Young You Wrote A Bank Software Thanks To AI

Coding AI may be trained on the code you wrote when being young. Now, it’s perceived as safe and used by coders, while it’s not safe

In this corner of the internet we discuss how AI will lead to the destruction of the world. Well, at least that’s what you may think after my one, or two critical articles.

But I’m not writing about AI critically just for the sake of it. I just want to know what’s ahead.

And it’s quite interesting.

Tim Anderson wrote an interesting article lately, where he quotes Snyk quoting Stanford University Study revealing that 15% of analysed mobile apps had snippets copied from the Internet of which 98% had security vulnerabilities.

What it means is that we should be both greatful and not happy about all the people how gave publicly wrong, misleading, insecure answers on online questions.

Now, it means that every AI based on online resources will be as bad as these answers.

So we can relax after the latest news that some AI scored better than human programmers in some educational challenges because practice may be different for benefit of humans.

We should be also worried, because Tim indicates other research that states that we tend to think AI generated code is safer than human written.

The perception bias leads me to a conclusion that our banks may soon run on that answer you gave on an online forum when being at primary school.

When you wrote it, you wasn’t aware how far it will go. But no worries. Online answers are tailored to the question, so no one expects to account for every aspect including security.

Security should be added while implementing. However AI does not know that, because it’s hidden behind closed curtains.

Well, at least for now. Microsoft bought GitHub, and at this very moment some companies surely think how to legalize training AI on private repos. (Microsoft and GitHub for sure act according to the law, security rules etc, awesome companies)