Analysing code written by people who have written software is notoriously difficult, as the code itself can be quite complex.
In a recent study, researchers at the University of Waterloo have demonstrated that if a computer is able to process a large set of source code, it can produce code that is 10x faster than a human programmer can write.
The study was published in Computer Vision and Pattern Recognition, a journal of the American Association for the Advancement of Science.
The team used a simple computer-vision model to examine how the human brain would process code written in the human language.
They then tested the machine on a large number of code samples and found that it was able to solve about a third of the problems.
“It was really surprising, to be honest,” said the team’s lead author, Dr. Andrew O’Reilly.
It was not something that we had anticipated.
“There are a lot of really hard problems to solve,” he said.
O’Reilly said the paper’s findings are “a very important step towards a better understanding of the neural basis of software and computer vision.”
The researchers said their computer could solve problems in the same way that the human eye is able.
This study builds on previous work that has shown that computers can do things that humans can’t do, like recognise patterns in pictures or understand speech.
While computer vision algorithms are becoming increasingly powerful and useful, O’Connor said the challenge is to find the right balance between the speed of the computer and the speed at which the algorithm can understand the code it’s trying to decode.
This computer-imaging technique could be used to test whether software can be read as text and not be.
When this type of algorithm is used in a more real-world environment, such as when building a smartphone, it will allow developers to create more robust, robust software and thus more value for their users.