Is there any activity that resists artificial intelligence? Less and less every day, actually, and although there is still no general-purpose AI similar to human AI, the developments dedicated to all kinds of specialties show us, day by day, that there seem to be few areas that are beyond its reach, and uses that a few years ago might have seemed crazy, today have been shown to be fully functional.

The latest example of this can be found in the project of Leila Character, a PhD student in Geography at the University of Texas at Austin College of Liberal Arts who, in collaboration with the Underwater Archaeology Branch of the U.S. Navy, has developed an artificial intelligence capable of recognizing shipwrecks on the seabed from images taken from the surface (by ships) and aerial photography.

As is usually the case with learning an artificial intelligence algorithm, the first step in this project was to collect images of already identified wrecks, using public information from the National Oceanic and Atmospheric Administration (NOAA), the U.S. scientific agency that monitors oceanic and atmospheric conditions. NOAA is responsible, for example, for the NHC (National Hurricane Center), the reference entity for hurricane information in the Atlantic and Pacific.

A key element at this point was that the artificial intelligence was able to distinguish between shipwrecks and the topography of the seabed, so it was also necessary for it to learn to recognize the ocean floor, so the algorithm was also generously fed with images of it, covering everything, or at least much of what can be found there, and what it looks like in this type of images.

See also  The Twitter font has changed

Inteligencia artificial para explorar el fondo marino

The result of putting this artificial intelligence to work after training was to get an accuracy of no less than 92%. Of course, this may seem little, if we take into account that we are talking about the analysis of images of sinkholes already identified in a specific region, but the potential it offers when analyzing larger regions that have not yet been explored is undoubtedly most remarkable.

And it has even more value, since with the necessary modificationsartificial intelligence such as this could be useful for many other purposes. For example, the search for the black box of Air France Flight 447 was a real race against the clock, to the point that if it had taken any longer, the data from the voice and flight data recorders could have been lost forever, making it impossible to investigate what had happened.

Worse still is the case of Malaysia Airlines Flight 370, which remains a huge unknown more than seven years after its mid-air disappearance. Although over the years some wreckage supposedly from the Boeing 777 has surfaced, it is likely that much of it still lies on the ocean floor, so an artificial intelligence dedicated to locating unnatural debris on the ocean floor could finally locate the plane and try to find out what happened.

Read more: The Conversation