tarpitt

I don't have the ability to downvote, and I disagree that the downvote should be used, but I disagree with your disagreement about the downvote.

The proposal that seems to infringe on free speech, but mostly I take issue because there just isn't really a precedent for what such a thing would look like. What constitutes doxxing? It is easy to accidentally doxx someone by mentioning some offhand fact in conversation, because you may not know what series of facts can be connected to form a doxx. Typicially, people doxx themselves by accident, and someone else is merely pointing it out. I know I have many times. You live and learn.

There isn't really a right to privacy in the way that there's a right to free speech. Free speech is important to processes of public transparency and justice which I think this would interfere with. But it's also true that justice is blind and doxxing can interfere with a judicial process.

Also, if everyone is going to have access to super-stylometry tools in the future, which is the premise of the article that GP reacts to, it will be fruitless to uphold such a right because anyone can just run doxxyou.exe themselves. There's no need to spread doxx because it can be reproduced individually.

The term doxxing came from a certain hacker culture where it was implied that you connected some real-life identity to a criminal pseudonym. It essentially meant "snitching". So the idea that doxxing itself would be a crime is interesting. It's a reversal of the original meaning.

But it's also true that the public has discovered sybil-suseptible techniques (like swatting) where you can screw with ordinary people's lives by knowing their identity. Which is interesting because we live in a new culture of "share everything online" vs old the hacker ethos of "don't use your real name online". The attackers have become stronger and the defenders weaker.

I wonder if this could be used to unmask satoshi. I remember a piece about applying stylometry to satoshi's writing, but they just compared him to the usual list of suspects (finney, back, etc.)

I remember this! it got me pretty good. I have a bad habit of generating alts because I forget the password.

It makes me wonder if we could use non-instruct LLMs to slightly alter the wording of text while keeping the meaning the same meaning. Perhaps by using perplexity or some other metric. I don't know, maybe you compare the distance of the "meaning" vectors.

You might also want to have some "style" vector associated with each pseudonym. For example, I might want it to produce british english under a certain pseudonym, and simulate an ESL speaker under another.

Essentially, you would want some way of re-styling text. The basic way to do this would be to run the same sylometry tools the hunter uses and manually make synonym word/phrase subsitutions to lower your similarity.

It's a cat and mouse game, but I think the mouse eventually wins. Consider a program that translates your english writing programmaticially into a low-entropy symbolic form and then translates them back to english in a procedural manner. Basicially you design an intermediary language that cannot contain style. It would be boring to read but it would remove all the style.

Eh, I've got 200mb/s fiber for cheap. It's pretty good and definitely bottlenecked by crowded wifi and upstream sources moreso than the ISP. Ethernet helps somewhat.

At the same time, I do kind of want more bandwidth just so I can download massive files like model weights quickly, host a web service out of my own house, seed torrents, etc. What might cryptocurrency look like if typical residential internet speeds were measured in gb/s? Perhaps bitcoin might be capable of more than 7 tps!

But to be fair, I am a nobody.

Probably not that many lives, maybe like a handful of hikers every year I would guess? I think what attracts hikers in the first place is the danger, and the idea that they're exploring an area that is "outside of civilization"

Say what you will, but it's staying above cloud providers.

I feel like I already have internet access pretty much everywhere with cell towers, and even then if I went to the middle of alaska or montana I could already get sattelite internet before starlink with hugesnet which is fine as long as you're not gaming or something.

But at the same time I think the low-earth-orbit is pretty nice in terms of latentcy, it's a pretty innovative approach.

I just don't get the idea behind AI datacenter sattelites and moving all this non-comms equipment up in space.

Do ISPs have to comply with the 1st ammendment? My understanding was that they have some sort of common carrier law but net neutrality did not hold up.

Why not both? Three cheers for escape artists!

I have epyc 9654 ES and a 7900 XTX. I was running the numbers, and even if I maxxed out the ram to like 12x32 gig sticks, it would cost me thousands more and I could only run GLM-5.2 at a couple tokens per second at q3. So this project is very promising because it suggests I could get pretty high speed and this CPU/motherboard combination suggests I have a lot of pci bandwidth that is unused.

I think another route might be looking at holding an even larger chunk of model weights in ram, and taking advantage of RAM<->GPU bandwidth, perhaps using a PCIe 5 GPU. This was my first thought since I have dedicated GPU.

If you are using Laptop, you're looking at shared memory between the iGPU and CPU. I've also tried that route, but I have always been skeptical of killing flash with too many reads, it essentially uses SSD like it's a consumable item.

I'm going to benchmark this right now with what I have and I'll get back to you on github.

I also just edited my comment with more ideas in the beginning, sorry

I am curious if it's possible to adjust this to use more RAM, as i've got a machine with 64GB RAM and 24GB VRAM. Or perhaps I could run Gemma/Qwen on the GPU and have GLM-5.2 delegate smaller tasks to it. It might take some retraining of GLM-5.2

I'm also curious if you can speed this up by using many disks in parallel to increase bandwidth.

