeting on the title topic (Anticipating Rare Events of Major Significance)
which consisted of talks by 12 experts followed by question-and-answer sessions with the organizing committee (which included myself). A detailed summary report (28 pages of content) has recently been
published and is freely available here as Proceedings.
Medium contains many articles about future risks, mostly rather speculative. Readers wishing to engage this field more seriously are invited to read this report, to discover how actual experts think about some specific topics. Hence my subtitle “Eavesdropping on the Adult Table”.
Note that the report format is different from the familiar style of committee report which seeks a consensus view and therefore tends to be rather bland. Instead the talks combine background facts with discussion of policy implications, and of course we can all have different opinions regarding the latter.
The meeting was convened under the auspices of the (rather James Bond-ish sounding) U.S. Defense Threat Reduction Agency (DTRA). Despite that name, the content is essentially not about military intelligence: rather DTRA was seeking an overview of how the civilian side thinks about such matters. The title word anticipation is important. The goal is not merely to observe that a risk exists or estimate its likelihood, but more importantly to make advance preparations to reduce the likelihood or mitigate the effects.
This post seeks to convey the flavor of the report via fragments from some of the 12 speakers. One can hardly generalize across very different risks, but the following conceptual framework is common.
There are real-time sensors. That is, ways of getting data — often massive and noisy data, or data primarily intended for other uses — suggesting that a risk event may be happening or is becoming more likely to happen in the near future. This information has to be conveyed to the people who need to take real-time decisions. And one might consider using AI to help decision-makers deal quickly with a flood of data. Of course all this needs much advance planning and organization.
Samples from the report:
(1) As a conceptually simple example, speaker Arvind Satyam notes that climate change has increased the frequency of megafires: “California, for example, has had six of the ten largest fires in the state’s history occur since August 2020.” Here one wants sensors to detect fires quickly.
“Our approach was to deploy pairs of ultrahigh-definition rotating cameras and use software to simulate the experience of a person sitting on a traditional fire watchtower. The result is an AI system that analyzes the continuously rotating image to spot texture, movement, and gradients in the images indicative of the smoke plume emitted by even the smallest fire.” This system is already being deployed. “It not only aids the local fire authority, but also enables the local utility to understand where the incident is relative to its transmission lines”.
(2) Speaker Alice Hill emphasized that localized climate predictions are needed by smaller cities and rural areas, to plan adaptations to mitigate future extreme weather events. Indeed such adaptation is well described as ““under-resourced, underfunded, and often ignored.”
(3) Speaker Nestor Alfonzo Santamaria discussed disaster scenario planning in different OECD countries. In this context, he noted it seems wise to use “an exercise that plans for common consequences of various incidents, such as disruption to utilities, transportation, and health care, rather than having in place specific arrangements for responding to specific risks”.
(4) Speaker Madhav Marathe discussed analysis of “the first 3 days” after a disaster cuts transportation, communication and power networks in a large city. One conclusion was “even a partially restored communication system has a disproportionately positive effect on overall behavior and reducing anxiety. Partially restoring communication can be done today … using what is known as a cell on wheels, which are networks that can be brought up quickly.”
(5) Speaker Charles Clancy gave a high-level account of “projects with IARPA aimed to leverage emerging big data capabilities and advances in AI and machine learning to develop anticipatory analytics for predicting the next war and forecasting events such as a terrorist attack, social unrest, and local election outcomes.” He acknowledged that predicting rare events was difficult. For example “(There is a) large gap in the ability of AI to do contextual and causal reasoning that would enable AI to tackle the small data problem. Some in the AI community have suggested using AI to detect anomalies, but the problem is that an analyst still needs to look at the anomaly to determine whether it is a false positive. Cyber-security uses this approach, but it generates so many false positives that analysts do not find them useful because it creates too many leads for them to examine. He added that anomaly detection is not likely to be useful for rare events because of the large number of anomalies that occur every day.”
(6) Speaker Elisabeth Paté-Cornell discussed a general model for optimizing a warning threshold to avoid giving too many false positives or false negatives, and to take account of the fact that people may react differently based on the past accuracy of warnings. An example concerned monitoring satellites in low Earth orbit to avoid collisions with debris.
This is currently done based on combining information from two separate sources, U.S. Space Surveillance Network and the international scientific optical network. “As an illustration, they compared the options to spend $100 million to add one sensor to the U.S. Space Surveillance Network or to add 35 sensors to the international optical network at the same cost”.
(7) Speaker Jeffrey J. Love discussed geomagnetic storms, which (perhaps once or twice a century) will be strong enough “to cause significant damage to and interference with military satellites; widespread disruption of GPSs, radio communication, and geophysical surveys; widespread and prolonged loss of electricity resulting from damage to the electric grid; and a possible economic impact to the United States of $1 trillion to $2 trillion”. The storms are of course not preventable, but Love and his collaborators are working to create real-time geoelectric hazard maps, as an important tool for managing the power grid system during an intense magnetic storm. Speaker John Organek continued this theme by describing work aimed at the developing the ability to re-start the electrical grid after a widespread outage.
(8) Speaker Seth Baum discussed global catastrophic risks, in particular a model for the probability of nuclear war. “The model outlines two paths to nuclear war, one in which there is an intentional first strike, the other involving unintentional or inadvertent nuclear war scenarios …. For instance if the victim of a terrorist attack country misinterprets the attack as one from a rival country or one sponsored by a rival country and launches its own nuclear weapons in what it believes is retaliation, but is in fact, the start of a nuclear war.”
