teravor

    > immediate redemption with the issuer once it's been spent/accepted. This is a terrible model for physical cash

not when everything is now online.

also the obvious way such cash would work is that "redemption" is just the new minting of coin. the central authority will mint a new coin by blindly signing your secret after you "destroy" the spent coin by giving them the unblinded signature.

it would depend on the government and the situation. I suspect most don't realize this is possible. a small government can generate immediate demand for their currency doing this.

when you unblind a blind signature all the signer knows is that it's a signature they signed at some point, they know nothing (true zero knowledge) about when or where they signed it among all the other signatures.

combine with a central authority server network and make it blind signatures and you have untraceable e-cash.

you would probably need special engine optimizations to really make use of it, and VR hasn't been exactly an attractive option for AAA games.

I suspect the answer is actually very simple, when you compare this device to one that is made deliberately dumber there are far too few actual differences to make it make sense.

usually, the implication of ZKP is that you buy coupons and claim them without attribution. in this coupon scenario the ZKP can just be a blind signature scheme.

however signal has an obscene fondness for TEEs (secure enclaves) so they may actually be doing something stupid here which will require trust beyond the ZKP.

    > inference is already being subsidized

idk where this comes from but it's laughably false.

the only place where actual subsidization (below cost) might be happening are the subscriptions. even that is unlikely because to be truly below cost you either need to offer below cost of electricity which isn't happening, have potential API users using multiple subscriptions or have opportunity cost loss due to saturation.

with Wayland it's generally not any safer.

if you use Wayland's security context to prohibit privileged protocols such as arbitrary clipboard access then an application will either not be able to grab clipboard content until you focus on it or the attempt will be noticeable as it spawns a short lived window in an attempt to grab focus.

when an LLM is trained backpropagation reaches into the entirety of the key value table. or rather, all the layers (not just the tokens) which come before the current tick.

langauge is only produced by the final layer every tick. and every layer can only interact with other layers on the same level.

the mechanics of LLMs and the restriction in how we can train them makes it appear as though all we are doing is forcing language onto them but once RL gets involved all bets are off regarding what's happening inside them (it's quite possible that a static corpus alone is sufficient for all the bets being off).

    > They're still autocomplete

LLMs are simulations and the tokens are the ticks.

if we transcribe your brain into a simulation and give it a tickrate, you will be just autocomplete too. the argument could be made that you are autocomplete anyway - neural dynamics.

the autocomplete reduction is vacuous.

    > Many graduate students (I know) are having a crisis if any of their research worth it? If AI can (or will) do everything, what's the point of doing experiments and all? This will eventually deter a whole generation of curious minded students from research.

should they not be deterred?

we stumbled into a way of brute forcing intelligence with gradient descent.

it's only faster in a medium, which technically isn't even light but rather electron interactions.

apparently they weren't even properly configured until almost touching down.

if such a method gains traction then so will the effort to bypass it. either by stealing private keys or just projecting light onto the sensor.

some VMMs just straight up translate vsock from the guest to UDS on the host.

    >  “you are saying I am doing X bad thing? What about Y bad thing you’re doing, let’s discuss that instead”.

no one is generally expecting or even asking for "let’s discuss that instead". what they are doing is trying to point out that your lack of of care about Y implies that you never actually cared about X in the first place but rather you only care about me doing X.

often times X is actually bad but human nature is such that no one actually cares about it while wanting to appear to when locked down on it. for this reason arguing about X directly is bad optics.

whataboustism is basically an effort to draw attention to the selective enforcement of norms, laws, morality, whatever. hypocrisy is implied.

false security. if there aren't LLM's monitoring patches with disassembler MCPs ready, there will be soon.

well that's why setting it up is hard, because you would want to do it in a way that what they want doesn't matter.

ideally, an auction and the vendor or a government can bid against malicious actors (which can also be a government). hard to set up though.

RCE inside sandbox, so requires chaining with another 0day.

there have been people who took existing LLM's and conducted an algoritmic search to find out which group of layers they can duplicate in order to improve performance, and it worked.

the only real way to have a decentralized domain system is something like https://github.com/pubky/pkarr

but you lose the ability to have short domains.

also until such a time that pkarr is widely adopted, you are better off using .onion domains anyway. it becomes a question of requiring custom DNS client vs. Tor browser.

both approaches use a DHT.

    > Researchers have confirmed Graphite attacks through WhatsApp and iMessage, although the complete method remains secret.

unsurprising that it's closed source software which is vulnerable like this. LLMs can now decompile binaries very efficiently so it's not even security by obscurity anymore.

you really have to screw up to not carefully trace the path of incoming messages to ensure they are processed safely.

it has been amusing watching you try to pretend that you knew what survivorship bias was. first you assumed it had something to do with death. then you invoked the plane anecdote.

now you link the wikipedia without realizing that over 90% of it has nothing to do with planes.

    > Before attempting to attribute fraud, one must first determine that it exists.

this is an assertion not an argument. I suppose there is no further point to this discussion.

survivorship bias is a statistical error, why would you invoke planes? especially after saying this:

    Huh? Survival of what? Are you claiming that Democrat fraud was not found only because the perpetrators are all dead?

also, you are the one invoking domestic political factions. I'm merely wondering why a system where it's difficult to attribute fraud to anyone should be allowed to continue. it's probably more likely that a state actor is going to be committing the fraud just to saw chaos and to perhaps one day be too obvious about it.

    > because what you describe is not survivorship bias

then you should look up what survivorship bias is. the visible group in this scenario are the individuals who were caught. which says nothing about any conspiracies or even other more competent individuals. the sampling is also poor.

    > mail-in voter fraud cannot be detected, then why do people claim that mail-in voter fraud exists

isn't that obvious? if it were easy to counterfeit currency in an undetectable way for a conspiracy while stupid individuals can get caught. would you be asking this question or would you assume the currency is in a bad shape? this example may actually be too on the nose as there is reason to believe exactly this happens with respect to certain state actors while being mitigated by the relative low volume of cash.

he is implying that the focus on the republican cases (the survivors) is a red herring because the system is presumably such that an actual conspiracy cannot be detected.

I don't know anything about elections, a question worth asking is what would it look like if a conspiracy set out to exploit mail ballots? could detection only originate from a defector?

such decentralized systems would have to rely on a DHT and the domains would be public keys, somewhat defeating the purpose of a domain.

an effort to do this: https://github.com/pubky/pkarr

that means all you had to do was gate the content behind a javascript snippet with a timer.