Topic: [ANN][MOTO] Motocoin - page 48. (Read 178256 times)

Here is a (very rough) fit of work requirement measurements so far.




The network is simulated with "half difficulty" challenges, where the coin is half of the normal distance from the rider.  We can make a pretty safe assumption that this will require approximately half of the amount of work at any given TT as on the real network rules.

The horizontal axis shows TT as the number of frames required below a 60 second maximum. (So easiest TTs are at the left of the graph and the right edge of the graph is at 15000 frames).

The blue points are the work samples, measured along the vertical in "replay work units" required to solve one block, each one of which is equal to  the average number of frames in all replays.  Each block had either a specific TT interval or a random TT and combinations of human miners, bot miners, and "cyborg" miners were given opportunity to compete.

The red line is a logistic regression fit of all data points, yellow is a fit of only the "interesting" data points past 7500 frame TTs, where the curve is more discernible.

This gives us a very rough approximation of work requirement.  It would be nice to take more work measurements at lower TTs, but as you can see from the estimated curve this quickly becomes a very slow process.

However, we have a bit of a conundrum anyway.  The newest bots on the network are using some improved heuristics, which may or may not significantly outperform this work curve estimate - and we can't really know for sure without having their bots to measure it.  We may have to either intermittently re-measure this curve against the best known solvers at any given time, or otherwise we may need to use an even more conservative estimate and find some way to do something along the lines of what DeepCrypto had suggested in having a slow relaxation of the work requirements.

In the immediate term, I think we should just go ahead with an approximation until we have a better option on the table.  I think no-one would argue that this wouldn't at least be a significant improvement over the curve in use now!

Excellent, thanks!

See my latest commit: https://github.com/motocoin-dev/motocoin/commit/00fdb3e0e1c09c576d9632df0ba7c38f58576bbb

True, but I'd like to do as much as possible to avoid any inconveniences to users.  I agree it is unlikely that miners wouldn't adopt this particular patch, but if it were to happen it could create some confusion and strife with users on a "wrong" fork, that's all.

@WilliamLie2 Not sure if I've asked this before, but could you do me a small favor and provide the algorithm used to generate the g_sin curve LUT.  It would be very helpful for a few things.

TIA!

i say make it a sidescroller and add shooting targets while driving part of gameplay to increase difficulty.
it will be more enjoyable on a human level...

So what if they will reject it? Do you think that they will perform attacks on our new hardfork or what? I don't see any reason why we should care so much about how many miners will accept it, it is important that community and services accepting motocoins will accept it (in our case the only such service is C-CEX which wouldn't be a problem). If miners will want to mine coins that worth something they will need to accept it.

I think we're pretty much already at conclusions about what needs to be done.  William had more or less already proposed the solution to the second warp before he even posted about the variation on the original attack itself.  We already have what "is supposed to be" the solution in the BTC/LTC ancestry code, it is just using a work definition that (although arguably valid by itself) contradicts the difficulty adjustment's definition of work under some cases.  In some sense we can not definitively declare a correct work definition, but we can use something much closer to expectations based on measurements.

MOTO had a paper that was actually pretty good relative to what most put out and call "whitepaper."  I think it was pulled for revisions that never ended up happening, but psychocoin and I had discussed reviewing and re-working it some.  I have a copy of it if you'd like it.


Sure, but this has to take a back seat to actually fixing the issues at hand!  That being said, here's a quick rundown on recent events:

A) Will has brought up a secondary attack vector that is similar to the first, but needs a more detailed solution than my patches offer.  Still no set fork point, but we will set one as soon as we have both solutions ready.

B) Coinader was likely a scam, and appears to have absconded with funds.  MOTO was one of the few coins that actually had liquidity there.
  (IMO they are likely the ones who most recently dumped.  Market is again showing almost unprecedented buoyancy for an alt.)

C) New or changed bot miner appears to be dominating production with a very slick new bot.  Their map filtering appears to be almost inexplicably fast for the quality of maps produced.  Other bot miners appear to be either tuning and failing to compete, or essentially giving up entirely on mining for the moment.  I haven't personally checked in great detail, but there's apparently no evidence that this dominant miner is behaving any differently from the earlier miners.  They appear to be processing transactions altruistically as well.

D) Will has proposed increasing the minimum work requirement in my patches.  This both adds security to the network and may slow bots more significantly even when the network is not repaying a work deficit.  I agree that this would be a good change, but want to be sure that the other miners would be willing to adopt it.

