Has anyone ever used a sanity check?

diorthotis

New Member
Joined
Apr 14, 2021
Messages
18
Reaction score
6
Credits
139
A sanity check in programming seems very strange. If a sufficiently advanced AI were to chat and help develop a program, we may have difficulty in detecting it is an AI, and it would appear to be a human. Who put the sane_behavior file in linux? That definitely is strange to me. Any thoughts?
 


I'm not sure if this will help anything, as it doesn't answer a darned thing you're asking about. I waited to see if someone else would respond to this, but nobody did. So, we'll see if this helps.

I'm a retired mathematician. My company modeled traffic. We used sanity checks.

Let's say you're trying to find the square root of 224. A sanity check tells you that the answer can't be 224. It can't be 1. It can't be 2340. It has to be within a certain range. Anything outside that range fails the sanity check. By simply knowing what the answer should be in approximation, let's say somewhere between 10 and 20, you can eliminate large swaths of answers as being the wrong answer simply by doing a sanity check.

In programming, if we know it's outside the bounds of sanity, we exit the program with a fault. We put sanity checks in because humans are flawed and it's a good way to reduce errors.
 
That sounds like something for keeping an AI system within reasonable bounds. It seems ridiculous for me to need sanity checks, since my program will either be correct or incorrect, and I can find out by compiling it.
 


Latest posts

Top