android auto

First I heard of this, but since it seems to be just some software that runs on the hardware of car manufacturers it seems rather unlikely. But very theoretically possible, if the car manufacturer was using default process scheduling in a CPU constrained machine and now switches to real-time scheduling in an update. But that was possible for years before this news, the code has just been mainlined to the default kernel now. If the car manufacturer cared about that they would probably have done it already with a patched kernel.

[…] a public institution is really not a great example of the general population […]

Which I touched upon in my disclaimer, but in some ways it is a great example. Public institutions are defined by the general population, indirectly through their representatives creating the rules that govern them, and directly through contact with the public at large. Now if all our institutions still use this very outdated technology, and you can have trouble convincing them - during a global pandemic mind you - that using email is just as safe as using fax (so not safe at all basically), then that speaks to a larger mindset in the general population.

Many in the general public are also a lot quicker, some might even say careless, with adopting new technology of course. But as a society we are rather slow, and there are surprisingly many individuals who are hesitant or entirely resistant to adopting new technology. The fediverse usage is a bubble in a bubble here.

The internet infrastructure is another good example for this on the societal level, as there were plans in the 1980ies [!] to lay out a glass fibre network between every publicly used building in the country, which would have gotten us a good part of the way towards adopting this new material at scale. But in the end it was deemed unnecessary and too expensive and the project got canned (mixed in with rumours of “close friendship” between the chancellor and a major copper producer). Instead now we have people running around thirty years later and collecting signatures at the door for last-mile fibre network projects that seldom make quorum and thus almost never materialise public funding.

2. […] But also how are Germans technologically behind regarding common personal life?

I bet you wherever in Germany you are, if you go to the website of your local city government right now they will have a still active fax number in their contact information. I guarantee it. Well if they have a website that is.

Which is a bit silly as an example but highlights the central problem, which is that adoption of new technology happens at a glacial pace, especially in public institutions. There are many reasons for that of course, some good, like the aforementioned inclination towards privacy, some bad like whatever allows fax machines to still be around.

And don’t get me started on internet infrastructure… In an international comparison we certainly aren’t leading the field regarding adoption of new technologies.

Oh yeah, can’t use the same IP range as your LAN, that will lead to problems. :D Glad it’s fixed.

Out of curiosity, does forwarding work now without the output (-o) command in PostUp?

Like I said in another thread on this post, I’m pretty sure that’s because they are forwarding input but not output in the PostUp rules. Setting a /32 in AllowedIPs works fine for me.

What are you trying to say? That reply also shows AllowedIPs set to a /32 on the server side.

I don’t think that’s what the setting does. Anyway, I have them set to a /32 IP in my server config and it works nonetheless. I get full access to the /24 behind the server from the client.

You have ALL traffic being routed over Wireguard here.

Please correct me if I’m wrong, but isn’t it the other way around? All Wireguard traffic is forwarded to the local interface.

I think the problem might be your PostUp/PostDown lines have an in-interface (-i) but are missing an out-interface (-o) for the forwarding. Try this:

PostUp   = iptables -A FORWARD -i %i -j ACCEPT; iptables -A FORWARD -o %i -j ACCEPT; iptables -t nat -A POSTROUTING -o ens3 -j MASQUERADE
PostDown = iptables -D FORWARD -i %i -j ACCEPT; iptables -D FORWARD -o %i -j ACCEPT; iptables -t nat -D POSTROUTING -o ens3 -j MASQUERADE

The number of littered bottles, with or without a cap, is greater than the number of loose caps,

That smells like survivorship bias. Your dataset is skewed by loose caps being way harder to find due to being smaller. It stands to reason that all those bottles without a cap you find will have also had their cap littered in the vast majority of cases.

Linux users can’t even agree […] so how am I supposed to pick one with any confidence?

Easy. You make a post like the OP, count the positive mentions of distros in the comments, and bam, you have your distro of choice. It’s called the Linux newbie roulette and works kind of like the magic hat in Harry Potter that sorts you into your house.

So, I think it’s pretty stupid to argue whether “convicted felon” should be in his opening lede line for Wikipedia.

