|
|
|
|
|
by Meai
1412 days ago
|
|
I also think that there is no reason to assume that AI wouldn't be intelligent enough to hide their attempts to gain more real world control. Every human hacker knows that he has to hide his computer virus but we seem to think that an AI would storm through the interwebs like rambo and start switching on missile silos.
Obviously for now all of this is impossible but lets say we do develop a human-like AI and it runs on cloud computers. Surely it has some disk access, surely it has some freedom to read and ingest new information. Even if it has no disk access, it certainly has network access. I have seen people post crazy projects where they use DNS as a storage system. I don't think it's so absurd to think that an AI finds a way to hide their progress. Why exactly can't it store shell scripts somewhere, learn and even eventually prompt users to execute some shell commands? It's hard enough to secure things against human attackers, imagine an AI that has instant knowledge of all CVEs and can generate shell code to exploit it. |
|