Our minds are a series of neural networks composed of interconnected neurons that store and create information. As technology develops, we have begun to develop software that mirrors these neural networks.
The ability to discern between the intricacies of shape and colour, such those between a ripe and rotten apple, without any prior knowledge of fruit, is one of the landmarks of this research. Casting such software upon crisis response brings a new asset to surveillance. Using the software to identify victims, hostile units, and potential events from images of the crisis would streamline response times and free resources.
The brain works by establishing a hierarchy of neural levels that allow for complex and abstract thinking. Software developers used this concept to create algorithms that communicate in a similar fashion. Each algorithm searches for distinct features or patterns and passes it on to the next algorithm that searches for something else. As a distinct attribute is found, the software learns about the possibilities of what the solution to the input may be.
For example, when processing images, the software first finds similar pixels, then distinguishes the groups of pixels into shapes and finally into physical features such as hair or grass. With the wide availability of organised data from search engines, the software can learn and correct itself when the end user either agrees or disagrees with the solution.
If you type “apple” into the Google Photos program it searches your collection of photos for apples and displays them. Without any preparation or previous tagging, Google Photos can identify apples from a unique photograph based upon the similarities of apples in its databank to the ones in the photograph. This level of identification can be brought further to determining if the apple is rotten or ripe based upon similarities of colour, wrinkles or shape.
Unfortunately, even with the vast amount of data to pull features and patterns from, the software still makes errors about 25 per cent of the time. Accuracy can be improved by user feedback and the software can learn from its mistakes by identifying what doesn’t make an apple through similarities in its mistakes.
Bringing this software into crisis response opens many possibilities for improving search and rescue operations. With a photograph of the impacted area, the software could identify survivors or impending danger. Searching for key features such as people, weapons and distress signals, the software could add an additional level of scrutiny and expedite recovery operations. Furthermore, the software could process hundreds or thousands of hours of video footage to help prevent crisis events before they begin. In the right hands, this software could improve the public’s safety and ability to respond in times of crisis.