It is a truth universally acknowledged that an artificial intelligence-based system can perform calculations and other data-based analyses faster than a human. But when an AI creates something in hours, which took humans hundreds of years to discover, is a new feat altogether.
Now, an AI programme called Atom2Vec, developed by Stanford physicists has successfully learned to distinguish between different atoms after analysing a list of chemical compound names from an online database. The programme uses concepts from natural language processing to cluster the elements together according to their chemical properties.
Shou-Cheng Zhang, a physics professor at Stanford’s School of Humanities and Sciences said, “We wanted to know whether an AI can be smart enough to discover the periodic table on its own, and our team showed that it can.”
Get started with deep learning using MATLAB
Professor Zhang further explained that they modelled Atom2Vec on a Google programme called Word2Vec. Here, the AI works by converting words into numerical codes or vectors. By analysing these vectors, the AI programme can figure out the probability of a word appearing in a text given, in relation to the co-occurrence of other words.
Zhang explained with a simple example:
The word “king” is often accompanied by “queen,” and “man” by “woman.” Therefore, the mathematical vector of “king” might be translated as:
king = queen – woman + man
“We can apply the same idea to atoms — instead of feeding in all of the words and sentences from a collection of texts, we fed Atom2Vec all the known chemical compounds. From this data, the AI program figured out that potassium (K) and sodium (Na) must have similar properties because both elements can bind with chlorine (Cl). Just as the king and queen are similar, potassium and sodium are similar too,” said Zhang.
Try deep learning using MATLAB
Provide your comments below