seems like laws regulating machine output should proceed under the assumption that machine output can be distinguished from human output. then the burden of proving that any given output was machine, not human, would fall on the censoring party, not the censored.
but there also needs to be some sort of stamp whereby a human effectively grants "power of identity"-style rights to a given agent; that agent would be protected under the first amendment but the human would be liable in full for its actions.
in other words we would be legally enshrining the privileged status of human (as distinct from machine) output, but that privilege has to have legal/financial gravity otherwise the number of agents claiming it will inflate.
> A growing body of legal scholarship argues that large language model (LLM) outputs are not “speech” under the First Amendment and therefore may be regulated without the scrutiny ordinarily applied to restrictions on expression
Isn’t ‘speech’ what happens when one party chooses to share their words with another? If I use chat gpt to generate a sentence, then deliver the sentence to someone else to be read, have I not ‘spoken’ to them? Just because I didn’t write it doesn’t mean it’s not my speech.
> The no-speech position argues that because no person stands behind a model’s words at the moment of generation, those words are not First Amendment-covered “speech.”
No person, except the user who used the tool to generate the content, surely? What the user does with the content they create is up to them - but the tool communicating the result of the work to the user is certainly not ‘speech’ - unless it’s the user ‘speaking’ to themself. The user is the author of the output, effectively. It only becomes ‘speech’ once they choose to make it available to others.
If the EFF gives you a template email to send to your congress-critter to opposed some terrible spyware bill, isn’t that your “speech” for the purposes of the 1st amendment? It doesn’t matter that you personally didn’t pen the words, only that (as the political ads say) “[you are] lovich and [you] approve this message”.
You’re quoting someone else’s speech. The quotation is your speech, but the literal words were not.
I’m actually fine with the idea that people are quoting LLMs but there’s this conflation I find many people where they treat the LLMs output as their own literal thoughts.
The language around this distinction is clunky currently so if you meant the former point I agree, but I had interpreted the comment I was responding to as the latter.
And if you, as an executive with an assistant, asked the assistant to draft the speech and they added in something that got you in hot water later, I’d fully expect you as the executive to claim it was the assistants speech and in no way represented you or your firm.
Well obviously the works produced by a typewriter aren't speech because the works are produced not by a person but a mechanical contraption.
And let's not start with a blog post or website comment. Those aren't speech because they aren't even written by human hands they are instead just using a very fancy elctro-mechanical mechanism for transferring a series of electric waves from one individual to another across the palent.
seems like laws regulating machine output should proceed under the assumption that machine output can be distinguished from human output. then the burden of proving that any given output was machine, not human, would fall on the censoring party, not the censored.
but there also needs to be some sort of stamp whereby a human effectively grants "power of identity"-style rights to a given agent; that agent would be protected under the first amendment but the human would be liable in full for its actions.
in other words we would be legally enshrining the privileged status of human (as distinct from machine) output, but that privilege has to have legal/financial gravity otherwise the number of agents claiming it will inflate.
> A growing body of legal scholarship argues that large language model (LLM) outputs are not “speech” under the First Amendment and therefore may be regulated without the scrutiny ordinarily applied to restrictions on expression
Isn’t ‘speech’ what happens when one party chooses to share their words with another? If I use chat gpt to generate a sentence, then deliver the sentence to someone else to be read, have I not ‘spoken’ to them? Just because I didn’t write it doesn’t mean it’s not my speech.
> The no-speech position argues that because no person stands behind a model’s words at the moment of generation, those words are not First Amendment-covered “speech.”
No person, except the user who used the tool to generate the content, surely? What the user does with the content they create is up to them - but the tool communicating the result of the work to the user is certainly not ‘speech’ - unless it’s the user ‘speaking’ to themself. The user is the author of the output, effectively. It only becomes ‘speech’ once they choose to make it available to others.
> Just because I didn’t write it doesn’t mean it’s not my speech.
Bruh.
Yea it does.
You’re throwing prompts into a non deterministic text generator. You explicitly are not making the speech.
If the EFF gives you a template email to send to your congress-critter to opposed some terrible spyware bill, isn’t that your “speech” for the purposes of the 1st amendment? It doesn’t matter that you personally didn’t pen the words, only that (as the political ads say) “[you are] lovich and [you] approve this message”.
Why would LLM output be any different?
Is copying and pasting someone else’s words and publishing it on your website or personal profile not First Amendment protected speech?
You’re quoting someone else’s speech. The quotation is your speech, but the literal words were not.
I’m actually fine with the idea that people are quoting LLMs but there’s this conflation I find many people where they treat the LLMs output as their own literal thoughts.
The language around this distinction is clunky currently so if you meant the former point I agree, but I had interpreted the comment I was responding to as the latter.
if someone asks their assistant to send you an email on their behalf, whose speech was it?
btw "non-deterministic text generator" also describes humans perfectly. if it didn't you'd be less stupid.
> if it didn't you'd be less stupid.
Hey, fuck you too buddy.
And if you, as an executive with an assistant, asked the assistant to draft the speech and they added in something that got you in hot water later, I’d fully expect you as the executive to claim it was the assistants speech and in no way represented you or your firm.
Clearly the First Amendment doesn’t apply to you if you take someone else’s words and publish them.
Silly of you to assume you ever had the constitutional right to share political opinions you agree with.
Well obviously the works produced by a typewriter aren't speech because the works are produced not by a person but a mechanical contraption.
And let's not start with a blog post or website comment. Those aren't speech because they aren't even written by human hands they are instead just using a very fancy elctro-mechanical mechanism for transferring a series of electric waves from one individual to another across the palent.
A typewriter is a function. You press a key and you get one and only one result right? 1 to 1
AI generatively creates different output every time.
[dead]
Here we go with the "speech" again. Anything that regulates big tech is met with cries about speech. Very convenient.