ath92

So do you start sweating because of excess heat? Do you suddenly feel more energetic?

The paper is about biases in GPT-4. They're not talking to GPT-4 to determine things about the real world. They're talking to GPT-4 to determine things about GPT-4.

> I think there are techniques you can use for less noisy selection of points

Do you happen to know what those techniques might be called? Kinda hard to search for these kinds of things without knowing the right terms.

That's a funny coincidence, I recently made something similar: https://webgl-particle-life.netlify.app/ (source: https://github.com/ath92/particle-life)

This version does everything in webgl shaders and keeps all state for the simulation in textures / uniforms. This allows it to simulate and draw more particles. Unfortunately it may not run on all devices because it uses some less supported webgl extensions.

That “game loop” exists in the browser too. It’s just the main event loop, and it’s been there since the inception of JavaScript. Not saying everything that web developers aren’t reinventing the wheel, but I think the reason for it isn’t a lack of tools, it’s a combination of poor ergonomics of the existing tools and (perhaps more importantly) people not understanding those tools and building frameworks for things that are included by default.

> upward mobility in German society is very low compared to the US

This appears to be untrue: https://en.wikipedia.org/wiki/Global_Social_Mobility_Index

Germany ranks 11th on social mobility, while the us ranks 27th.

There’s a difference between your own democratically elected government spying on its citizens and some other government you have no control over doing the same. Not saying that it’s good that the Dutch government was spying on its citizens, but it’s worse when it’s a foreign super power.

I’d never seen this explanation of what a cross product is before. It really clicked for me, even just reading it as text. Thanks!

This implementation does not support view-dependence though (mentioned in the readme)

The 1999 equivalent for Nvidia stocks would have been something like Sun (as mentioned in the article) or Cisco (because they sold routers which everyone thought were essential).

A full body scan is a bit like p-hacking. If you have a hypothesis of what might be wrong and perform a specific test to see if some value is within normal range, there’s relatively little chance of a false positive.

If you test everything at the same time, some values are bound to be outside of the normal range, just because you’re measuring so many things at once. Doing any kind of procedure to further investigate or treat the symptoms has a risk associated with it.

In the end, doing a full body scan, with current medical technology (both in terms of the test equipment as well as our means to treat anything that comes up as a result of the test) is a risky thing to do. It sounds nice in theory, but unfortunately we do not live in the utopia you speak of.

> There may be a future where an AI is smart enough and demonstrates enough analytical, cognition, and deduction ability that it can take over specifying what a computer should do

When we reach that point, anyone's job will be at risk. Give that AI an AWS account and a credit card and it'll take over the world, not build websites when someone asks it to.

source: https://github.com/ath92/particle-life

Here's a demo from about 3 years ago that does something similar to the experiment in the article: https://compassionate-murdock-82fdbb.netlify.app/

It's interesting to see that these face detection models have been getting a lot better over the past couple of years, making it easier to stable and cheap webcam-based tracking.

That said, if you look closely at the videos in the original article, there's some noticeable delay between the camera moving and the program responding. The original wii controller based prototype that this was based off doesn't suffer from this delay because it used infrared LEDs. It's annoying to have to place those somewhere near your face, but once you have them there the face tracking because basically instant.

A NeRF is not the same as an SDF though. NGP (the paper by Nvidia linked here) can train NeRFs and SDFs, but I don't know of any straightforward way of extracting an SDF from a NeRF.

And while it's true that there are methods for extracting a surface from a NeRF, achieving a high quality result can be challenging because you have to figure out what to do with regions that have low occupancy (i.e. regions that are translucent). Should you consider those regions as contained within the surface, or outside of it? Especially when dealing with things like hair, it's not obvious how to construct a surface based on a NeRF.

Amazing work. I think there's a whole host of apps that just store some small amount of data for each user that could work entirely on something like this.

Some nice things:

- Anyone can host their own Matrix homeserver, which can then sync its data within a federation, allowing users pretty fine-grained control over what data they share and with whom.

- Add end-to-end encryption, and it becomes even better: data stored somewhere in the cloud (i.e. the user doesn't have to manage a server), but the server knows almost nothing about what you've stored aside from some metadata.

For some additional context, when the original NeRF paper (https://arxiv.org/pdf/2003.08934.pdf) was published 2 years ago, it reportedly took at least 12 hours (depending on hardware used of course) to train on the scene with the bulldozer. This has now been reduced to about 5 seconds (!), with realtime rendering of the result.

There's also yjs:

https://github.com/yjs/yjs

For most people, having only the equations means they cannot build an intuition for what a fourier transform actually is, because they don't have a solid enough grasp of math notation to easily make sense of what the dense equations actually represent.

For those people it's very helpful to first get a sense of what it all means (even if it means that initial mental model is somewhat imprecise), because it makes understanding the equations easier.

That example you mentioned would be just a couple of lines of html with css (which you could inline if you wanted), and there are plenty of drag and drop tools for the web (WebFlow is a modern example, but they have been around for decades).

I think part of the problem is also that it's sometimes more difficult to implement a nice and usable design. Instead, it may be easier to write code that does everything it needs to, but forces the user to jump through hoops to get what they want.

Sometimes making software more user-friendly may even cause the code to get more "ugly", because instead of a nice and elegant piece of software that only solves the core problem, you now have to deal with all the weirdnesses of how people perceive the problem as well. Taking away complexity on the user's side almost always adds complexity to the code. And if you're the developer, it's very easy to think that the tool you created is perfectly usable, because you understand everything about the mental model that went into it.

Getting paid properly is important of course, but intrinsic motivation (of which a sense of autonomy is an important driver) is very important to keep workers happy, independent of whether they're doing crowdsourcing work or just a regular job.

I find it hard to believe that this platform will be competing against Amazon Mechanical Turk on price, so they'll have to compete in other dimensions. Increasing worker motivation to improve the quality of the output is a win-win for all the stakeholders, so IMO it's worth pursuing.

Before Airbnb, your neighbour couldn't just start renting out (a part of) their house to tourists. In many European cities, and I imagine other places too, zoning laws and permits were in place that prevented that from happening. Then Airbnb came along and disrupted the market by rewriting the rules, or rather selectively forgetting about some of them.

This has a real impact on the people that live in those cities. Without it, there would still have been plenty of people that wanted to travel to these places, but without part of the housing supply meant for citizens going to tourism, the price of a room would simply go up.

Neither zebras nor white pages in a hardback book emit light, they only reflect it.

I think frameworks like tailwind are great for rolling your own components actually. Tailwind sits somewhere between just writing your own styles from scratch and using a css component library like Bootstrap. It provides you with sensible atomics (e.g. a nicely defined range of margins/paddings that allow you to create UIs with consistent spacing), but it still allows you to mix and match where needed.

This approach allows tailwind to be terser than raw inline styles, while giving you the ability to write a lot of your styling inside the markup, which prevents context-switching. It's a nice balance between many of the approaches that have come before it.

Well JSX is not really just javascript/typescript - it needs to be transformed first, just throwing a .jsx file into a script tag won't work. It does compile down to just javascript, but the same can be said for svelte.

I can see the argument for JSX being more flexible, given that you can store little bits of JSX in js expressions, something you typically cannot do with the other component frameworks. But tools like svelte have their own DX improvements that make things like state management / reactivity arguably a lot easier than React.

Do you think the suicide rate could be related to the cultural pressure to work long hours?

> who let themselves be underpaid

So what should these people do instead? Stay where they are and get paid much less?

It's the combination of being a Nordic social democratic + having large oil reserves.

Energy is not an infinite resource. The energy used to mine bitcoin could have been used for something else instead.