4 ms·
The article totally understands this distinction of fictionality. That's why it defines personhood thusly: The what is easy. It’s personhood. By perso
by gizmo 4y ago
The article totally understands this distinction of fictionality. That's why it defines personhood thusly:
The what is easy. It’s personhood.
By personhood I mean what it takes in an entity
to get another person treat it unironically as a
human, and feel treated as a human in turn. In
shorthand, personhood is the capacity to see and
be seen.
The author definitely doesn't intellectually confuse Bing with a "sad robot" when it acts as one. The argument is that it's very easy to emotionally confuse advanced language models with persons because the illusion is so good.
- throwaway4aday 4y agoHonestly, that's a terrible working definition of personhood. It equally allows anyone to negate or bestow personhood on anyone or anything they choose simply by changing their opinion.
- velcrovan 4y agoThat…is exactly what happens in real life
- mecsred 4y agoUnfortunately when your working with concepts that can't be measured/only exist in the eye of the beholder, any definition you make will have that problem. The only litmus test for "personhood" is if you think they're a person.
- qup 4y agoI can't wait for PETA-for-things-we-say-are-persons
- pegasus 4y agoBut it's not easy at all to get confused, unless one decides to consciously suspend disbelief, in spite of what they know. If they do know how LLMs work. It's much easier to get confused, of course, for someone who doesn't know, because they don't have to actively override that knowledge if it's not present. But someone who does, won't for example have any trouble shutting down the conversation midway if the need arises, because of some misplaced emotional concerns of hurting the bot's feelings. At least that's my experience.
- BlueTemplar 4y agoOn the Internet, nobody knows if you are a ~dog~ chatbot. So basically, impersonation and emotional spam might become a problem. (Depending how easily ethically compromised people will be able to profit from it.)
- pixl97 4y agoEh, it appears this thread is ignoring the Chinese room problem, which is what you have defined with your post. I personally reject most of Searles arguments regarding it. If a black box is giving you 'mindlike' responses it doesn't matter if it's a human mind or a simulated one. In any virtual interaction, for example over the internet the outcome of either type interacting with you can/could be exactly the same. Does it matter if you were manipulated by a bot or a human if the outcome is the same?
- BlueTemplar 4y ago[dead]
- lukev 4y agoIf the argument is that it's very easy to emotionally confuse language models and persons, than I reject that argument on the following grounds: No works of fiction are persons. All "I" statements from the current generation of LLMs are works of fiction. Therefore, no "I" statements from the current generation of LLMs are persons. Premise 1 is in conflict with the author's premise that personhood can be ascribed at will; I'm happy agreeing to disagree on that. I do not think it ever makes sense to ascribe personhood to fictional characters (for any meaningful definition of personhood.)