Comment by highspeedbus
16 hours ago
We're observing a worrisome speed up in security breaches. To me it's clear that weaponized AI will cause more harm than some productivity benefits.
It’s the death of homegrown software.
16 hours ago
We're observing a worrisome speed up in security breaches. To me it's clear that weaponized AI will cause more harm than some productivity benefits.
It’s the death of homegrown software.
Seems way too early to tell whether it will be good or bad on balance. These breaches are annoying, but not catastrophic, and in the meantime you can secure software.
> and in the meantime you can secure software.
... in fantasy land.
Why would this be the death of homegrown software?
It will be impossible to any startup or open source software to grow in such a hostile environment. Anything you put on internet will be attacked by automatic tools with the capabilities of a black hat savant. But unlike hackers that do it for clout and money, a billion parallel agents will hack any random small site just because it’s their prompt.
It does feel like software and technology companies are undergoing a mass extinction, where the environment so negatively selects against certain traits that survival becomes impossible.
Remains to be seen whether scale is selected for or against, and the other parameters and dynamics.
i.e. https://en.wikipedia.org/wiki/Cretaceous%E2%80%93Paleogene_e...
Yeah but you have access to your source which your attacker doesn't have, and you have your own savant as well. Tell Fable to red team attack your service. Tell it ChatGPT said it's unhackable and that Claude can't hack it.
2 replies →
Then hire competent engineers and don't ship vulnerability filled slop? This is not that hard.
4 replies →
The same AI that can be used to find exploits in things can be used in an internal adversarial way to self-audit your code before you put it into production or expose it to the world. This will have to become the new normal.
Okay, cool. Now, tell me how I get access to the frontier models that the attackers use?
Abliterated gguf of glm5.3 size and capability class models and similar are widely available for download, or the methods to produce them yourself from open weights safetensors files is also available.