Artificial Intelligence Made New Scientific Discoveries From Old Research Papers

A tablet is seen in the center of a high tech art installation at Salt Galata on May 6, 2017 in Istanbul, Turkey.
Chris McGrath / Getty Images

A recent study published in Nature reveals that artificial intelligence (AI) in the form of a machine learning algorithm uncovered hidden scientific knowledge that was missed by humans in old research papers. The experiment was conducted at the Lawrence Berkeley National Laboratory using the Word2Vec algorithm and resulted in the discovery of potential thermoelectric materials, which are used in heating and cooling applications.

Although the algorithm didn’t receive training in materials science — it didn’t even know the definition of thermoelectric — it was able to make its discoveries by using word associations. According to the paper, the discoveries might have uncovered materials that are better than the current thermoelectric standards.

“It can read any paper on material science, so can make connections that no scientists could,” said researcher Anubhav Jain, senior author on the paper. “Sometimes it does what a researcher would do; other times it makes these cross-discipline associations.”

Prior to the discovery, Vice News reported that the research team fed the algorithm about 500,000 words that the team deemed important the material science. Afterward, the AI began learning the relationships between each word — each of which stemmed from over 3.3 million abstracts. For example, the neural network used to train it might remove a word from a sentence and teach the AI how to predict the word after it.

The team reports that this training procedure helped the AI learn how to understand scientific concepts such as the periodic table and molecular structure.

In addition to materials science, the algorithm could be used in other areas of research.

“This algorithm is unsupervised and it builds its own connections,” said Vahe Tshitoyan, senior author on the study.

Loading...

“You could use this for things like medical research or drug discovery. The information is out there. We just haven’t made these connections yet because you can’t read every article.”

The use of AI to conduct procedures in the place of humans has been a source of both fascination and fear for people in the United States. CNN reported that Democratic presidential candidate Andrew Yang has built his platform on the effect of AI on worker displacement. According to the report, he suggested that the country is in the midst of the “greatest economic transformation in our history.”

“This may seem like science fiction to you. But you are reading this on a supercomputer, Trump is our President and your local mall might have just closed. We are living in unprecedented times and must rise to these 21st century challenges.”