For the world of science, that’s amazing news. Science writers, on the other hand, might have a different take on the matter.
To address this new scientific breakthrough, the Qatar Computing Research Institute joined forces with MIT to develop a neural network capable of reading and interpreting scientific papers and rewriting them in humanly-readable terms, which – up to this point – used to be the job of science writers.
You see, science papers are often not the most engaging kind of content to read and often end up being incomprehensible to your average reader. Due to how cheap it is for the neural network to scan and interpret a large number of scientific papers, this digital transformation breakthrough may soon force science writers to start considering switching over to a new career.
Interestingly, the idea sprung up when the researchers were working on another project; namely, the challenge of developing an AI capable of solving difficult physics problems. Somewhere along the way, they found that their RUM approach could address the problem of natural language processing, which is something that computers are still struggling with.
To help illustrate the principle, imagine a multi-dimensional space with words in it, where each word is represented by an individual vector. The system can remember every word and grasp the concept of a sentence by finding patterns while assigning meaning to them.
Preslav Nakov from the Qatar Computing Research Institute claims that the RUM AI can do two things very well when combined with neural networks:
– It boosts memory
– It allows them to recall information more accurately