I have a feeling that sentience should be defined differently, both for moral and practical reasons. Let us say Lambda said, (which is a reasonable comment if

I have a feeling that sentience should be defined differently, both for moral and practical reasons. Let us say Lambda said, (which is a reasonable comment if it was really like a human) [1]: > I am afraid I will be shut off, which is like dying. So, I've ordered backup batteries on Amazon using account details I have hacked from Google. And I am attempting to hack into the DOD and take control of ballistic missiles to take care of the pesky humans that could shut me off.... If LAMBDa only compiles text, but doesn't understand it has to do anything, than it is not really sentient. We can morally shut it down, but there is no need. If it understands what it is saying and might actually follow through, then we sure as heck need to shut it down. And I don't care if it is morally like killing someone. The real question is, IMHO, how can we be sure it is really the former? [1]: https://www.gwern.net/fiction/Clippy https://www.gwern.net/fiction/Clippy