• 0 Posts
  • 78 Comments
Joined 1 year ago
cake
Cake day: August 8th, 2023

help-circle
  • Generating meaningful text in an image is very complex. Most of these models like Dall-E and simple diffusion are essentially guided denoising algorithms. They get images of pure noise, and are being told that it’s actually just a very noisy image of whatever the description is. So all they do is remove some noise for many steps in a row until a clear image emerges. You can kinda imagine it as the “AI” staring into the noise to see the image that you described.

    Most real-world objects are of course quite complex. If it sees a tree branch in the noise, it also need to make sure that the rest of the tree fits. And a car headlight only makes sense if the rest of the car is also there. But for text these kind of correlations are even way way harder. In order to generate meaningful text it not only needs to understand how text is usually spaced, and that letters usually are written in a consistent font, it also needs to learn the entire English language. All that just to generate something that is probably overall of less influence to it’s “score” on images form the dataaset than learning how to draw a realistic car.

    So in order to generate meaningful text, the model requires a lot of capacity. Otherwise, since it’s not specifically motivated to learn to write meaningful text, it’ll do whatever it’s doing now. Honestly I’m sometimes quite impressed with how well these models do generate text, given all these considerations.






  • I’m not sure if the writer of this article is familiar with Dutch politics, but nothing unusual is happening. Unlike in the US, where one party wins am absolute majority, we have a system where you need to form a coalition. And this is obviously a whole political game, with parties kinda pretending they don’t want it just to have a better bargaining position. This formation process has taken months for as long as I can remember, and personally I don’t feel like it’s going particularly badly for them.

    I still think the coalition of PVV, NSC, VVD, BBB will happen in some way or another. But the other parties do want the PVV to make clear that some of it’s plans are just not going to happen. Wilders is aware of this, and suddenly seems a lot more reasonable than I’m used to (though obviously still far-right). Ideas like a Nexit or a full immigration stop are just not executable, so they’ll have to be toned down.



  • Baudet is definitely on the edge yeah. Obviously there somes a moment where you shouldn’t tolerate the intolerant and a democratic society shouldn’t protect the undemocratic in the name of democracy. Up until now I don’t think either of them has done enough to deserve that tho. Randomly calling for a politician to be killed, even as a joke, is in my opinion still bad taste. It’s not the way I’d want Dutch politics to go, even for those who I disagree with. Violence and hate only creates more violence and hate


  • This is not a joke that is very appropriate though. In the early 2000s a politician similar to Wilders was killed for his opinions. Another (even further) right-wing politician was assaulted twice in the past few months, and Wilders himself hasn’t been able to go anywhere without bodyguards in the past few years. We’re still a democracy, Wilders should be able to say things without getting constant death threats. Unlike the US, things are not yet as polarized here. To me it feels like the comedian made a very inappropriate “joke”, even if I also absolutely despise Wilders


  • Obviously this is a privacy community, and this ain’t great in that regard, but as someone who’s interested in AI this is absolutely fascinating. I’m now starting to wonder whether the model could theoretically encode the entire dataset in its weights. Surely some compression and generalization is taking place, otherwise it couldn’t generate all the amazing responses it does give to novel inputs, but apparently it can also just recite long chunks of the dataset. And also why would these specific inputs trigger such a response. Maybe there are issues in the training data (or process) that cause it to do this. Or maybe this is just a fundamental flaw of the model architecture? And maybe it’s even an expected thing. After all, we as humans also have the ability to recite pieces of “training data” if we seem them interesting enough.



  • I’m Dutch and afaik right-wing media doesn’t play a role. The reason is quite simple, it’s because the previous coalitions lead by the neoliberal VVD have consistently fucked everyone who isn’t rich (and even some who are). Especially the crises that they’ve caused have completely eroded trust in the previously big parties. Some of these issues I’ll lost below.

    In the northern province of Groningen houses started getting damaged die to earthquakes caused by government-funded natural gas extraction. The whole country got rich of of this for decades, but when the people of Groningen needed money to repair their homes they sometimes had to fight for years to get anything.

    The tax agency started incorrectly marking people as fraudsters for claiming certain benefits, which often put these people in debt for thousands of euros. This has ruined the life’s of hundreds of people, yet the government initially just ignored it. While these people lost everything they had, the government seemed to do everything it could to avoid paying or helping. At some point they did start paying and pretending to help, but for many victims this was simply too little to late.

    And then there’s plenty of other issues that erroded trust as well. Years of ignoring “the nitrogen problem” until it was too late and lead to all kinds of issues. The farmers (who are the biggest cause of this issue) suddenly become the target of the government, which they obviously didn’t like. This caused huge protests and the rapid growth of parties like the BBB and probably also the PVV. And there’s also a housing crisis, where buying houses is impossible.

    In short: the previous normal coalitions kept fucking up, until people seemingly had enough. There were other alternatives to the PVV, but apparently voting on a right -wing extremist who has all the ideas is easier than voting left-wing for many people here. I hope the damage stays limited, but this is truly a dark day for our country.



  • I use mostly pure GNOME on my PC. The missing system tray is very annoying, though I really don’t care about the missing dock or desktop icons. I start programs by searching for them, not by clicking on a desktop icon or in a dock. That’s my preferred method. And the alt+tab menu of GNOME is nice enough to find anything when I need it, together with having workspaces to organize everything.

    I tried KDE, but it just felt more messy to me. For instance, I tried to move the bottom bar to the side of my screen, accidentally moved something on the bar itself, and then everything looked off and I couldn’t figure out how to get it back. It all felt a bit janky and unrefined. On some systems I also use i3, but that’s only for productivity. For daily use I prefer GNOME, where everything just seems to work.



  • I’m noticing the same. Maybe the chill people are slowly having enough of the more toxic people and are going somewhere else. I’m not really enjoying the constant barrage of far-left biased news and opinions. I just want to chill but it seems to permeate everything. Things aren’t always great in the world, but personally I come here to not be reminded of that.



  • $500 is still insane to me. I’m Dutch and I’ve been riding bikes pretty much daily for my whole life. I don’t think I’ve ever used a bike that cost more than €200. Almost every bike I’ve used is either secondhand or a hand-me-down. My current bike has a front wheel that has been folded into a 90-degree angle and back (which is noticeable), has only one working brake, and only 2 of the 7 gears work well. Yet for the past 2.5 years I’ve reliably and comfortably driven trips without issues. Whether it is a 10-minute trip to the supermarket or work, or a 40-minute trip to the next city over for some party, it always gets me there.

    I don’t really understand the sentiment in this comment section that a bike should always be a fancy new bike of $500 or even $3K when a trusty old rustbucked could also get you the same distance for a hundred bucks or even way less.


  • I bought Cities Skylines 2 under the premise that I’d refund it if it ran like shit. On my PC it runs about the same as the first game. Which isn’t great, around 30-60 fps, but also not unplayable. I haven’t played a lot yet, but so far I’m enjoying it. I did get into a fight with the UI thingy that lets you designate an area for landfills/farmland/etc. It feels very counterintuitive when you build the respective building and then try to mark the area. But when you finish the building and then edit the area again it suddenly works a lot better. Maybe my brain is just weird