NASA engineers came up with a new model, which itself decides what kind of research to do and what data to send to scientists. This approach will save tens of millions of dollars.
Scientists from the NASA Goddard Space Flight Center have announced the first test results of new intelligent systems that will appear in space probes – for example, they will be able to determine the geochemical signs of life from rock samples. At the same time, the systems themselves will choose what they analyze and what information according to the results of the analysis they send to Earth. They will begin to work in 2022 or 2023 on celestial bodies more distant from the Earth.
Researchers have explained that sending data to Earth is very long and expensive. Therefore, scientists cannot conduct an unlimited number of experiments and analyze as many samples as they want. Using artificial intelligence (AI) to initially analyze data after it is collected, NASA can optimize the size of the data without sacrificing its quality.
“This is the right step for long-term space exploration. We are gradually moving away from the idea that a person should do everything. In fact, it’s not there anymore, computers are equipped with intelligent systems, they are trained to make decisions in priority order and highlight the information that scientists need to show”.
Victoria Poyan, NASA
The team used the raw data from initial laboratory tests with the MOMA ground instrument to train computers to recognize useful information. At the same time, when receiving new raw data, the software informs scientists what data they can relate to.
The first results show that when a neural network algorithm of a system processes raw data, it can classify with an accuracy of 94%. He can connect new data with old in 87% of cases.