Disclaimer: We may earn a commission if you make any purchase by clicking our links. Please see our detailed guide here.

Follow us on:

Google News

Microsoft Develops an AI Meant to De-Bug Code Errors Called BugLabs

Bipasha Mandal
Bipasha Mandal
Bipasha Mondal is writer at TechGenyz

Join the Opinion Leaders Network

Join the Techgenyz Opinion Leaders Network today and become part of a vibrant community of change-makers. Together, we can create a brighter future by shaping opinions, driving conversations, and transforming ideas into reality.

Microsoft Research Principal Researcher Miltos Alamanis and Senior Principal Research Director Marc Brockschmidt just introduced their newly developed BugLabs artificial intelligence. This newly introduced AI will find developers find errors in their code and debug the applications likewise.

Microsoft introduced the new two networks they built in an article titled, “Finding and Fixing Errors with Deep Learning”. One of the networks will introduce small errors into the code, and the other one will find these bugs.

In recent times, the use of AI has become more prevalent, and in this case, the use of AI means that there is no need for developers to self-monitor the data throughout the process. Miltos Allamanis and Marc Brockschmidt mentioned in the report, “In theory, we can apply it extensively to the game of “hide and seek”- teaching machines to recognize complex errors in tasks. Unfortunately, these bugs are usually beyond the scope of modern artificial intelligence methods. In view of this, the research team decided to focus more on a set of common mistakes-including incorrect comparisons (such as using <= instead of symbols, inappropriate Boolean operators (and/or), abuse of variables ( Misuse i instead of j), etc.”

The researchers at Microsoft focused on Python code for system testing. Once the detector has passed all the necessary training, it can both detect and fix bugs in the actual code. However, Microsoft also mentioned that the researchers also had to manually annotate certain types of small error data sets in the Python Package Index. The “hide and seek” training model of the AI is 30% more successful than other alternatives of identifying and debugging codes. That being said, Microsoft was quick to mention that there are still many false positives in the test, and as such, some more improvements need to be carried out before this AI model can be put to practical use.


Partner With Us

Digital advertising offers a way for your business to reach out and make much-needed connections with your audience in a meaningful way. Advertising on Techgenyz will help you build brand awareness, increase website traffic, generate qualified leads, and grow your business.

Power Your Business

Solutions you need to super charge your business and drive growth

More from this topic