1
Replies:
5
@wirepair I read this earlier and a lot of the "research" cited is promotional material, including stuff where the claims were found to not check out at all.
1
@gsuberland yeah i kinda glossed over that part, but i've beeng using it in my day job for determining vulns and it's really fucking good at finding root cause of 1days. So I still think he's correct even though his references aren't the best.
1
@gsuberland i've also used it to reverse engineer games (unity il2cpp) and it was also really fucking good at finding various points of interest that i had it drill into (netcode handling etc).
1
@wirepair there's stuff it's good at, for sure, but I think the assertion that "vulnerability research is cooked" is complete nonsense regardless. both from the actual discovery/triage side and from more fundamental bug class discovery angles.
2
@gsuberland I really want you to be right. But I’m comparing where these things were a few months ago to now and it is so wildly better than it was. If they continue to get better, which these new models allude to being, i do think there will be less need for human VR devs. Also let me be perfectly clear if I could go back in time and erase LLMs from existing I would in a heartbeat. I think they are, and will continue to, make everything worse
2
