Take the pledge to vote

For a better tommorow#AajSawaroApnaKal
  • I agree to receive emails from News18

  • I promise to vote in this year's elections no matter what the odds are.
  • Please check above checkbox.

    SUBMIT

Thank you for
taking the pledge

Vote responsibly as each vote counts
and makes a diffrence

Disclaimer:

Issued in public interest by HDFC Life. HDFC Life Insurance Company Limited (Formerly HDFC Standard Life Insurance Company Limited) (“HDFC Life”). CIN: L65110MH2000PLC128245, IRDAI Reg. No. 101 . The name/letters "HDFC" in the name/logo of the company belongs to Housing Development Finance Corporation Limited ("HDFC Limited") and is used by HDFC Life under an agreement entered into with HDFC Limited. ARN EU/04/19/13618
LIVE TV DownloadNews18 App
»
1-min read

AI Discovers Something Fascinating in Old Scientific Papers that Humans May Have Missed

What makes it more intriguing is that the algorithm did not know the definition of thermoelectric and received no training in material science.

Trending Desk

Updated:July 12, 2019, 3:03 PM IST
facebookTwitterskypewhatsapp
AI Discovers Something Fascinating in Old Scientific Papers that Humans May Have Missed
What makes it more intriguing is that the algorithm did not know the definition of thermoelectric and received no training in material science.
Loading...

The powers of AI keep on amazing us every day. In a new study, published in Nature, on July 3, researchers revealed that using the language in millions of old scientific papers, a machine learning algorithm was able to make completely new scientific discoveries.

According to experts from the Lawrence Berkeley National Laboratory, they used an algorithm called Word2Vec to sift through scientific papers for connections humans may have had missed.

Surprisingly, the algorithm spat out possible thermoelectric materials, which convert heat to energy and are used in many heating and cooling applications.

What makes it more intriguing is that the algorithm did not know the definition of thermoelectric and received no training in material science.

Turns out, the AI was able to provide candidates for possible future thermoelectric materials, using only word associations.

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

Jain further revealed that the study shows if the algorithm was in place earlier, some materials could have conceivably been discovered years in advance,

The researchers assessed the language in 3.3 million abstracts related to material science, thus ending up with a vocabulary of about 500,000 words. They fed the abstracts to Word2vec, which used machine learning to analyse relationships between words.

According to Jain, by training a neural network on a word, one can get representations of words that can actually confer knowledge. Using words found in scientific abstracts, the algorithm was able to understand concepts such as periodic table and the chemical structure of molecules.

The algorithm linked words that were found close together, thus creating vectors of related words that helped define concepts. In some cases, words were linked to thermoelectric concepts, but had never been written about as thermoelectric in any abstract the team surveyed.

The team tested the algorithm on old papers, seeing if it could predict scientific discoveries before they happened and found that, in one experiment, the AI was able to predict thermoelectric materials a few years before it was actually discovered in 2012.

The lead author of the study, Vahe Tshitoyan, said, "This algorithm is unsupervised and it builds its own connections," adding that one can use it for things like medical research or drug discovery.

Read full article
Loading...
Next Story
Next Story

Also Watch

facebookTwitterskypewhatsapp

Live TV

Loading...
Countdown To Elections Results
To Assembly Elections 2018 Results