Defense Advanced Research Project Agency is looking for an artificial intelligence and machine-learning model that can help scientists and researchers push their work to new limits.

The Automating Scientific Knowledge Extraction (ASKE) program, announced Aug. 17, is the first contract opportunity DARPA has released as part of its new AI exploration program. The goal is to establish the feasibility of new AI concepts and do it fast ― within 18 months of award ― to help DARPA outpace global AI science and technology discovery efforts.

Specifically, the ASKE opportunity is looking to develop an AI system that can rapidly aggregate scientific data over a number of complex systems (physical, biological, social) and identify new data and information resources automatically. Science depends on equations and complex computations of large data sets. The proposed AI system would be able to interpret and expose scientific knowledge and underlying assumptions in existing computational models to extract useful information, like causal relationships, correlations and parameters. This information would then be integrated into a machine-curated model that generates more robust hypotheses.

To ensure the system is working with the full-breadth of scientific information available, DARPA is interested in a system that automatically verifies published scientific results and can monitor “fragile economic, political, social and environmental systems undergoing complex events,” in real-time. For such a system to be viable, DARPA believes advanced AI techniques such as “natural language processing, knowledge-based reasoning, machine learning, and/or human-machine collaboration” are needed.

Although rapid and real-time aggregation of data from a variety digital sources may have military applications, for now DARPA maintains its “overriding interest is in innovative approaches to extracting knowledge from scientific models.”

The winner will be awarded a contract worth as much as $1 million for a prototype. Proposals are due Sept. 17.

Daniel Cebul is an editorial fellow and general assignments writer for Defense News, C4ISRNET, Fifth Domain and Federal Times.

More In IT and Networks