The highest GDPR fine was 1.2 billion.
This isn't the GDPR but the DMA. That said, fines there are even steeper, 10% of global revenue for the first offence, 20% for repeated offences.
Muehe
joined 2 years ago
They had the most unhinged comments
Don't know the user, but apparently they just believe humanity might end any day now and nobody is even realising. Which to be fair is a somewhat of a fair assessment? One I have to keep at the very back of my mind in order to not constantly scream at everybody around me, so I feel this fishmaboi. The worst thing is that GHG emissions could have triggered the cascade inevitably leading to such a feedback loop decades ago and we wouldn't even know yet, because the cycles of the systems involved are so long. I mean sure, "Venus by Sunday" is a very weird way of framing it, but the danger seems to be rather real sadly. Consider that we don't need to be at Venus levels but just approaching them enough in order for society to break down. The hothouse earth scenario is rather long-term, but the feedback loop itself would probably be characterised by rapid global temperature shifts on the magnitude of 1°C per year (!) or more in either direction, as climate subsystems fail, rebalance, thereby get others to fail, therefore fail again, and rebalance again or fail permanently. This would take time to settle, which would probably be deadly for globalised society, might be deadly for humanity as a species, and already is deadly for a multitude of other species despite "the feedback loop" not even really having started yet.
Absolutely, what we have on the Discord in the way of documentation is a straight forward install guide with one screenshot and a download link (of the GitHub release) and a FAQ channel, which is basically just links to the GitHub for a good part of the answers. We also automatically mirror our changelog there. But that's it, and it's all on GitHub as well.
What gets sadly lost on GitHub sometimes is "emerging events" like a new release of ours or the game we mod breaking something, where we will get yelled at on the Discord immediately and might have a hotfix release out before anybody even managed to create a proper GitHub issue.
Edit: Oh and temporary workarounds. If we figured something out on the Discord it doesn't get posted to Github necessarily even if there is already an issue. Hence why I'm looking into having a bot for that instead of literally having to copy and paste a message.
I hate how devs use Discord for documentation. All the info on there is fleeting.
Guilty as charged, but in our defence we mirror most of the info from/to GitHub best we can. Also you can make the information somewhat less fleeting by pinning comments to a channel, using forum channels, or creating channels where users only have read access. Of course this doesn't prevent the data from going away if Discord does, but to be fair the same can be said about almost all other services as well. GitHub servers get ransomwared and they don't pay? Yeah your changes until their last uncorrupted backup are gone now unless you had backups of your own.
The reason why we use Discord in the first place though is network effect. The amount of reports and questions we get on Discord is simply no comparison to GitHub. It's more simply because more users already have Discord than do GitHub leading to a lesser barrier of entry (account creation/program installation), especially for gaming related projects like ours. Of course this creates some added bureaucracy for keeping track of important reports from Discord. It's kind of manageable to do manually, but I have been looking into ways of having a bot transfer messages/threads to GitHub by simply replying with an !issue 4321 command or something. Sadly I'm pretty sure we wouldn't get half the reports we do on Matrix/IRC/XMPP/whatever, same diff if we were to switch from GitHub to GitLab basically.
Lastly, a server owner (or someone given the rights by them) can get an API key that enables them to dump the full server logs to disk. So if you really want your Discord server content to be indexed by search engines the possibility to just host a copy of your logs as a static website is technically there (we admittedly don't do this yet, not sure if there are existing projects for this).
Know what data source isn’t fleeting? Forums.
Guess you never were a member of a forum with private sub-forums that went out of maintenance? That info is just as gone as our Discord logs if the company croaks tomorrow. And the public part is only available if it was mirrored to web.archive.org or something, which isn't guaranteed either.
In summary, yes Discord isn't the shit, it's just shit, but the people are there. If the mountain won't come to you, then you must go to the mountain. ¯\_(ツ)_/¯
https://knowyourmeme.com/memes/loss
TL;DR: Somebody made a really weird episode of a web comic and the configuration of the figures in the panel has become a meme named "Loss" after the name of that episode. Pic related:
The joke here is that evolution gave you pattern recognition to avoid predators, but now you are using it for useless things like recognising this comic is spatially organised in the same way the Loss comic is.
Soweit ein Passwort im Klartext gespeichert wird und ausgelesen werden kann, kann es voraussichtlich gar keine Sicherungsfunktion erfüllen. Das bloße Geheimhalten eines Passworts genügt nämlich nicht, mithin kan auch das “schlechte Verstecken” nicht genügen. Damit ist die Rechtslage weit entfernt von der Klarheit, die du suggerierst.
