Is artificial intelligence a threat? Experts weigh the risks
July 02, 2015
As several tech luminaries express worries about the future of AI, researchers met at a D.C. think tank to discuss whether advances could pose a threat.
Travis Korte, analyst at the Center for Data Innovation, is a FedScoop contributor.
Weather data saves lives. Better sensors, along with more sophisticated models and greater processing power, have helped sharpen the accuracy of the United States' hurricane warnings 10 times over since the 1990s. Aside from its government uses, weather data is also the cornerstone of many important commercial applications. Public utility companies use it to plan for likely outages, farmers use it to inform planting decisions and air traffic controllers use it to keep planes safe and running on time.
One of the most important sources of weather data is satellite-mounted instruments, which collect visible and infrared observations of the Earth and its atmosphere; temperature, pressure and moisture data in three dimensions -- ozone-level readings, radiation-level readings and other data.
But a series of planning failures, delays and cost overruns in the development of a next-generation U.S. government-owned Earth observation satellite has put applications that rely on high-quality weather data at risk, heralding a likely gap in some of the country’s critical weather data-collection efforts.
The U.S. relies on satellite-collected weather data from two sources. One source is geostationary satellites, which orbit above a fixed point on the planet's surface. Geostationary satellite data is highly granular, and accounts for short-term weather forecasts of the continental U.S. The second source is two polar-orbiting satellites, which orbit at an angle perpendicular to the equator and record data from the entire planet twice over the course of a single day as the earth rotates beneath them.
Polar-orbiting satellite data is used for large-scale numerical weather prediction, and feeds three- to seven-day forecasts. The U.S. weather modeling efforts rely on data collected by polar-orbiting satellites twice a day, in the morning and afternoon. The projected gap would affect the afternoon orbit; although, the morning orbit would still be covered, a gap in the afternoon would mean weather data supporting critical U.S. applications would only be available half as often, and less frequent data means less accurate predictions.
For example, the National Weather Service conducted an analysis to estimate how much its forecasts of the 2010 mid-Atlantic "Snowmageddon" storm would have been off without afternoon polar satellite data, and found its models would have underestimated the amount of snow by at least 10 inches.
The satellite currently tasked with collecting the afternoon data is the Suomi National Polar-orbiting Partnership spacecraft, which was launched in 2011 and whose useful lifecycle is expected to last until in 2016, after which the likelihood of its failure exceeds that of its continuing to operate. The National Oceanographic and Atmospheric Administration has scheduled the next-generation satellite, the Joint Polar Satellite System, to launch in mid-2017, but it must undergo several months of testing before it is operational. Future delays or equipment failures could increase the length of the gap; a NOAA review anticipated the gap would likely last between 18 and 24 months, with some scenarios suggesting a gap of up to 53 months.
Just how likely is a gap? Estimates differ: Dr. Kathryn Sullivan, assistant secretary of commerce for environmental observation and prediction, deemed the gap "near-certain," while Mary Kicza, NOAA assistant administrator of satellite and information services, estimated the probability of a gap to be 50 percent. The discrepancy arises from the fact that satellite lifetime estimates are inexact. The Suomi-NPP satellite was designed to operate for seven years; if the craft continues to function until the end of its design life, the chances of a coverage gap will be greatly diminished. In practice, however, satellite operational lifetime estimates are typically years shorter than the lifetimes for which they were designed, due to external risks such as rocket failures and satellite component malfunctions.
So what can be done? In October 2012, NOAA issued a plan for mitigating a gap of up to 18 months. The plan details options for collecting critical data, both in terms of alternative data sources (such as other countries’ polar-orbiting Earth observation satellites) and improved modeling to derive more insights from data that will likely be available. Several lawmakers balked at the idea of piggy-backing on a foreign-government’s data, particularly as China's Feng-Yun 3 satellite is the leading candidate among foreign data sources, citing fears that China could threaten to withhold the data or even provide fake data in the event of a disagreement with the U.S.
But given the importance of weather data, if a gap should occur, this may be the best option. This looming threat should be a wakeup call to policymakers of the severe lack of redundancy in the U.S. weather observation data infrastructure and the serious consequences of failure.
Long term, policymakers should consider the costs and benefits of transferring more of the collection of weather data to the private sector, such as by purchasing weather data from commercial providers and sharing satellite infrastructure (it is unlikely that the private sector can supplant NOAA data for the projected 2016 gap). Legislation, such as H.R. 2413, the Weather Forecasting Improvement Act, introduced by Rep. Jim Bridenstine, R-Okla., would enact some of these reforms.