> Bendels claimed the meme, posted by his newspaper's X account, was satirical.

> But the judge in the case said during the verdict that Bendels published a 'deliberately untrue and contemptuous statement about Interior Minister Ms. Faeser (...) that would not be recognizable to the unbiased reader and is likely to significantly impair her public work'.

Spotify has discovered there is a big market for music where the quality isn't that important and they can serve it themselves. Same as supermarkets do with many products.

This existed, it just didn't exist in the traditional media. Look in the right discord servers and there were new videos being posted every 5 minutes as the convoy was moving around.

What’s harder is explaining why ChatGPT would lie in this way. What possible reason could LLM companies have for shipping a model that does this?

It did this because it's copying how humans talk, not what humans do. Humans say "I double checked" when asked to verify something, that's all GPT knows or cares about.

Because the very changes in brain chemistry you're talking about affect your ability to determine whether something is real or not.

Many psychedelics cause Apophenia, Pareidolia and other disorders. In loose terms, the ability of your brain to determine what information is meaningful or not is affected and you begin to over interpret the meaningfulness of things that happen and your own thoughts. So your idle thought becomes some earth-shattering revelation.

Something similar can happen to people with Schizophrenia. They'll be walking down a street and see a license plate XUE-383 and the function of the brain that tells whether something is meaningful or not misfires, telling them that this is very important. The brain then tries to explain why this is incredibly important, hence hidden messages, spies, all sorts of things.

You can of course be on drugs or schizophrenic and understand this is what is happening but it's hard to override an emotion with logic.

The simplified and (generally) commonly accepted narrative is that a cohesive group arrived here 45-60,000YA, lived peacefully and in harmony with the land right up until Europeans arrived

Is absolutely what a lot of, if not the majority of people, believe.

Yes, real life is hard and complicated and there's no easy way to decide it. We collectively do the best we can and keep trying to improve. That's just reality. That's why free speech absolutism is naïve- it appeals to our desire for one, simple, always enforceable rule but it's not the way the world works that one simple rule is always the best outcome.

My first attempt at creating a simulation of water and erosion before I saw this paper was pretty bad as well.

Yes, it's a grid based system. When you choose map size (e.g. 1024x1024) that's the resolution of the underlying simulation. My refinement of this approach is there is an extra step that "smooths" everything out. The vertex points for rendering are offset depending on how the water flows, which allows rivers to appear to flow at angles other than 90 degrees. Without this, it would either be very blocky or cover a much smaller area- there are other implementations of that paper floating around and the scale is an order of magnitude less, as a river needs to be many grid points across. In my implementation, you can get a nice looking river from a sequence of single points.

The drawback is this complicates a lot of other things. E.g. you can no longer easily tell what the height is at a point on the map is- there is no easy way to know what vertices to interpolate between since they can be offset in different directions. So when rendering trees, there's a whole GPU calculation to produce a separate data layer just to know what height the ground they're sitting on is at. But it's worth it to get a map that covers a big area with a full water flow simulation.

In general terms, the simulation is done as a series of grids of values (called layers, in reality just textures on the GPU) and operations that take some of them as input to generate another. Some layers persist always (e.g. the water flow) and update themselves iteratively. Other layers are produced on the fly when needed, and can be at different resolutions. E.g. you zoom into an area of the map- the water layer is read to determine the biome layer for just that part of the map. The water layer is read to determine the heights layer I talked about previously. Then the biome layer, water layer and heights layer are combined into a trees layer where one grid vertex is one possible instance of a tree. The values represent whether a tree is there, the position of its base, the type of tree, etc.

That's used as input for the rendering system, there's a mesh that uses that trees layer to render all the trees. It's also used as input for the animals system, so the animals walk around the trees, etc. It's basically a big dependency graph of data layers and the operations needed to generate them. Most of the work so far has been building all that tooling and getting it running asynchronously so the game doesn't lag if it takes multiple frames to generate what is needed.

I might gamify it a little bit more, but personally I love the fun that comes from exploring a set of mechanics by yourself without anyone telling you what you're trying to achieve.

My aim is to build a set of simulation mechanics that are accurate enough to teach you a little bit about how the processes work in real life but also responsive enough to let you play around with them and create what you'd like.

I get this feedback a lot, but I'm unsure what formats would be most useful- there are a lot of different heightmap / GIS formats to choose from.

The rivers don't meander in the way that real rivers do- real rivers meander because of differences in flow velocity across the width of the river, but the simulation doesn't run at that sort of resolution. If it did, the area covered would be far smaller.

I just released a new version after first releasing it 18 months ago. It's been a hard slog to scrounge time as it's not my day job but that's life.

There's a ton of obvious improvements in the future (weather systems, temperature, varying rainfall, ice/snow, more biomes, more plate tectonics, lava, etc) but all suggestions/feedback is welcome.

Comments URL: https://news.ycombinator.com/item?id=34396026

Points: 229

# Comments: 79