IARPA RFI – Teasing Future Performance from Historical Data

Home / Articles / External / Government

dcma_connections

June 24, 2019 | Originally published by Date Line: June 24 on

Exploring methods to predict the future performance of critical technologies will require historical data on technologies’ performance over time, ideally over the span of decades.

The Intelligence Advanced Research Projects Activity (IARPA) believes they may be able to predict the future performance of critical technologies by analyzing developments of their constituent systems and components over time in an objective and quantitative manner. To help explore development of methods to accomplish this, IRAPA is looking for datasets and metrics that document performance trends of a constituent technology, system or component with at least a decade of historical coverage and accurate information on when the technology was first developed. All technology areas are of interest, particularly those related to:

Artificial Intelligence
Collection & Sensing
Communication
Computing
Cybersecurity
Energy & Power
Human Performance Modification
Materials
Quantum Sciences
Space Sciences

As an example, under Energy & Power, IARPA cites advances in capacitor technology performance as an indicator for the ability to deliver high pulsed power.  Capacitor technology can effectively be represented by the performance metric of power density. “As capacitor technology improves, increasing power density is a good indicator of performance improvement, and ultimately historical improvements in power density may be used to predict future improvements in power density.”

Solicitation Link

FedBizOps Link: RFI for Seeking Data Sources for Technology Performance Prediction

Solicitation Number: IARPA-RFI-19-07
Agency: Office of the Director of National Intelligence
Office: Intelligence Advanced Research Projects Activity

Response Date: Aug 06, 2019 4:00 pm Eastern

Synopsis: IARPA requests information on datasets that document the historical performance of technologies. Datasets must have historical coverage of at least a decade and, most importantly, have accurate information on when the technology was first developed. For data about products, first development of the technology means “first offered for sale”. For data about pre-commercial or non-commercial technologies, date of journal publication or other public disclosure will be defined as first development. A temporal resolution of at least 1 year is needed for any data occurring after 1900; earlier data may have coarser temporal resolution. Performance metrics for technologies. These must be objective, quantitative, and recoverable from the historical record. Many performance metrics will likely be multi-dimensional and represent tradeoffs in some aspect of the technology (e.g., FLOPs per Watt).