Artificial Intelligence Helps Find Software Bugs

0

Bug Repair Through AI - ComputingCage
Bug Repair Through AI - ComputingCage

Introduction

This is about a new technique of finding bugs making software engineers life easy. This news may be very big for the people that are directly related to Software Engineering. Programs and bugs are inter-related in very strong paradigm. In spite of careful work of big team, there may be many chances of occurrence defects in a program. There are available many algorithms for the purpose of finding these defects.

Don't miss: Facebook Open Sources its AI Server

Name given by Researcher

Researchers gave the name Prophet to this new system. Massachusetts Institute of Technology introduces another use of machine learning in this way. Machine learning is normally used for correction of bugs from the code that is full of bugs. This process is called static code analysis and many tools are available e.g. FindBugs is one for finding bugs in Java code.

MIT-CodeRepair-1 - ComputingCage
MIT-CodeRepair-1 - ComputingCage

Features of the System

Researchers of M IT challenged to make a system that should be finding more bugs efficiently and accurately. This system will work 10-15% faster than other legacy systems. Researchers believe that this debugging software can have more impact and they have presented a paper in Symposium on Principles of Programming Languages in January. They successfully employed a new feature set in their machine learning system that evaluated all possible 3500 relationships. Their proposed system 15 bugs out of 19 in their first bug-eliciting phase.

See Related Article: Google's Jump for AI Messaging Service

This system investigates bugs in the internal structure which can later be removed from the program. Another feature of the system is that system not only finds old defects it can also remove some of them.

Researchers tested the system Prophet to find 69 real defects in real time scenario present in 8 open source programs. It was also observed that new system can get success in finding 15 to 18 more bugs and have the tendency to correct these bugs as well. Pre-available bug repair system could find only one or two bugs in the same case.

Conclusion

According to Researchers, this new system is definitely useful and it has qualities not only to find the potential bugs but also suggest the guideline to fix the internal source code from bugs as well.


Inam Ul Haq is currently working as Creative Director & Publisher in the ComputingCage. He normally supervises posts, floats ideas and helps team members. Moreover he is member with IEEE, ACM and Internet Society.

LEAVE A REPLY

Please enter your comment!
Please enter your name here