stub AI Can Be Trained To Independently Make Scientific Predictions Based On Previous Knowledge - Unite.AI
Connect with us

Artificial Intelligence

AI Can Be Trained To Independently Make Scientific Predictions Based On Previous Knowledge

mm
Updated on

There is an ongoing debate among AI researchers whether Artificial Intelligence, as TheNext Web (TNW) notes, “will soon be able to develop the kind of general intelligence that humans have,” with heated arguments for and against.

But there is yet another field of knowledge where AI is making giant steps forward, and that is with Natural Language Processing (NLP), a part of a much larger umbrella of machine learning, with “an aim to assess, extract and evaluate information from textual data.” To that effect, TNW points to a paper recently published in Nature which reports that an AI has now “managed to predict future scientific discoveries by simply extracting meaningful data from research publications.”

Researching and understanding a specific scientific question requires the obvious step of consulting books, specialized publications, web pages, and any other relevant sources. Of course, this can be an extremely time-consuming exercise, particularly if we have a very complex problem or question at hand. That is where NLP comes in. By using “sophisticated methods and techniques, computer programs can identify concepts, mutual relationships, general topics and specific properties from large textual datasets.”

As is discussed in the aforementioned study, “so far, most of the existing automated NLP-based methods are supervised, requiring input from humans. Despite being an improvement compared to a purely manual approach, this is still a labor-intensive job.” But researchers who prepared this paper were able to create an AI system that “could accurately identify and extract information independently. It used sophisticated techniques based on statistical and geometrical properties of data to identify chemical names, concepts, and structures. This was based on about 1.5 million abstracts of scientific papers on material science.”

Then, this machine learning program “classified words in the data based on specific features such as “elements”, “energetics” and “binders”. For example, “heat” was classified as part of “energetics”, and “gas” as “elements”. This helped connect certain compounds with types of magnetism and similarity with other materials among other things, providing insight on how the words were connected with no human intervention required.”

This method made it possible for the AI to “capture complex relationships and identify different layers of information, which would be virtually impossible to carry out by humans.” This made it possible to give insights well ahead in comparison to what the scientists dealing with the field are able to do at this moment. AI actually recommended materials “for functional applications several years before their actual discovery. There were five such predictions, all based on papers published before the year 2009. For example, the AI managed to identify a substance known as CsAgGa2Se4as as a thermoelectric material, which scientists only discovered in 2012. So if the AI had been around in 2009, it could have speeded up the discovery.”

 

Former diplomat and translator for the UN, currently freelance journalist/writer/researcher, focusing on modern technology, artificial intelligence, and modern culture.