Or an alternative: The whole background why in Dune thinking machines are outlawed and how it came to pass, that humanity had to fight the Butlerian Jihad.
I think GPL was though of and is an answer to the idea of copyrighting code. If there was no copyright for code, there might not be any copyleft license. If we were all allowed to freely copy and modify and redistribute etc., then there would be no need for licenses enshrining these rights.
Not to forget the European Human Rights Convention as interpreted by the European Court of Human Rights (ECHR), although it's not an institution of the EU but the Council of Europe, spanning all EU states and then some.
The Court's case law is recognized as is by EU, and includes some landmark privacy rulings, e.g., against access to contents and metadata of electronic comms without a court order (or comparable legal safeguards), or any forms of bulk surveillance.
Yes, the court can overturn laws that are deemed to infringe on privacy rights but this path takes years. The last time a privacy invasive law was enacted it took something like 8 years to be overturned.
Are you willing to live in the EU knowing that for the next 8 years the EU agencies will have access to every single message, email and picture you send your loved ones just in case one of those could be CSAM?
Even if that was overturned, how will one know if he data they intercepted and illegally obtained will be destroyed or if the text messages you sent to your wife/husband are not going to end up sold to the highest bidder on the darkweb?
To add some practical results to that: the Data Retention Directive[1] was struck down by a bunch of national constitutional courts before being completely annulled by the EU Court of Justice.
There are forces pushing for chat control again and again and again, until one day people are not vigilant enough to prevent this BS from happening. Then it will take ages to repeal and repair the damage done. Once there is some chance to vacuum up some personal data, businesses, government, and security forces will be very quick to grab that chance, and very slow to give that data up again, should it turn out, that the whole damn thing was illegal all along, due to some constitution. The malpractices might continue for years even after things have been declared illegal.
Yea, they only need one successful attempt. I'm still quite not sure about the motivation, the fact that politicians have an exempt from this law is scandalous at least.
Yeah. Fortunately Germany has a quite firm stance against surveilance, for well known reasons ;), and with Germany against it will be extremely difficult to pass this.
Not so sure about that. We have enough politicians here, who don't know shit about technology, or want to seem like they don't and will gladly hold the hand open to be bought by forces, who would like nothing more than get their hands on all that data.
In practice it _might_ still save you, if the country is willing to pay annual fines to the EU for not implementing some law or policy. But communication tools are always at least a 2 way thing, where you communicate with someone else. If that someone else has their chats leaked due to living in another EU country, where chat control is implemented, then even your constitution will not protect you.
Scenarios like these make me consider, whether one day I need to vote for my country to also leave the EU, as it might become a surveillance nightmare. The moment chat control impacts me and my chats on my devices, I think I will have had enough of this. But the economic consequences are scary of course. So damned if you leave and damned if you don't. Sucks.
The constitution of Italy says that we can't use war to solve international controversies. And yet we were in iraq… constitutions are somewhat meaningless if you have authoritarian governments that can ignore them.
And also a "proprietary data set" hahaha... Probably just means they don't want to show it, and it is data, that either they shouldn't have, or that there is nothing special about their training data and it is just meant to sound like there is some secret ingredient, while there is none.
Not sharing the data is pretty standard because 1) it tends to get the lawyers involved and 2) good data is critical for getting good results.
Imo you can get better results with great data and generic modeling techniques than with incredible modeling techniques and crappy data. Because if you have crappy data, you won’t even know if your model is good because your evals will also be bad.
This is why Anthropic is throwing a fit about the Chinese distillation “attacks”. Clean reasoning traces are gold.
Companies pay lots of money for proprietary agentic trajectories which are used during RL. These are things like "Task: summarize stock levels for months end accounting" which then traces the task though using SAP to look at different SKU stock levels, exporting them and generating summary Excel spreadsheets.
This is very different to the "scrape the internet" datasets that a table stakes for training a LLM.
Data has copyright issues, so one can't share it generally without getting permissions from all of the copyright holders. The data is not theirs to share, anyways. The derived (learned) weights are a different matter.
True for the pre-training data scraped from diverse sources. Less so for the later stage data for RL which is by all account more of a differentiator. In most cases the labs themselves produced the data so they are the copyright holders (or they are borrowing it from other labs via distillation). A lot of it is synthetic data, and since you can't copyright AI output, it becomes less about copyright and more about trade secrets.
If I train a model on a song's lyrics, and the user asks the model to recite the song, and it does, so in RLHF I downvote the generation because it refused-to-refuse to regurgitate the song's lyrics... is that RLHF session copyright-encumbered?
this is very normal for frontier lab companies. you need good data either synthetic or labelled (all the chinese open source models have their own armies of data labelers)
It would be funny, if there was a website anywhere, which accumulates data about all the tech giant C levels and shares it with the public, just like they share data about all of use behind closed doors to manipulate us and sell us shit, or sell our data to the next tier of data hungry businesses.
Probably wouldn't last long though, as they would be furious, that us lowly human beings are able to glean anything about them. The double standard of this is not obvious to them.
Probably the decisive difference is that the web needs to be delivered over network, while the OS is installed once and then is just there. If everyone needs to download your very special implementation of an otherwise pretty common type of widget, it becomes a nuisance and a waste of bandwidth.
In my opinion not being open about which data is ingested and trained on, and trying to make that a repeatable thing for a third party, is not worth being called "open". Glad they did that.
to be fair, this discussion has been had numerous times here and the industry has arrived on "open-weight" to describe the practice of releasing the post-training weights in an open manner but not releasing the data it was trained on.
That's what they call this and I think it's a pretty clear definition these days to people in the industry.
reply