Topic: Bitcoin adoption slowing; Coinbase + Bitpay is enough to make Bitcoin a fiat - page 12. (Read 67180 times)

Fact: iron was a precious metal

Fact: commodities are on a 2000 year decline in price

Read the thread for the facts.

Knowledge is growing in relative value. Facts don't lie.

A grain of sand isn't worthless, but for all practical purposes it is. Meaning tangible things can become relatively devalued to the point where they are essentially worthless. Through enhanced knowledge tangible things can become nearly limitless (i.e. robots build robots who build tangible things), while knowledge is not fungible. Not every person has every great insight.

After reading this thread and a few of the links provided by Anonymint it really struck me how important not only anonymity is but also the fact that the next generation crypto should be cpu only plus zero transaction fees. The point of mining being slightly -EV but still a highly valuable feature since it allows everyone to convert fiat to crypto without any third party involved was a heureka moment for me. It's just an amazingly simple but yet extremely important feature. Not only would it solve the exchange/trust problem of bitcoin but it would also greatly facilitate the adoption of the masses.

Siegfried, I understand it isn't personal. Some (probably many or most?) don't want a proliferation of crypto-coin supply, because they think this will diminish the value of crypto-currency.

I don't have time for more debates. Besides none of those guys are interested in debating with me. And much better we not debate and each attempt to prove our points in the market pronto.

---

Vociferation:

Currency is not a heirloom, hold gold (or silver) for that purpose when you need it. Remember money generates no income and will always fall behind production in return of investment. Warren Buffet is correct on that point.

I think that is an incorrect understanding of macroeconomics and money. They ignore Gresham's Law. The purpose of currency was never as a very strict store-of-value, rather to facilitate trade by removing the mismatch friction between barter-in-kind.

I explained in great detail, that reasonable rates of decentralized debasement of the money supply is not deleterious for the productive people. Decentralized debasement actually can dilute the lazy large capitalists who intend to capture everything by riding stored capital, and productively transfer that capital in a competitive proof-of-work to the more productive knowledge capital producers. This is done at a very gradual rate so that money still retains a store-of-value quality as well. Bitcoin's debasement rate has been very high in recent years and still going at 11% per annum, yet the price appreciated due to influx of adoption.

There is not going to be dozens of very successful crapaltcoins, i.e. coin supply will not spiral out-of-control. The market will end up with one, two, or maybe three top tier coins.

Make sure you pick the best horse to ride. Study carefully the attributes of your horse.

The only way to fund security of proof-of-work without the Tragedy of Commons of transaction fees, is perpetual debasement. What is the fair amount to pay for security of our money? 2.5%? 5%? (per annum)

Consider we pay it to ourselves with a cpu-only coin, because we mine on a level playing field.

Consider all the losses we incur due to not having a secure, decentralized money. It far outstrips even 10% per annum. Think of all the BS government is doing all propped up by central banking. All the unnecesary wars we've funded by not having decentralized, well secured, untaxable money. It is economically impossible for a government or King to make a war if they can't tax to fund it.

USA Government has grown from 10% of the GDP before 1913 creation of central banking, to probably north of 70% when all regulatory aspects are considered. Europe is even worse.

Before central banking in the 1800s depressions were frequent and quickly over with. USA had a depression in 1919, but it was over in 2 years, because there was no central bank to backstop and prevent the defaults. Since we've had central banking, debt has grown to what the IMF admits is a 200 year high at $223 trillion in total global debt, with a $quadrillion ($1000 trillion) of derivatives to prop it up, and another $quadrillion of unfunded social liabilities from developed nation governments.

It is collective insanity. I think 2 - 5% paid to ourselves is a very excellent win-win design decision. And it coincides with gold's debasement rate. And for those who say the supply of gold is finite, I already won that debate against bitfreak!.

---

I came to this thread to ask your opinion about that, but you just said it. Nothing personal, but I hope you are wrong. I would like to see you carry on a debate about this "sidechain" concept with someone very knowledgeable.

So much going on. Adam Back and some Bitcoin core developers are throwing their hat into the "let's improve Bitcoin" arena:

http://letstalkbitcoin.com/blockchain-2-0-let-a-thousand-chains-blossom/

Problem with this idea is it can't innovate mining because it is merge-mined with Bitcoin, which is the crux of the problem with Bitcoin.

Have fun Adam. Sorry you are not going to be able to save Bitcoin, nor restrict crypto-coins to 21 million. Your goals are flawed.

Me thinks the Winklevoss twins either don't understand or ignore technological details. I don't think Bitcoin can scale to micro transactions and still fund mining with transaction fees because I argued upthread that transaction fees are a Tragedy of the Commons. The transaction fees can only be controlled well via centralized control, but then how do we stop that central control from being captured and then milking the economy for parasitic rent.

http://www.investopedia.com/articles/active-trading/040814/winklevoss-bitcoin-payment-system-worth-400-billion.asp

I wish the "No" (likely BTC) voters would post in the thread and tell us why.

Haven't seen any reasonable logic yet to support why the OP is wrong. Maybe there is some?

Perhaps multiplexing. Slow memory latency bound means we don't need very fast transient response.


What is the justification for memory interface saturated at 32 cores?
 

Tilera cores are 400mW, so 12.8W for 32 cores. That might be load dependent though. Are ARM or Atom CPUs more efficient?

DDR4 main memory is coming next year bringing claimed 40% improved energy efficiency.


Hopefully the community can arrange to provide to you access to a TILE-Gx now or in near future.


I didn't guess it based on your photos.   Explains why you are very knowledge though. Experience.

"I still look like a teenager" (yeah right, we keep such delusions as we get older).

The asic could have thousands of cores. But a chip can only have a few 100 IO pins, of which each memory interface needs at least 32.
So you'd be limited to running at most, say, 16 Cuckoo instances, each only needing 32 cores to saturate their memory interface.

 

Memory chips run at about 1W each, I think. That's 1W per Gbit. A cuckoo instance (512MB) would take 4W, which I think is already
more than the 32 super-simple cores would need. So I disagree with you on where the power mostly goes.


I would love to bench Cuckoo Cycle on either a Xeon Phi, a dual 15-core Xeon, or a TILE-Gx processor.
All are rather hard to find in the wild though (I had acces to a Phi, but it kept crashing on my code:()

PS: I corrected the hashing overhead in the new algorithm from 16x to 4x since it only hashes edges which survive
the filtering in each round.

I measured the fraction of runtime spent NOT accessing main memory (mostly hashing) as 32.5 %
So it's still latency dominated, just not as extremely as before (where it was 5% vs 95%)


I'm fast approaching 50 myself:-(

If the main per hardware $ cost is memory, then the more threads the lower the cost. Since an ASIC can discard unneeded generality, it should support more threads.

Since memory latency bound is not consuming much electricity, the main electrical load comes from the computation of the hashes, thus ASICs due to their dedicated circuits should be per Watt more efficient than manycore CPUs.

The simpler the CPU cores, e.g. AMD's business model, the less per Watt advantage an ASIC would have. And the more and the simpler the CPU cores the less per hardware $ advantage. However, I believe Intel's more complex CPU cores model is winning on the desktop which is what matters now because last time I checked even mobile devices didn't have more than 4 CPU cores.

For mobile computing, the requirement is power efficiency. Thus the simpler and more CPU cores model may win. Thus your Cuckoo Cycle may become the correct choice in the future for making sure proof-of-work on a mainstream device is not at a disadvantage to economies-of-scale, ASICs, nor GPUs. I urge you to measure and determine if there is an asymptotic limit to the parallel threads, so we can determine at what point your hash becomes the superior choice. Perhaps you can only do this with a simulation, unless you can get your hands on a Tilera CPU but which currently top out at 72 cores.

P.S. to the extent your work is used in crypto-currency, I believe you will be monetarily rewarded. You put considerable effort into Cuckoo Cycle hash, including providing test results in your whitepaper. I believe it is essential to maximize breadth of development contribution in crypto-currency. As of today, per my citations upthread the USA may have spent as much as $4 trillion unaccounted black budget on covert goals. Hopefully we can get $4 trillion spent on open source goals and hopefully apportioned via decentralized decision making. I'm pushing 50 next year, blind in one eye, moderately (not debilitatingly) suffering from slowly progressive systemic illness (unknown if curable) which causes Chronic Fatigue Syndrome every nth day or so (n ranging from roughly 3 to 14), so to any extent at all that I am a defacto leader, I would be working to share development effort and offload it to the younger guys as much as makes sense. I urge leaders to do the same.

And pleeeassseee don't use Windows 8 as it calls home and is porously backdoored. If you must use Windows, please Windows XP SP3 (or if you must then Windows Vista but I trust it less). Much better Linux.

For Windows XP, after installation go to the Control Panel, select Administrative Tools, then Services, then right-click Automatic Updates, select Stop from popup menu. Right-click Auto Updates, select Properties from popup menu. Then choose Startup type = Disabled. Also click the Windows Firewall icon at the right side of the bottom taskbar, and make sure Automatic Updates is really off.

Also click Start menu at left side of the bottom taskbar of your computer, then Run.... Then type "msconfig" and press Enter. Click the Startup tab, then remove any programs you don't want to run automatically at startup (such as that pesky Google Update which spies on you).

You can still obtain a copy of Windows XP SP3 and also there is a crack just search for "Anti_WPA__XP___Vista_.exe" that enables you to avoid registering it with Microsoft's website and such that you don't need a unique serial number. I am not advocating piracy, just informing you what exists and hopefully you only apply that patch to a legitimate copy of Windows that you purchased.

Spread this information please, but please don't quote my pseudonym.

I highly recommend automatic disk imaging restore software such as Deep Freeze or Baseline Shield which resets your computer to virgin on every reboot (other than any rootkit virus that might have got in during your session). Does anyone know of similar software for Linux? (Edit: yes look up Ofris)

It parallellizes well upto 20 threads at least. Haven't tested the new version on anything with more than 20 cores.
I expect memory will be saturated before reaching 100 threads.
A cuckoo ASIC will be a vastly simplifed manycore cpu. It will do away with the need for virtual memory support, page tables, and big
caches (which the current implementation uses just to reduce page faults). It will attach to a bunch of memory chips, but just load
and write 32 or 64 bit words rather than whole cachelines. It will have a siphash instruction, and no floating point support. It will
be tiny and cheap, a small fraction of the price of the memory it needs. But it won't be all that much faster or cheaper than some similar
construction of commodity parts, considering how manycore cpus will become popular soon.


Cuckoo Hashing is great for where simplicity is key and you don't need to achieve high loads.
For better locality and high loads I think you can't beat Hopscotch hashing...

And this is one of the reasons why I say I don't want to trust some public ceremony to generate global cryptographic setup parameters for Zerocash, because we can't be sure the computer is doing what our open source code says it should be doing. And for Zerocash we have no way to measure whether we succeeded, because the anonymity is so extreme there is no way to measure the money supply. We will always be running on faith only. Btw, I am thinking about ways that Zerocoin technology could be modified (but I can't do everything simultaneously and have to stay focused).

There is even research showing a computer can be controlled remotely and even when is it turned off. Google search will help you find the information.

Ken Thompson's Trusting Trust, pointed out it is even possible to covertly infect the compiler into producing an output program which has malware embedded, even though the original open source code did not.

It is much more difficult (impossible probably) for the NSA to attack every one of our computers individually. Thus I prefer technology where the creation of cryptographic trapdoors (i.e. our private key) is done billions of times on individual computers distributed throughout the world, and not depending on the security of global parameters.

Thus I think as long as we don't use global cryptographic parameters (i.e. not Zerocash) and each of us has a dedicated computer for issuing our anonymous transactions and often reformats the hard disk and reinstalls the operating system (and reflash the BIOS on your motherboard), or for simplification buy a new computer every year or several months, then I don't think realistically the NSA can't target all those computers. For micro payments where we aren't paranoid, the masses can simply use their existing computer.

Hopefully we the people will gain the upper hand economically and then we will open source the CPU designs and fabs, to get the covertness out of out foundational computation layers.

The fundamental disadvantage is the generative model is money requires spontaneous (ephemeral) trade-weighted matching backing of money. Thus we lose timing degrees-of-freedom which was the point of money in the first place. Generalizing Gresham's Law to its more generative essence tells us that money with the least friction will displace all other forms of money. I am referring to the autonomy of money.

In summary, the value of autonomy and decentralization is it scales by N x N (where there are N users) in the network according to Reed's and Metcalfe's Laws.

Ah I see Dave mentions using a 'bitvector':

http://da-data.blogspot.com/2014/03/a-public-review-of-cuckoo-cycle.html

Good to see confirmation that my analytical capabilities are still in touch with reality.


If you've raised the computational load relative to memory latency bound, then you've made it more computationally costly, but still that doesn't necessarily defend against an ASIC.

The most important question is it still parallelizable (multiple threads on same memory) even if slightly sublinear? If yes, then some economy-of-scale can be applied to outperform personal computer CPUs per Watt at least, if not also per hardware $. Even if no longer parallelizable, this might still be true (as you know ASICs can be more power efficient than CPUs because they are dedicated circuits)

Btw, your design helped me refine my design, not because they are at all similar, but because by analyzing yours I was able to more coherently conceptualize mine in a unified analytical framework. Will open source my white paper soon so you can analyze mine.

As you've noted, the significant advantage of your design is the asymmetrical execution time of calculating versus verification. Yours is the fastest verifying hash (relative to finding a block or pool share) I've seen proposed for proof-of-work. That probably means your hash has applications where verification cost is critical. Perhaps it should be proposed for Bitmessage.

Also thanks for introducing me to Cuckoo hashing. Might be able to apply that to my programming work.

AnonyMint, what do you make of this?

http://www.infowars.com/intel-ceo-refuses-to-answer-questions-on-whether-nsa-can-access-processors/


"Last Summer, shortly after the Snowden leaks began to escalate, Steve Blank, recognized as one of Silicon Valleys leading experts, noted that he firmly believed NSA has backdoor access into Intel and AMD chips. He noted how the leaks highlighted how backdoor “hacking” is the NSA’s go to technique because it is much easier than trying to crack encryption."

I have some ideas not yet written down on that the key on scaling is a protocol for the way pools can cooperate to contain both the orphan rate and the transaction scaling. I believe this is the way to reduce transaction delay to less than a minute (perhaps lower than 30 seconds) and also scale volume by orders-of-magnitude. Remember one of my assertions is we must have a mechanism to keep pools small enough size that they can't easily collude to do evil.

Just to get some concept of how many micro payments might occur over the web, 4 billion youtube videos are watched per day. That is roughly 50,000 per second.

Once you put the decentralized micro payment paradigm out there to compete with the advertizing (i.e. spam!) funded website model, I believe it will skyrocket beyond all fathomable calculations due to the NxN scaling of networks as I explained upthread in the discussion of Reed's (Metcalfe's) Law.

alxs, I am not ignoring you, just a wait a bit for your answer while I catch up on a few unstated matters.

It was in fact subverted in exactly that way!
See https://bitcointalk.org/index.php?topic=405483.40

The latest algorithm has an edge-trimming filter that uses a bitvector
to record which edges cannot be part of a cycle.
As such there are no "empty memory locations" at this stage.
It gets by with 21x times less memory, and 4x more hashing,
at roughly the same speed as before.

Maybe you can stop calling it a memory-only coin...

Why not allow the user create currency ex nihilo like how Central Banks do it.  That way the currency is worthless and people would only create it temporarily in order to engage in trade

Lets say I've got pies that I wanna trade for socks.  But the sock guy don't need pies.  But lets say there is an electronic exchange w order books w people with things to trade for something else.  The exchange would match up all the trades necessary for all transactions to come full circle.  Each trader generates some amount of currency (credit for their goods) then the exchange looks for trades to match up the orders.  Then the orders are executed all at once and each user ends up with a coupon for what they need and gives a coupon redeemable for exchange of hard goods.  In essence each user is issuing currency but only redeemable in assets that they can back.  So I issue my pie note convertible for 1 pie.  Sock guy issues sock note convertible to 1 pair of socks.  The exchange has to find all the orders so that there is no excess currency in the system.  All balance sheets close to zero after the orders are complete.

The currency itself is ephemeral and not worth anything.  No value ever so only the backing asset has intrinsic value.

I'm just toying w the idea conceptually as a mental exercise.  Probably in practice it would require too much energy compared to centrally controlled currency