IBM seeks to create code that programs autonomously

IBM published a database with 14 million samples as part of its Project CodeNet initiative, with which the company seeks to help train Artificial Intelligence (AI) mechanisms capable of programming code or translating it autonomously.

The use of a current technology such as AI in code programming seeks to modernize the sector and make it possible for machines to help in tasks currently performed by humans, such as cleaning, maintaining and updating code, as reported by the company in the framework of its event annual IBM Think 2021.

“The language of the machines is the code and, in the same way that we have advanced a lot in the processing of natural language as in the translators, we can teach the machines to program“, said the senior vice president and director of IBM Research, Darío Gil.

Now, the company publicly released Project CodeNet databases on the open GitHub platform, with 14 million data samples, 500 million lines of code in more than 55 different programming languages, from some modern ones like C++, Java, Python and Go to old ones like COBOL, Pascal and FORTRAN.

The use of this database aims to train new AI tools that are capable of programming code or translating it, overcoming current obstacles to this, such as contextual information, which makes the larger a program, the more difficult. it turns out to be translated for a machine.

K. Tovar

Source: VentureBeat