While many users still marvel at ChatGPT or Midjourney, the scientific race to create AI-based tools is moving forward. The latest of these creations aims to analyze something particular: sand.
SandAI. A team of researchers at Stanford University recently unveiled a new AI tool: SandAI, a neural network designed to study sand. According to the experts, it will help them learn how grains of sand get into a given environment.
The developers say SandAI will make it easier for geologists and archaeologists to analyze strata where other “clues,” such as fossils, are lacking. They even point out that forensic investigators could use it in analyses related to illegal sand mining.
“Working on sedimentary deposits that haven’t been disturbed or deformed feels about as close as you can get to being in a time machine–you’re seeing exactly what was on the surface of Earth, even hundreds of millions of years ago. SandAI adds another layer of detail to the information we can pull from them,” Michael Hasson, a member of the research team, said in a press release echoed by SciTechDaily.
“Telltale signatures.” SandAI is a tool that studies microtextures, such as the surface of sand grains, to infer their history.
For example, it can determine the transport mechanisms of sand grains through “telltale signatures” on their surface. It can also tell experts which sand grains traveled the farthest before reaching a surface.
Microtexture analysis is very popular in this context. However, “manual” examination presents some problems, such as the time it takes to carry out or the presence of a certain degree of subjectivity. The team responsible for the development of SandAI emphasizes that the new tool allows these studies to be “more quantitative, objective, and potentially useful across a wide range of applications.”
Microtextures of a grain of sand
Training SandAI. Like any AI tool, SandAI had to be trained by its developers. Numerous researchers assisted the team with images of sand grains extracted in various contexts. As EurekAlert! reports, the goal was to have a representative database of the major terrestrial environments, such as fluvial, aeolian, glacial, and coastal areas.
The database also covered different geological periods. SandAI achieved high accuracy in tests with samples up to 200 million years old. That is, with sand preserved from the early Jurassic period. In later phases, the tool was able to identify the origin of older grains up to 600 million years old.
According to its creators, SandAI achieved a prediction accuracy of 90% during its development. The research team detailed their work in a paper published in Proceedings of the National Academy of Sciences.
SandAI is an online tool available on GitHub for researchers around the world. However, the work continues. The team explains they’re still working on the tool based on user input.
“The fact that we can now offer detailed conclusions about geological deposits that weren’t knowable before, I find kind of mind-blowing. We’re looking forward to seeing what else SandAI can do,” Hasson concludes.
Image | Mathieu Lapôtre, Michael Hasson
View 0 comments