Blog

DeepCoder : A Prolific AI Algorithm

Ever wondered that an AI would be able to do programing by itself? It was merely a dream until Microsoft and the researchers of University of Cambridge came up with the idea of the “DeepCoder”.

These researchers include Matej Balog (From University of Cambridge) and Alexander L. Gaunt, Marc Brockschmidt, Sebastian Nowozin, Daniel Tarlow (From Microsoft Research) who have built this algorithm and presented a paper on its use.

From the paper:

A dream of artificial intelligence is to build systems that can write computer programs.

DeepCoder is an algorithm, built on IPS (Inductive Program Synthesis). It is given a sample for input and output. It, then searches the codes from already completed programs and after finding a suitable match of codes it applies to its program to produce the desired output. In the building of an IPS it required to solve 2 most important issues. One, how to search for a set of suitable programs. For this a program space and search procedure is to be defined. The second issue to solve is how to decide which program among the suitable set is more applicable. For this, researchers have used Domain Specific Languages, Search Techniques and Ranking Methods.

DeepCoder writes the codes within seconds saving the time and efforts of users. Currently DeepCoder can write up to 5 lines of simple codes but there would a time when it will be able to write plethora of complex codes by its own.

This eases the work and efforts of developers but will it lead to job risks of developers in future is a big question in itself!

You can read the published research paper by Microsoft over here

Copyright © 2019 by Binary Bakers IT Solutions