| Middle East Object Footage taken by an MQ-9 of an apparent silver, orb-like object in the Middle East. Due to limited data, the object remains unidentified. The appearance of U.S. Department of Defense (DoD) visual information does not imply or constitute DoD endorsement. |
 |
 |
4 Which scientific analysis techniques currently in production could be employed to assess the nature and origins of UAP? Which types of analysis techniques should be developed?
FINDING
AI and ML, combined with NASA's extensive expertise, should be utilized to investigate the nature and origins of UAP.
Artificial intelligence (AI) and machine learning (ML) have proven to be essential tools for identifying rare occurrences within vast datasets. These methodologies, combined with NASA's extensive experience and expertise, should be utilized to investigate the nature and origins of UAP by examining data from sources such as satellites and radar systems. However, the effectiveness of AI and ML in studying UAP depends critically upon the quality of the data used to train the AI and in subsequent analysis. At present, UAP analysis is more limited by the quality of data than by the availability of techniques. As a consequence, it is a higher priority to obtain better quality data than it is to develop new analysis techniques.
Once AARO and other agencies, including NASA, accumulate an extensive and well-curated catalog of baseline data, these can be used to train neural networks so that they can characterize deviations from normal. The panel finds that standard techniques that are routinely applied in astronomy, particle physics, and other areas of science can be adapted for these analyses.
When it comes to detecting anomalies–such as UAP–within datasets, there are two approaches. The first approach involves constructing a model that represents the expected signal characteristics then searching for any matches against this model. The second approach involves using a model of the background properties and searching for anything that deviates from that model. The panel notes that the first approach is difficult as we do not possess a consistent description of the physical characteristics of UAP. The second
15
