Now I don't know if they ever changed anything since launch. But if you judged the max speed by the first flying saddle you got you didn't actually experience anything close to max speed. Pals that have their saddles unlocked at a higher level (usually) have a much higher speed when mounted.

So is the example with the dogs/wolves and the example in the OP.

As to how hard to resolve, the dog/wolves one might be quite difficult, but for the example in the OP, it wouldn't be hard to feed in all images (during training) with randomly chosen backgrounds to remove the model's ability to draw any conclusions based on background.

However this would probably unearth the next issue. The one where the human graders, who were probably used to create the original training dataset, have their own biases based on race, gender, appearance, etc. This doesn't even necessarily mean that they were racist/sexist/etc, just that they struggle to detect certain emotions in certain groups of people. The model would then replicate those issues.

Source

People mention Spore because the official FAQ mentions Spore.

Thrive is never gonna be “from puddle to space adventures”-type of game.

People also mention Spore because this is exactly what the devs are envisioning. To quote the FAQ:

Gameplay is split into seven stages – Microbe, Multicellular, Aware, Awakening, Society, Industrial and Space.

I'm still not 100% trusting that. Any time a dev comes up with a new feature like this one, they might forget to implement a check if the game is privated (or do the check and mess up properly hiding it).

It's not as accurate as you'd like it to be. Some issues are:

• It's quite lossy.
• It'll do better on images containing common objects vs rare or even novel objects.
• You won't know how much the result deviates from the original if all you're given is the prompt/conditioning vector and what model to use it on.
• You cannot easily "compress" new images, instead you would have to either finetune the model (at which point you'd also mess with everyone else's decompression) or do an adversarial attack onto the model with another model to find the prompt/conditioning vector most likely to create something as close as possible to the original image you have.
• It's rather slow.

Also it's not all that novel. People have been doing this with (variational) autoencoders (another class of generative model). This also doesn't have the flaw that you have no easy way to compress new images since an autoencoder is a trained encoder/decoder pair. It's also quite a bit faster than diffusion models when it comes to decoding, but often with a greater decrease in quality.

Most widespread diffusion models even use an autoencoder adjacent architecture to "compress" the input. The actual diffusion model then works in that "compressed data space" called latent space. The generated images are then decompressed before shown to users. Last time I checked, iirc, that compression rate was at around 1/4 to 1/8, but it's been a while, so don't quote me on this number.

edit: fixed some ambiguous wordings.

I think it's much more likely whatever scraping they used to get the training data snatched a screenshot of the movie some random internet user posted somewhere. (To confirm, I typed "joaquin phoenix joker" into Google and this very image was very high up in the image results) And of course not only this one but many many more too.

Now I'm not saying scraping copyrighted material is morally right either, but I'd doubt they'd just feed an entire movie frame by frame (or randomly spaced screenshots from throughout a movie), especially because it would make generating good labels for each frame very difficult.

Scarlet and Violet sold 10 million copies during the launch weekend, so I think they'll be pretty content to keep churning out mediocre games at a much faster pace.

Or into a horse.

I've seen this reposted on the original 196 on Reddit multiple times so saying this is the worst 196 on the basis of this meme being posted makes no sense.

Also nazis are very unlikely to be participating in this community anyway, and if they are then they are either hiding to the point of indistinguishably or getting the ban-hammer really quickly. In the latter case, the problem is solved by the mods and in the former case, with the internet's anonymity, someone fully to be a member of a digital community is just a regular member of the community.

Tankies on the other hand share many more values with the core demographic of this community so they might be less inclined to fully hide their views and their views simmering through might not immediately get them a ban (depending on what they let shine through, of course).

I disagree. The placeholder should be 0 (or even better, a negative number if the system allows it).

Otherwise you'll have people who will end up paying the $1000, believing it to be what they owe.

In a strange way, having it be ridiculously high was a benefit since it got this person to complain. If it was something high but still within in the range of plausibility, he might have overpaid. That being said, it being ridiculously high could also (as the article suggests) be used nefariously to scare people into showing up in court.

I think it would be fine as an official extension. Shipping it built-in feels weird to me.