Das sah das Landgericht Aachen aber anders, weshalb sie den Fall zwecks Revision an das Amtsgericht zurückgegeben hatten:
Die Daten seien besonders gesichert gewesen, da ein Passwortschutz vorlag und das "Abrufen" der Daten "zudem nur nach einer Dekompilierung möglich war", heißt es in dem Beschluss des LG Aachen. "Die Sicherung des Zugangs mittels Passwort reicht als Zugangssicherung aus", damit sei der Straftatbestand erfüllt. Damit folgt das Gericht der seit Jahren auch an anderen Gerichten vorherrschenden Rechtsmeinung.
Der Rechtsauffassung der Aachener Richter nach "ist auf die allgemeine Sicherung der Daten gegenüber Zugriff Unbefugter abzustellen, nicht darauf, ob Eingeweihte oder Experten leicht auf die Daten zugreifen können". Der Beschluss folgt außerdem der Argumentation der Staatsanwaltschaft Köln, die in ihrer Beschwerde argumentiert hatte, dass die Dekompilierung einer Binärdatei "ein tiefes Verständnis über Programmiersprachen und Softwareentwicklung" voraussetzt, "um mit dem Ergebnis der Dekompilierung umgehen zu können".
a neural network with a series of layers (W in this case would be a single layer)
I understood this differently. W is a whole model, not a single layer of a model. W is a layer of the Transformer architecture, not of a model. So it is a single feed forward or attention model, which is a layer in the Transformer. As the paper says, a LoRA:
injects trainable rank decomposition matrices into each layer of the Transformer architecture
It basically learns shifting the output of each Transformer layer. But the original Transformer stays intact, which is the whole point, as it lets you quickly train a LoRA when you need this extra bias, and you can switch to another for a different task easily, without re-training your Transformer. So if the source of the bias you want to get rid off is already in these original models in the Transformer, you are just fighting fire with fire.
Which is a good approach for specific situations, but not for general ones. In the context of OP you would need one LoRA for fighting it sexualising Asian women, then you would need another one for the next bias you find, and before you know it you have hundreds and your output quality has degraded irrecoverably.
Yeah but that's my point, right?
That
- you do not "replace data until your desired objective".
- the original model stays intact (the W in the picture you embedded).
Meaning that when you change or remove the LoRA (A and B), the same types of biases will just resurface from the original model (W). Hence "less biased" W being the preferable solution, where possible.
Don't get me wrong, LoRAs seem quite interesting, they just don't seem like a good general approach to fighting model bias.
First, there is no thing as a “de-biased” training set, only sets with whatever target series of biases you define for them to reflect.
Yes, I obviously meant "de-biased" by definition of whoever makes the set. Didn't think it worth mentioning, as it seems self evident. But again, in concrete terms regarding the OP this just means not having your dataset skewed towards sexualised depictions of certain groups.
- either you replace data until your desired objective, which will reduce the model’s quality for any of the alternatives
[...]
For reference, LoRAs are a sledgehammer approach to apply the first way.
The paper introducing LoRA seems to disagree (emphasis mine):
We propose Low-Rank Adaptation, or LoRA, which freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, greatly reducing the number of trainable parameters for downstream tasks.
There is no data replaced, the model is not changed at all. In fact if I'm not misunderstanding it adds an additional neural network on top of the pre-trained one, i.e. it's adding data instead of replacing any. Fighting bias with bias if you will.
And I think this is relevant to a discussion of all models, as reproduction of training set biases is something common to all neural networks.
Just here to leave the daily reminder that API reimplementation may constitute fair use under certain circumstances.
Wundert mich jetzt nicht, da das explizit verfassungswidrig wäre nach Artikel 14 Absatz 3 Grundgesetz:
Ich meine Satz 3 gibt in der Hinsicht auf die Höhe vielleicht ein wenig Spielraum, aber gar keine Entschädigung geht halt einfach nicht. Der Artikel 14 unterliegt im Übrigen auch der Ewigkeitsgarantie nach Artikel 79, daher ist der nicht änder- oder abschaffbar. Wenn du den als Politiker also in Frage stellst ist das wie ein Freifahrtschein für den Verfassungsschutz dich unter Beobachtung zu stellen.