A recent study that was published in Nature has revealed that the artificial intelligence (AI) that is in the form of a machine that is learning hidden scientific knowledge which is algorithm uncovered and was missed by scientists in the old research papers. The conducted experiment has resulted in the finding of potential thermoelectric materials that are being utilized in heating as well as cooling applications.
Although the alogrithm did not receive any training in the materials science, it was not even aware of the definition of thermoelectric, but had the ability to make its discoveries by utilizing the word associations. The paper states that the discoveries had the ability to uncover materials that were better than the already existing thermoelectric standards.
Anubhav jain, researcher as well as senior author of the paper said that it could read anything on material science, and hence has the potential to make connections that no scientists can. He further added that sometimes it would do what a researcher would, and at other times it made these cross-discipline associations.
Before this discovery was made, the research team would feed the algorithm almost five-hundred thousand words that the team thought were important to the material science. Later, the Artificial Intelligence began to learn the relationships between each of the words. For instance, the neural network that was used to train it, could remove a word from one of the sentence and then teach the AI on predicting the word after it.
The team has reported that this training procedure greatly helped the AI in learning how to understand the scientific concepts like the periodic table and the molecular structure.
In addition to the materials science, the algorithm can also be used in various other areas of research.