Hacker Newsnew | past | comments | ask | show | jobs | submit | AstralStorm's commentslogin

It will be challenged in the Tribunal if it passes, and will lose as it is incompatible with baseline EU laws.

Specifically, the charter: https://eur-lex.europa.eu/eli/treaty/char_2012/oj/eng

Article 8 point 2. They're missing the provisions of data retrieval. If available, that would immediately take down the whole monitoring system as a full message history request would have to be available since they don't know which data is private or personally identifying.

Additionally, this falls under GDPR legitimate concern category so must be possible to reject by the user. Otherwise it's an involuntary measure making the law contradictory.

With this in mind, it's toothless if corrected.


If you haven't started doing the thing while researching it, you're doing archeology or historiography, not research. (Different scientific fields. Research requires experiment.)


It's called documentation. Somehow people forget to include a non-external service dependent version with their code these days.

Probably because "nobody is paid for doing that".


Yeah sure. The tricky part is if you even know which action to take. To actually do that you need to research enough, conceptualize the end goal and envison a rough but workable plan.

Most people don't and blindly stumble going after short term to mid term reward. Then later life crashes them hard, except for the lucky few.

And the truth is, now some people have many more luck coupons to spend than others.


Who's paying to school them at googling? Certainly not you...


Unless you wasted years by going in the wrong direction which could have been avoided by a week of serious thinking and planning... (Or even a year. Some mistakes are extremely costly.)


It's even more legally funny than this. They are also allowed to enact entrapment to some degree and it passed court muster in the US. That because it's extremely hard to use it as a defence.

Note that these actions are illegal in most continental jurisdictions as stings must be devised ahead of time against specific groups of people. There's also Article 6 of ECHR.

In other words, FBI cannot run a sting off an EU site like this, at least definitely not a German one.


How can you even sue without any legal identity? This website and an organisation does not happen to have any. Might as well be some shell company in the Carribeans with no legal standing in France. It's not even good enough for public prosecution, as the tip would then go through French services.

This law is completely backwards, and worse than a SLAPP. If you cannot respond to a report in any way, it should be null.


If you can sue shark fins, why not a website? https://en.wikipedia.org/wiki/United_States_v._Approximately...


Amazing, here is a list of other similarly hilariously-titled “in rem jurisdiction” cases: https://en.wikipedia.org/wiki/In_rem_jurisdiction#Examples

Some good ones: - United States v. One Solid Gold Object in Form of a Rooster - United States v. 11 1/4 Dozen Packages of Articles Labeled in Part Mrs. Moffat's Shoo-Fly Powders for Drunkenness - South Dakota v. Fifteen Impounded Cats


My favorite one: United States v. Article Consisting of 50,000 Cardboard Boxes More or Less, Each Containing One Pair of Clacker Balls

https://en.wikipedia.org/wiki/United_States_v._Article_Consi...


"More or Less" maybe takes the cake for me.


I've spent a lot of time in forfeiture court and it's always a chuckle to hear these cases get called. Especially the defendants' lawyer "Yes, your honor, I represent the cats."

Always wanted the cat, or the Honda Civic or whatever to ask to represent themselves. I guess if there was a foreclosure against an Nvidia Spark with a local LLM it might be able to give it a worthy try.


My favorite is United States v. One Solid Gold Object in Form of a Rooster.

The Rooster won.


Fantastic. Each case is basically an SCP object.


French law is based on an entirely different legal system compared to US (and Anglosphere law):

https://en.wikipedia.org/wiki/List_of_national_legal_systems...

It might not be possible to do something like that in France (though I assume there are other mechanisms available in that case).


Louisiana has a bunch of French law: https://en.wikipedia.org/wiki/Law_of_Louisiana


And your point of trivia is not applicable here:

https://en.wikipedia.org/wiki/United_States_Court_of_Appeals... -> Louisiana isn't in the list for the court that handled that trial.


Holy crap, 30'000 sharks kills for a bloody soup. Insane and that wasn't even their only journey.


> How can you even sue without any legal identity?

The images of the various messages on the adguard page are not lawsuits.

They are threatening messages that threaten to create legal issues, but until and unless they carry through on the threats, are simply "threats" to the extent we've been given any visibility into the messages contents.


US courts let you sue objects under "in rem" jurisdiction.

In rem = the thing is the defendant. You're not suing a person, and you're asking the court to decide who owns or controls a specific property.

The quintessential case is United States v. $124,700

https://en.wikipedia.org/wiki/United_States_v._$124,700_in_U...


Interesting, so the US already has the tools to go after AI, self driving cars and robots.


I remember when publishers were suing individuals using nothing more than a list of IP addresses. Those crazy times seem to have come around again.



Lawmakers are gonna have to figure that out soon (years hopefully) since it’s not unlikely that AGI will have the same issue.


In the U.S., “John Doe” is typically used when cannot (yet) identify the person to name as a defendant. Once the case is filed then the plaintiff can execute the necessary subpoenas to identify the defendant specifically.



I wonder if there is a real person out there called John Doe who regularly received legal threats for a myriad of alleged crimes.


The store owner visibly responds to the customers differently. Fingerprinting is invisible. It's more like the store owner recording everyone on hidden camera.

So no, you cannot steelman a broken analogy.


It's not even that. Only a kernel of the LLM is trained using RLHF. The rest is self-trained from corpus with a few test questions added into the mix.

Because it still cannot reason about veracity of sources, much less empirically try things out, the algorithm has no idea what makes for correctness...

It does not even understand fiction. Tends to return sci-fi answers every now and then to technical questions.


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

Search: