- coppsilgoldI always wondered how useful such tools are against a competent adversary. If you are a competent engineer designing malware, wouldn't you introduce a dormancy period into your malware executable and if possible only talk to C&C while the user is doing something that talks to other endpoints? Maybe even choose the communication protocol based on what the user is doing to blend in even better.
- Even if you plug the audio output into the input you would still be taking a quality loss by passing the audio through a DAC and then an ADC. Maybe if the quality of your hardware is good enough it wouldn't matter, but then you would be limited to only ripping 24 hours of audio per day...
- Correct. Note that the MTU will be further reduced and that WireGuard DIY multi-hop may be inferred.
- WireGuard limitations hurt the attempt however.
For example, multi-hop betrays the actual exit node to your ISP (or MITM) due to the port used.
- That just shows that GenAI is associated with poor quality content at this moment.
So the true opposition is to poor quality content, not GenAI.
Unfortunately for artists, actors, etc. GenAI right now is the worst it's ever going to be, and it was much worse just a year ago.
- Can you use DAITA with just gotatun (on linux) or do you require the Mullvad daemon?
- Stated preference vs. revealed preference: https://steamdb.info/app/1808500/charts/
- The parent attempted to excuse them by pointing out that the initial design was based on phone numbers. Putting aside the fact that initial design is irrelevant to present design criticism, they went out of their way to design usernames yet deliberately disallow signup without phone numbers.
> Not a very good case made since you obviously didn’t read the parent discussion.
This isn't an argument, do you have anything to back up your assertion?
- > Signal is demonstrably anti-privacy by design.
> You can reasonably disdain still having to expose a phone number to Signal, but calling it "anti-privacy by design" elides the origins of that design.
They introduced usernames without removing the requirement for phone numbers.
I rest my case.
- > have a database of all the phone numbers
They have exactly that. They rely on TPMs for "privacy" which is not serious.
> Perfect privacy would mean not sending any messages at all
Not sending messages is incompatible with secure messaging which is the subject of the discussion...
> ZKP system where you make a public record of your zero-knowledge proof sounds anti-privacy to me.
A zero-knowledge proof provably contains zero information. Even if you use a type of ZKP vulnerable to a potential CRQC it's still zero information and can never be cracked to reveal information (a CRQC could forge proofs however).
> especially when you're suggesting cryptocurrency as a solution
Would you elaborate on why cryptocurrencies are not a solution? Especially if combined with ZKPs to sever the connection between the payment and the account. When combined with ZKPs, they could even accept Paypal for donations in exchange for private accounts.
- If you intend to use SMS (phone numbers) as a resource constraint (sign up requires 'locking up' a resource that is worth at least a few cents) then at least you can offer a ZKP system where the 'consumed' phone number is not tied to an account. You could also offer to accept cryptocurrency for this function - call it a donation.
That Signal did none of those things implies that privacy was not their objective. Only secure communications was.
It's possible that the reason behind their anti-privacy stand is strategic, to discourage criminal use which could be used as a vector of attack against them. Doesn't change the fact that Signal is demonstrably anti-privacy by design.
- As VPN usage proliferates such discrimination starts hurting sites more. For example, a VPN may be left on by a user for whatever reason and when the site they visit doesn't work or makes them jump through hoops they are less likely to visit the site in the future or view it with contempt and abandon it a soon as they are made aware of an alternative.
It takes time for sites to realize the danger, especially with mobile users where fiddling with a VPN is often more hassle than its worth and its just left always on. It's often a good idea to impersonate a mobile user agent for this reason as some sites (or perhaps cloudflare?) started treating them differently. The impersonation needs to be done well (SSL and HTTP fingerprints should also match mobile).
Usually, the more expensive the VPN offering the better the reputation of their IP's. Avoid VPNs that have any kind of free tier like the plague.
- Isn't this type of "circular funding" equivalent to bartering? And why is this a problem?
Basically, the gold rush has reached a point where the shovel seller wants a stake in the operation.
I wonder if we would one day have a situation where NVIDIA no longer wants to sell chips to anyone and just use them themselves. Some special developments would have to occur to reach this point I think.
- > If everything have static address and deterministic static, you can have a exact copy on ground and debug there.
You can also have deterministic dynamic - the satellite could transmit its dynamic state (a few bits signifying which memory cells failed) and then you proceed deterministically on the ground.
- The attention mechanism is capable of computing, in my thought experiment where you can magically pluck a weights-set from a trillion-dimensional space the tokens the machine will predict will only have a tiny subset dedicated to language. We have no capability of training such a system at this time, much like we have no way of training a non-differentiable architecture.
- If you come up with a genetic algorithm scaffolding to affect both the architecture and the training algorithm, and then you instantiate it in an artificial selection environment, and you also give it trillions generations to evolve evolvability just right (as life had for billions of years) then the answer is yes, I'm certain it will and probably much sooner than we did.
Also, I think there is a very high chance that given an existing LLM architecture there exists a set of weights that would manifest a true intelligence immediately upon instantiation (with anterograde amnesia). Finding this set of weights is the problem.
- Is a brain not a token prediction machine?
Tokens in form of neural impulses go in, tokens in the form of neural impulses go out.
We would like to believe that there is something profound happening inside and we call that consciousness. Unfortunately when reading about split-brain patient experiments or agenesis of the corpus callosum cases I feel like we are all deceived, every moment of every day. I came to realization that the confabulation that is observed is just a more pronounced effect of the normal.
- > This can be bad if a particular memory cell has failed. If every variable has a fixed address, and one of those addresses goes bad, a patch can be loaded to move that address and the mission can continue.
This seems like a rather manual way to go about things for which an automated solution can be devised. Such as create special ECC memory where you also account for entire cell failure with Reed-Solomon coding or some boot process which blacklists bad cells etc.