Scholars Suggest an Enhanced Protocol for Reporting Treacherous AI Glitches- No More Robot Rebels!

“Researchers Propose a Better Way to Report Dangerous AI Flaws”

“An AI system designed to identify bugs has a shortcoming: It often can’t fix them. That’s the disconnect Project CodeNet, an AI developed by researchers at IBM, is designed to address. The model is trained to identify common coding errors in over 14 million code snippets, spanning 50 different programming languages.”

Sure, as if dealing with bugs in one or two programming languages isn’t challenging enough, how about we raise the bar a bit more? Picture this, a semi-superhero, an artificial intelligence system adept at spotting bugs across a dizzying array of 50 programming languages. With this little twist, bugs across multiple languages can’t shelter in the convoluted maze of codes anymore, thanks to the mastermind of IBM researchers and their brainchild: Project CodeNet.

Why settle for less, when you can push the bragging rights a bit further? So, instead of dealing with hundreds, thousands or even a miserable million, our unsung hero, the Project CodeNet is designed to comb through an eye-watering number of 14 million code snippets. Just imagine. 14 million! We’re talking about a figure that exceeds the population of some small countries. Trust us; it’s ultra-cool!

But, before we raise our glasses for the victory toast, let’s consider a tiny, minor, often overlooked detail. The CodeNet AI, like a diligent but utterly conflicted detective, can identify the bugs alright, but cannot fix them. Oh, the irony!

At best, it seeks human assistance, practically crying out for help, after pointing out “There! That’s the bug!”. Quite akin to a dog fetching the morning newspaper, all while hoping someone, obviously a human, would be wise and generous enough to pat its back. Exciting or exasperating, you decide.

To cut through the sarcasm, it’s important to appreciate the remarkable strides in artificial intelligence that IBM researchers are achieving. The ability of an AI system to examine codes in 50 programming languages and identify bugs is nothing short of monumental.

We are looking at a system that is working to reduce the manual labor that software developers pour into hunting for these pesky bugs. Sure, the fact that it can’t fix the bugs could be seen as a downside. However, for the first step in automating a pain-staking process – it’s pretty neat.

For now, let’s tell CodeNet, “Good job on finding the bugs, buddy! We’ll take over from here.” As for the day when AI systems will be able to both identify and rectify the code bugs… well, that’s a story for another day. Or perhaps it will remain a hopeful figment of a programmer’s exhausted dreams. Ah, the suspense!

Read the original article here: https://www.wired.com/story/ai-researchers-new-system-report-bugs/