Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

What percentage of this is lawslop?

https://github.com/righttoprivacyact/bill/tree/main/tests

There’s clearly a non-trivial level of LLM involvement.

I want to say 100% lawslop. I can’t figure out who’s behind this to ascertain their qualifications and acumen in the space.

100% seems like a safe place to start speculating from but I can be talked down.

 help



I think this kind of argument is a modified version of an ad-hominem attack.

When you disagree with an argument, you are supposed to address the argument itself, not the thing making the argument.


That went out of window and provably failed even before LLM. The strategy of flooding everything with cheap false claims and arguments while demanding that the opponent spends increasing amount of effort and time was a success even before. It became worst with LLM.

So, no, you are making the claim, first prove it is worth any of that effort.


This seems like a chicken and egg problem.

Ad-hominem (literally: "to the person") requires a person on the other side of the argument. This wasn't made or written by a person, thus ad-hominem does not apply.

Thats a bit pedantic. Youre still arguing against an entity rather than addressing the argument.

If you prompted an LLM to make a PSA that people should brush their teeth, is it a fair to argue that brushing your teeth is bad because an LLM made the argument?


Let's call it "Ad machinam".

A literal form of "to the machine”: none of the rights that a person has.

Example usage:

“That’s not a rebuttal; it’s an argumentum ad machinam -- you’re rejecting it just because AI wrote it.”


And really when you think about it, all AI is is just the a statical recombination of (almost) everything that (almost) everyone has written.

So it's a kind of mechanical recombination of ideas.


Almost like reject the argument because the logical reasoning was supplied emotionally.

Presumably why the person you are responding to called it a modified ad hominem.

Maybe we need a phrase like ad-machinam to attack LLMs.

Made me coin a term.

Ad hominllm: the dismissal of a work or an argument because an LLM was, or may have been, used in its construction or editing.


Pronounced: ad homin-ellem?

It all makes sense, I don't care if it's LLM-generated or not. I'm fine with 100% LLM-written bill if it protects me from Flock et al.

I agree, and it wont... We will all be tracked at all times as the new global economy is created.

Biblical times.


Yeah, most charitably, it seems to be some sort of an LLM art project. And it's another day when we engage with slop because it happens to say something we like.

Who has ever written like this:

    (a) In general.—There are authorized

As opposed to bills sponsored by real people that completely decimated privacy? Patriot Act? I’ll take slop that has some semblance of reasonable privacy protection any day.

There's some obvious issues that come with LLM generated bills. LLMs tend to stay very generic, to have a lower chance of being "wrong".

I could also ask my 5 year old to draw a bill for me, that's as successful as giving a generic bill.


> LLMs tend to stay very generic, to have a lower chance of being "wrong".

It depends entirely on how you prompt them.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: