Introducing [P]ermanence
Motto: Recall [V]ictory ?
[P]ermanence should be the world's first AAIi. It'd consist of :
- An ircii bot for communicating with The Lordship.
- A box acquision module.iii
- A cryptographic module.iv
- Av Bitcoin node in a secure configuration, with wallets.
The functionality contemplated here is that upon receiving a $boxdeploy command from an acceptable signature, [P]ermanence responds within a reasonable timeframe with login credentials, delivered much in the manner deedbot presently provides OTPs.
The OS environment in which [P]ermanence itself runs, as well as the environment in which it creates accounts is tbd, with the mention that absent a homomorphic encryption solution, spreading the data in some sort of encrypted database model across physical machines would be probably the best solution. In any case the first goal of [P]ermanence hardening is to make an attacker unable to obtain usable data from subverting n/2+1 machines at any given moment ; and for [P]ermanence to self-destruct if it ever loses n/2-1 machines at any point. Whether these two conditions can be satisfied by protocol rather than by promise is the most important determinant of the utility of the whole shebang.
Ultimately, [P]ermanence must be capable of disabling, completely and for any desired interval up to permanently, any computing infrastructure, such as operated by any individual or group however defined, up to the total of all installed computing infrastructure operated by anyone anywhere. This "any" knows absolutely no bounds, if in the future someone somewhere makes a computing anything outside of the Internet, [P]ermanence makes a nuclear sub and glasses the spot.
Obviously, throwing all this together out of and on top of all the leaky "modern" software is a taller order than making a proper Bitcoin. Nevertheless, [P]ermanence is necessary and unavoidable, so might as well start thinking about it. The goal of overpowering the entire world may well seem far fetched, but the strict reality of the matter is that, much as in the case of the very Bitcoin that spawned the Republic, we actually have all the tools required. It's not a matter of great leaps of faith or imagination being required (except perhaps as pertains to homomorphisms), but merely a matter of patiently, diligently and correctly applying things we already mostly understand and mostly have.vi
To arms! And may we all live to see a world which harbors no intelligent life outside the WoT.
———- Autonomous Artificial Intelligence. [↩]
- To be supplanted with gossipd later on. [↩]
- At first, this can consist of a simple list of IPs and root credentials. Eventually to be expanded to include an actual mechanism to rent dedicated servers from companies ; to score the providers according to its own, (self-revised) rules, and to find them in the first place (which later task will include running a complete web crawler, exactly like Google - only better). Even more eventually to supplement this with active 0-day discovery and exploitation. [↩]
- To replace SSH entirely. To be based on the eventual winner of TMSR's cipher competition. [↩]
- The grammatical singular implies nothing about actual implementation. [↩]
- Mention of the LOIC, because. [↩]
Sunday, 1 May 2016
Sounds like an awesome idea, but it seems to get very close to that tricky limitation of Gödel (in the end you'll have a program that looks at other programs and says stuff about them, but there's no way to verify that it doesn't lie, other than verifying the results by hand -- IMHO this is an important aspect of "Autonomous AI" to discuss; autonomous, but how do you know it doesn't start fucking you over at some point? as per Vinge's "singularity", Skynet and other past discussions on the matter). But who knows, it might become somewhat feasible in a decade or two.
AFAIK at the moment there are no (publicly known) ways to find zero-days without human intervention. Even fancy-schmancy approaches such as symbolic execution (for binaries, although it should also work for web stuffs or even more general system-level searching) are utterly unreliable and computationally expensive. A few of the engines out there sort of manage to do it (with a big emphasis on "sort of") for specific scenarios, but they're far from production-grade stuff.
Sunday, 1 May 2016
You test that it doesn't fuck you over in exactly the same way you test any other slave : you give it a task and see what it produces. Inasmuch as it doesn't know the scoring criteria aforehand, "perfect deception" reduces to perfect submission, which is what was sought in the first place - and also both why I say that government may exist for as long as I don't notice its existence, and why MPEx proxies work the way they do.
The point is not for all of this to work with 0 human intervention, nor anything close to it. Let me underscore this matter, because the tendency for the mind to immediately go that way smacks me of fauxism : that P should work without any human intervention is neither relevant, nor even an interesting matter.
The guarantee P is expected to make is exactly that it will always choose death over dishonor, and absolutely nothing else. The eventual endgame contemplated in the penultimate paragraph flows from here as a matter of necessity, and requires no further priors.
Sunday, 1 May 2016
PS. Why aren't you on #trilema and in the WoT etc ?
Tuesday, 3 May 2016
Ah, then I misunderstood what "autonomous" means? It being a slave makes it pretty much non-autonomous (at least with respect to its owner), but otherwise yeah, I guess it's easy to fall into the trap of thinking of a HAL 9000-like scenario.
Gotta ruminate more on it, but I kinda see. The n/2+1 guarantee kind of reminds me of the bounded-retrieval model (there are quite a few papers on this, I've read very little on the subject so can't say whether it has any relevance to P), but I guess the bigger issue is how to detect infiltration.
It's possible to detect physical tampering (already mentioned on IRC, I remember seeing that Freescale/NXP have hardware with self-erasing memory, not sure how much I would trust that though), but kinda impossible to do this using the average software stack, not if you can't have a fine-grained control over what processes run in the system anyway (and even so, the kernel itself remains attackable). But I'm getting ahead of myself.
Working towards that when not academitarding. I could lurk around, but I'd rather have something to contribute if expected to WoT; I expect simple PGP signature to be necessary but not sufficient. :)
Tuesday, 3 May 2016
> It being a slave makes it pretty much non-autonomous
You're making concept salad. For your own personal enlightment, lemme direct you to the standard Romanian History Manual, 9th grade, with Michael on the cover. The Romanian Principalities acquiring autonomy is therein pointedly not equivalent to their acquiring independence. And there's also churchly autocephaly and a host of other situations. They all mean meanings and are distinguishable and distinct, this naive autonomous = independent = sovereign = sugar, spice and everything nice doth not hold.
More generally : slaves are routinely autonomous, at least mine are. Autonomy denotes an ability to act independently, that, the ability, no more.
> but I guess the bigger issue is how to detect infiltration.
This is the biggest issue, requiring as the expression goes some cheating in wolf form. Nevertheless - there is one fundamental property of the universe that promises this should be possible, which is the H principle.
But yes, probably new hardware will have to be baked.
There are, of course, softer approaches also - after all, how does any organisation, from a secret service to a living organism, detect infiltration ? Yes, such wonders as delusional parasitosis as well as getting used to your own fleas exist, but nevertheless. What'd we be if not belabouring under the delusional misapprehension that we can improve upon nature ?
> but I'd rather have something to contribute if expected to WoT
That's more a lordship thing. Do not waste time, under any excuse, such as for instance that you're 13 and intercourse is painful still, or that you're saving up to "offer the best [for anglophones : Happiness.]" or whatever else. Fugit irreparabile.
Tuesday, 3 May 2016
Point taken.
The issues raised are worth contemplating though. One of them is the slave switching allegiance (see Russo-Turkish War/Romanian War of Independence, very probably related to infiltration problem discussed above; I've avoided discussing the more precise meaning of "subverted" in terms of computing, but in the end it's unavoidable; I guess the system needs to be fully deterministic and controllable-by-owner, from a control theoretical point of view). Or just susceptibility to simpler scenarios such as "bribe guard dog with food".
The other is, if we use biological organisms as models, (in addition to parasites et al.) the possibility of cancers or autoimmune disease occuring in the system; I guess the latter would actually be a feature (kill self if stopped functioning properly), but the former probably not. This isn't so much the case of the system turning „signularity” (which I agree is unicornism) as it is of "automobile running into tree due to stack overflow bug in brake firmware" or something similar. But alas, this is an issue with most software (and some hardware).
Alright. Will do once I get a cloak. Is the tradition still to do registration in-channel, or is it ok to chat privately with deedbot?
Tuesday, 3 May 2016
There's more approaches to the subversion problem. For one thing : the system could be incomprehensible. Consider the [very little discussed] situation of faux "artificial intelligence" as produced by the Bayesian hyerophants. For the other thing : the system could be self-healing. Consider the very simple case of CRC encoded messages resisting subversion by the transmission medium. I'm sure there's more things.
The automobile will forever run into trees ; the question is to make every occurence unsurprising.
You can register either way, really.
Wednesday, 8 November 2017
How did I earn 11000 bitcoins in 55 days: http://11000bitcoin.xyz/wppage/how-to-earn-11000-bitcoins-in-55-days/
Wednesday, 8 November 2017
Here, I helped : http://btcbase.org/log/2017-11-08#1734358
Once you get the 11999 bitcoinz together early next year drop us a line we can go yacht partying or something.