awegio

In their FAQ

> How does it work?

> Using a "Virtual Distributed Filesystem" (VDFS), in other words; a decentralized database that emulates a filesystem. It indexes hardware filesystems to create a master database that is synchronized in realtime between your devices running Spacedrive.

> What makes this different to Dropbox or Google Drive?

> It is not a storage provider, Spacedrive is simply a database that exists on top of existing storage layers, from cloud services like Dropbox, Google Drive and iCloud to physical devices and external drives you already own. It doesn't provide you more storage, rather a supercharged view of your existing storage.

So more like Syncthing? Or rather Windows File Sharing/Samba? I don't really get it

It depends on your web browser. Just see what happens here https://webauthn.io/

Firefox on Desktop tells me to "touch my security key". Not sure how that works. Firefox Android gives me a few hardware options to store my passkey to. Chrome Desktop asks me to enable Bluetooth. Chrome Android asks which Google Account to use.

You need a compact encoding for chess engines that explore billions of states as fast as possible to plan the best move.

> There's exactly one hit

There are N possible sequences, and you try N times with a success probability of 1/N each (because it is a good hash function). This means the expected number of hits is 1.

A language model estimates the probability of a sequence of words P(w_1, ..., w_n) or equivalently P(word | context).

For compression, word sequences that have higher probability should be encoded with shorter codes, so there is a direct relationship. A well known method to construct such codes based on probabilities is Huffman coding.

This works whether you use a statistical language model using word frequencies or an LLM to estimate probabilities. The better your language model (lower perplexity) the shorter the compressed output will be.

Conversely, you can probably argue that a compression algorithm implicitly defines a language model by the code lengths, e.g., it assumes duplicate strings are more likely than random noise.

> put programs in custom sandboxes, where /bin/bash is found exactly where they expect to find it

NixOS can do that, I think it was necessary to get Steam games running

https://nixos.org/manual/nixpkgs/stable/#sec-fhs-environment...

Why? Only if you would reuse the same data multiple times, or if you use a unique random generator that is easily distinguishable from other data. So maybe don't use the 5x5 island, but when done right the approach shouldn't help fingerprinting?

That's probably because you confused Vim and Kakoune key bindings.

Kakoune

- n: next match

- alt+n: previous match

- shift+movement: extend selection by movement

Vim

- n: next

- shift+N: previous match

So if you press shift+N as you are used from vim, you start adding a lot of selections instead of going to previous match. I believe this difference is the most confusing for people who switch from Vim to Kakoune.

> Node packages work much better

Are you sure about that? I haven't seen a node app built from source on nixpkgs yet. That includes Electron apps like Signal Desktop, which is a bit disappointing.

There is this article about trying to package jQuery on Guix:

http://dustycloud.org/blog/javascript-packaging-dystopia/

DeepMind's blog post on AlphaDev says:

> AlphaDev uncovered faster algorithms by starting from scratch rather than refining existing algorithms

Finding that specific optimization, especially when given the comments, seems almost trivial by comparison.

Edit: I tried to understand the optimization in question. This is not the full sort3 algorithm, but only under the assumption that B < C. In that case the GPT-4 answer is actually wrong because it wasn't given that assumption.

Then just use an Ad Blocker. The actual problem is SEO spam though.

> mental models, each with a degree of uncertainty, to some overall business strategy

That actually sounds a lot like model based reinforcement learning. AlphaGo still had an explicit model of a perfect information game, but newer algorithms are capable of learning the rules of games like Minecraft with very little prior information.

The problem is that the real world is still much more complicated and hard to access for machines. However, I wouldn't say it is a fundamentally different problem, but rather a matter of scale. That's part of the reason why AI research spent so much effort on building game playing systems.

It looks like this is a high level interface more like Intents on Android, e.g. in the video it opens the Logo Templates page of Adobe Express probably without automated UI interaction.

> It sees what you're seeing.

In the video the user drag&drops a file into the copilot window. How do you know that it can see your screen? I believe "works with all apps" only refers to the third party plugins.

If I understand correctly, the proposed act restricts the use of spyware, and apparently some actors want to keep some loopholes.

The title reads like the EU wants to allow uses of spyware that were not allowed before, but I cannot infer that from the very limited information given in this article.

According to the internet what you wrote is the distribution of min(random(), random()), but the distribution of the product is different.

Product: https://math.stackexchange.com/questions/659254/product-dist...

Minimum: https://stats.stackexchange.com/questions/220/how-is-the-min...

I can't explain it either, but I also don't think there is a reason they should look the same.

In more formal terms, since it is a uniform distribution you can write

  random(x) = random()*x

then you can easily see that

  random(random(random())) = random()*random()*random()

The resulting distribution is not equal to random()^3, because e.g. the probability that all 3 random calls give you a number <0.5 is only 0.125.

How did you realize it here? This user has multiple comments in this thread but this one actually sounds more normal than the others.

I find it very uncanny to see comments like this that sound like ChatGPT but are surprisingly relevant to the discussion.

So like almost every news website they added a title picture to every article to drive engagement. The problem here is that they used a picture where a face was visible, they have to use some generic image instead.

The whole trend of adding stock pictures to everything only creates distractions. It's bad for the reader, good for the publisher. The pictures are either useless or misleading.

Well, it's about workforce, not the shareholders. There is probably some correlation between morale and stock price, though.

Related: The Hardware Lottery (https://arxiv.org/abs/2009.06489)

It goes both ways: processors are built to accelerate existing applications, and applications are built to run with the constraints of the hardware. This holds back progress towards better systems where both new kind of hardware and software is required.

This is part of the reason why research into neural networks was so slow for a long time, and took off only after GPUs for 3d graphics came around.

Here is the original post by the photographer in German:

https://www.alltageinesfotoproduzenten.de/2023/04/24/laion-e...

He sent an initial demand that LAION declined. He did not want to give up, then he made his lawyer send a letter, then LAION paid a lawyer to answer, and the invoice is for the cost of the lawyer.

What I find interesting is that the photographer admits in the comments that he sells thousands of AI-generated images, probably made with an AI trained on LAION's dataset. He only doesn't want links (with alt text) to his own images in the dataset.