>SSD Wear Warning

> Cold starts are heavy on random reads (~11 GB/token). Reads themselves are safe, but the OS page cache can generate writes. Heavy use may accelerate wear on cheaper SSDs. Use with caution and monitor your drive health.

Hmm, maybe a safe way to do this would be to make a separate partition for the model weights, and set them to read-only? Not sure how the page cache works, if it's like per partition or per disk. If it's per disk, maybe you could have a read-only data.iso formatted as a partition and mount it as a disk?

its open-weight. I think you can find a host for GLM-5.2 in the USA

GLM-5.2 is not as good as Opus, it's better. I can abliterate GLM-5.2 and have it work on projects that Opus refuses.

the latency on this can't be good

This site is pretty awesome

Your comment is merely a comment and not an argument

The problem is that your system can be changed at any time such that you can have the information.

I would say that they are snake oil because of that. Data breaches occur more often than rootkits because most developers see that this path adding easily-removable encryption does nothing in the long run.

I think the point is that you can apply it to any shameful-enough aspect of the libel/parody.

There is actually an older method for countering steganography and adversarial image generation attacks: https://en.wikipedia.org/wiki/Gaussian_blur

perhaps doomerism is just the logical reaction to our real-world conditions right now

>Why would China care about deflating the US AI bubble?

To weaken the stature of the USA on the global stage relative to themselves. Perhaps decrease US investment in AI and slow creation of some general AI superweapon I suppose.

Because the goal is to show that cheap chinese AI can compete with expensive USA AI, it's nessisarially a low-cost attack relative to the "damage" it could create.

>Why do we think there is a bubble for sure in 2025/2026?

Well that's the position that these chinese firms are trying to convince us of, and they can convince us by undercutting proprietary models in price/performance/openness.

In other words, we can be sure there is a bubble to the extent that open-weight models can successfully demonstrate that there is no moat.

>Why doesn't China also worry about their own AI bubble inside the country?

Because they haven't bet the farm on AI like the USA has.

Your right to 3d print has been taken away, but not your ability.

I think it is a reasonable term to use in the context of adversarial AI development. Does that mean I am getting paid by the CIA or something?

No, but you need some sort of blockchain to enforce uniqueness in petname systems due to zooko's trilemma.

If you're okay with another side of the triangle, where all of the identities are keypairs, then you don't need a petname system or any name system. the name system is an optional convenience layer on top of the web of trust.

Good question, this made me think.

You get a reward for being a mining node, not just any node. Even then, do miners have much incentive to share blocks, other than the ones they mine?

I think the incentive is mutual for most nodes (aside from the mining aspect). People will set up a node to accept transactions in an automated manner, or to have higher confidence in the state of their accounts.

It's like being on the floor of the stock market. People participating want to be where all the information is (for their own benefit), and there is incentive to bring others in and share information (because it increases the amount of information you have).

I suppose you could be a "selfish" node. The bitcoin-equivalent of someone who leeches and never seeds. But the advantage is low relative to the amount of money moving around. Most people don't care about the bandwidth of running a bitcoin node, they care about latency. Unlike bittorrent, there isn't a de-facto finished version of the file being synced: it's a constantly-updating list that everyone wants to have the latest version of. I can't find the words, but this seems to be the fundamental difference.

except for your domain registrar and your PKI certificate authority

I think you're right that transaction fees are a key problem. It's ultimately a bandwidth problem. You're bidding for the limited vbytes, and the bidding price only increases with traffic, kicking poor users out.

I think the key thing to recognise with petname systems is that there doesn't need to be this sort of "top-level consensus" as opposed to ecash systems.

You can have two instances of namecoin, say Namecoin1 or Namecoin2. You can just have different domains like alice.nmc1 and bob.nmc2 and have them interoperate properly. You can just keep forking blockchain-based petname systems to overcome the bandwidth/fee problem.

What this means is that Namecoin1 full nodes don't need to synchronize all the domain names on Namecoin2 and vice-versa. Similar to TLDs on DNS. We can imagine that there might be different petname TLDs for different global regions, and they might be merge-mined.

This isn't true for money applications like bitcoin or eth, because by forking BTC or ETH or something, you are creating new coins.

Perhaps some sort of namecoin or ENS-like petname system with multisig or some type of scripting that enables different recovery methods.

For example, you could set your petname up so it can be controlled by a single keypair, which can be overridden after a certain time by a ring signature based on keypairs held by friends, family, peers, and trusted computing devices you leave in a safe deposit box.

Or maybe you could trust your identity with some centralized entity, but only as part of a 2-of-3 multisig with yourself and another trusted entity.

Basicially namecoin with bitcoin-like scripting controls.