2018 Year in Review
IARPA's 2018 Fiscal Year Highlights:
- 5 new multi-year research programs covering diverse technical fields, including information extraction and retrieval, chemical analysis systems, superconducting electronics, unmanned aerial vehicles, and synthetic biology
- 15 single-year seedling studies researching topics, such as unattended ground sensors, high performance computing, satellite imagery, and automated latent fingerprint recognition
- 7 new public prize challenges offering cash awards to innovators across the globe
- 4 requests for information to inform on the state-of-the-art in credibility assessment, marine and coastal biosecurity, chemical sensing devices, and nuclear material collection
- 41 technical workshops with over 2,200 attendees
- 39 transition agreements to transfer IARPA-funded technologies to other government agencies
- Over 450 research publications from current and past IARPA programs, uploaded to the Defense Technical Information Center registered user site. This includes approximately 75 publications available to the public through DTIC’s public user site.
IARPA-funded research highlights:
The Babel program reported 150 organizations (44 U.S., 106 International) were granted access to the program’s speech recognition training data. These datasets, hosted by the Linguistic Data Consortium, consist of transcribed conversational speech in 25 languages, and continue to be a veritable gold mine for speech and machine learning research and development.
The Dual Frequency Optical Combs seedling developed four different fiber laser designs that allow two independent comb laser beams to be created in a single fiber chain, an enabler for cheap, portable dual comb spectroscopy systems that can provide extremely high sensitivity for detection of chemicals in the gas phase.
The Fun GCAT program advanced biosecurity capabilities by building a new database containing over 5,000 DNA sequences of concern and associated biocuration tools for automated ingestion of sequence information from published research. The database is being used to train artificial intelligence tools to conduct threat assessments and functional predictions of DNA sequences.
FELIX program performers generated, collected, and simulated tens of thousands of engineered organism sequences to develop experimental and computational tools to distinguish engineered and naturally occurring organisms. Initial data suggests the detection capabilities may be more sensitive than the current gold standard and new detection platforms may provide improved screening methods.
The Geopolitical Forecasting Challenge produced forecasting methods that improve the state-of-the-art of integrating crowdsourced forecasts and other data into accurate, timely forecasts on worldwide issues.
The GHO Battery Powered Flight study successfully flew a 450 pound unmanned aerial vehicle quieter than the study goal. It was not heard on the ground at an altitude that was less than 2% of the current US Army UAV altitude of the same size. (images available)
The LogiQ program successfully demonstrated all building blocks of a logical qubit, thereby significantly improving our understanding of multi-qubit systems intended for fault-tolerant quantum computing. This has also led to the development of the most complete quantum error models and quantum benchmarking protocols to date.
The Micro-Gas Chromatography seedling demonstrated chemical separation using voltage rather than heating, dramatically reducing power draw and eliminates the cycle-reset time before the column can be used again.
The MICrONS program continued to acquire the largest-ever maps of neocortical circuit structure and function to inform the development of next-generation machine learning algorithms. In parallel with this work, MICrONS performers continued to build seven startup companies with over $30M of private funding to commercialize technologies developed through the program.
The SILMARILS program demonstrated the ability to detect and identify trace chemical residue on roughened aluminum and glass surfaces at a 25 meter standoff distance, detection of heavy fingerprint residue on a car panel at 24m, identification of aerosolized chemical warfare agent simulants at levels <1 parts per million, and identification of trace narcotic residues inside plastic bags.
The QEO program discovered the first evidence supporting a potential scaling advantage in quantum annealing over classical computing on certain applications, with preliminary estimates of the enhancement factors up to 100x. World-record coherence times for annealing-compatible qubits were also demonstrated.
The UG2 Prize Challenge focused on image enhancement of imagery and video, and highlighted how challenging this research area is. Identifying only modest performance, this prompted the challenge to be run for a second year (UG2+ Prize Challenge) to provide the research community an opportunity to continue working on this problem.