In my experience, AI is really good at finding problems because of the large context window and speed. But it is really crappy at writing good code, because it can’t distinguished good code from bad. The way AI writes code is just to generate stuff and fix problems until all problems are gone. That is a terrible way to develop.
This is why you need strict quality control whenever you have AI making things.
Cybersecurity experts need to thoroughly examine all the code and how the software works…or doesn’t work and help fix it
Then you need ethical hackers to test the security and you need the cybersecurity experts to fix the problems the ethical hackers find.
it’s boring, grueling work but if you don’t do it everything falls apart
The article is about AI finding exploits.
Exploits in… What?
Software.
Within which is so so much slop code nowadays
sure, but you can’t just patch shit Willy nilly without also breaking things.
doing things correctly takes time and money, neither of which we have time for when we have the poor, defenseless, baby bird shareholder to take of!!
I mean i agree that proper quality control and review is needed.
Its just that the conclusion Zeus takes from this article is not really corroborated by the article.
It’s funny how the narrative until this point has been that AI is too crappy at coding to be useful.
In my experience, AI is really good at finding problems because of the large context window and speed. But it is really crappy at writing good code, because it can’t distinguished good code from bad. The way AI writes code is just to generate stuff and fix problems until all problems are gone. That is a terrible way to develop.
True, but let’s be honest, that’s also how a LOT of human programmers do it.
Yes, but all the guardrails we have to keep poor developers in line start to fail when you scale it up with AI
Is a mantra that exists across disciplines