3 ms·
This is the real threat of these large AI models. They will revolutionize the ability to find vulnerabilities. Especially if we achieve AGI.
by kvetching 4y ago
This is the real threat of these large AI models. They will revolutionize the ability to find vulnerabilities.
Especially if we achieve AGI.
- kibwen 4y agoIf we have AGI capable of finding vulnerabilities, then we can just ask the AGI to design a system so secure that even the AGI cannot find any vulnerabilities.
- walterbell 4y agoSome vulnerabilities have their own supply chain and incentives.
- varunjain99 4y agoIt will be an interesting race... I do think that it'll be easier to find vulnerabilities than to design systems that are provably secure. It's like evaluating a function f(x) vs finding the values of x for which f(x) = 0.
- lima 4y agoThey will revolutionize the ability to find and fix vulnerabilities. Threat actors can already afford expensive security researchers today. It removes some of the asymmetry.
- KeplerBoy 4y agoi worry more about GPT-assisted social engineering.
- tptacek 4y agoWriting programs to automatically find bugs is a pursuit almost as old as compilers, and the models used to do that in modern static/symbolic analysis tools are much more sophisticated than an LLM's next-token generator.