True though that may be, I don’t think it’s surprising that this would happen, and since making the post I have been falling down a rabbit hole of finding out how Wikipedia is handling situations like this, partly through taking more than a glancing look at the talk pages for the first time ever, and it’s fascinating.

Currently my deepest point of descent is this sub-thread on the Admin board about the “consensus” boxes on top of talk pages being an undocumented and unapproved feature.

If you rise anywhere above lever 5 or so, the difficulty ratchets up so much it makes the main quest nearly impossible to complete.

Didn’t Oblivion already have the difficulty slider? You could just adjust that, no?

I know level scaling is a big topic in the industry, but for me, the way it’s implemented nearly ruins what is otherwise a mostly great game.

Two of the first RPGs I played were Gothic and Gothic II which released approximately alongside Morrowind and Oblivion, and they just had no dynamic level scaling at all, so I don’t really see the appeal either. A tiny Mole Rat being roughly the same challenge as a big bad Orc just breaks immersion. If you were to meet the latter in early game it would just curb stomp you, which provided an immersive way of gating content and a real sense of achievement when you came back later with better armour and weapons to finally defeat the enemy who gave you so many problems earlier. Basically the same experience you had with Death Claws in Fallout New Vegas when compared to Fallout 3 - they aren’t just a set piece, they are a real challenge.

The games had their own problems, for example the fighting system sucked, and I’m told the English translation was so bad the games just flopped in the Anglosphere, putting them squarely in the Eurojank category of games. But creating a real sense of progression and an immersive world were certainly not amongst their weaknesses.

So WINE was just imagined into existence? Or maybe it was a wizard with a magic spell?

GP is simply wrong on this one. While it is an open source project with a lot of volunteer involvement, there are companies like CodeWeavers and Valve which directly or indirectly contribute to development. You can get support from CodeWeavers AFAIK, but that means paying them.

Why do people get so uppity when I simply ask questions? I never claimed that anyone owed me anything. I never asked for anything.

Well you did ask for something, which is replies to your questions. And your reaction to those replies, whether intended or not, comes off as “uppity” as well. Hence the downvotes and hostility (not to say that I support that from either side of the conversation).

I am unwilling to learn.

Then why are you wasting peoples time with asking questions?

I’ve wasted hundreds of hours trying to learn to use Linux for basic tasks after everyone assured me it was “so easy” and not gotten anywhere. I’m done trying to learn.

Running software on an OS it wasn’t made for is anything but a basic task. Try running various Linux software on Windows and you will see. If you want to run software made for Windows easily the way to do that is using the version of Windows it was created for.

What people mean by “basic tasks” is usually browsing and office, and there is Linux-native software for that.

Someone posted Zorin OS elsewhere, which appears to be exactly that.

Not really. It has deeper integration of Wine into the system by default, but it is still a Linux OS running a compatibility layer for Windows software. This will not save you if you are unwilling to learn, there will still be various problems. Some software will simply not work, or only partially work, or require additional configuration to work.

In summary, if your definition of “basic tasks” is running arbitrary Windows software then doing it on Windows is the way to go.

pandoc.org is probably what you are looking for, but you might have to create a custom reader/writer or find one on the internet.

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.

Depressingly, the message that GHG emissions are heating up the planet has been passed down for over a hundred years now. People just aren’t very good with passed down messages in general.

Yeah but that’s my point, right?

That

1. you do not “replace data until your desired objective”.
2. 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.

1. 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.

“Inclusive models” would need to be larger.

[citation needed]

To my understanding the problem is that the models reproduce biases in the training material, not model size. Alignment is currently a manual process after the initial unsupervised learning phase, often done by click-workers (Reinforcement Learning from Human Feedback, RLHF), and aimed at coaxing the model towards more “politically correct” outputs; But ultimately at that time the damage is already done since the bias is encoded in the model weights and will resurface in the outputs just randomly or if you “jailbreak” enough.

In the context of the OP, if your training material has a high volume of sexualised depictions of Asian women the model will reproduce that in its outputs. Which is also the argument the article makes. So what you need for more inclusive models is essentially a de-biased training set designed with that specific purpose in mind.

I’m glad to be corrected here, especially if you have any sources to look at.