IMO, the more the merrier.  At the very least we would like at least one more person to authenticate gitian builds with us.

I'm really enjoying reading about the Moto-specific algo oddities. Any chance we could soon come to conclusions and create a solid white paper outlining the solutions and the shape of the game and premise of the coin itself that we are ultimately going for? And a simple road map along with it. Summarizing the lengthy posts of the last pages that most people probably skipped over into bite-size chunks would surely spread some buzz around, re-inspiring investor confidence regarding the "new" Motocoin, perhaps even finally coming to peace with our new robot overlords. Would be nice, right?

In the original attack, a work deficit is created and left unpaid on the synchronization between wall-clock time and game time.  In the second attack no work deficit is ever created, the attacking chain is incorrectly credited with more cumulative work on the synchronization of game time and blockchain time, when less is actually done.  In a sense, the two attacks are inverses of each-other, a warp in one direction on one side of the synchronization, or a warp in the opposite direction on the other side.  (Some might be thinking that there might be a third kind of warp to "complete the triangle" on warps against the three-way synchronization.  Of course there is and this would just be the traditional time warp against wall-clock and blockchain sync!)

Their solutions are also independent.  Closing the unpaid debt on in-game time was relatively trivial, just require the debt to be made up by the network.  Closing the credit of surplus work estimate proves a bit more complicated since it requires a better definition of the work curve characterization than just the 1:1 inverse scale of TT currently in use.


Sure.  I'm entirely open to the idea of requiring a larger minimum of classical work.  There would seem to be a lot of advantages to it!


I worry that it may either be a short lived advantage, or too many bot miners may reject the network change.  In any case I feel that it is at least worth floating the concept out there and giving it a try!  Any thoughts from anyone else on this?

Fascinating posts, guys. Would it be worth it to try and recruit another capable dev or two to get things moving?

I don't see how it is different from the original attack. You can prevent decreasing of target time by generating specially crafted replays and use the fact that client cannot correctly estimate difficulty of the blockchain. Both attacks are based on these facts so I don't see any difference between them.

I think that we can make it necessary to solve classical proof-of-work challenge to generate map. HunterMinerCrafter's patch already does that but it is very easy there. If it would require say 0.1 second to generate a map then I think it will immidietly give advantage to humans and will force bot developers to use smarter algorithms.

THIS.  Although it was reasonably self evident as to what needed to be done to secure against DeepCrypto's original version of the attack, the secondary vector that Will has pointed out is significantly less straightforward.

I have been implementing a large simulation to approximate an appropriate work curve.  Although I initially thought this would be an "easy" task, it seems that the work curve is highly sensitive to the specifics of the approach taken by miners to finding solutions.  As such, it is not enough to just run minim1ner's bot and measure the work requirements as originally planned - the slightest change in the search parameters to the bot results in a different work curve.  I'm now simulating an approximated combination of different types/parameters of bots, as well as taking measurements of the work curve under both human mined play, and "cyborg mining" play, in order to best fit an average curve.

Further complicating (and slowing) the matter is the fact that at low difficulty runs, the curve is basically indiscernible.  This means that the simulations to measure work must not only include a composite of different approaches to mining, but must be run against higher difficulty challenges, which means that finding solutions takes longer.  Since many solutions must be found for each of many different TTs along the curve, this becomes very time consuming.  Runing each "sweep" from 60 seconds down to 10 seconds on TT takes about 6+ hours (and this is with pretty much all of my mining resources thrown behind the simulation) and we will need quite a few of each sweep for a reliable average.  Just the "run time" on approximating the curve will be several more days, actually making the code changes will be a few more days, and planning/deploying the necessary fork will be a few more days beyond that.  In general, these sorts of developments can't happen overnight.  In this specific case, it can't even happen over a weekend, or likely over a week or two.  The number crunching requirement in this case is just downright big.  (If anyone out there would like to contribute a few hundred CPUs to speed this work measurement process up, please contact me!)

I will likely post my first "rough" analysis of the data so far in the next day or two, along with a first stab at a polynomial to fit the curve, but we will need several more days of simulation running time before we can have any real confidence in the approximation.


Yes, specifically we are the first chain to synchronize three clocks, and this is kind of a big deal.  I've come to term this a "second order block-chain" and although it is very cool that it is even possible, it puts some unique constraints on the mining process.  Perhaps things would have gone more smoothly if the proof-of-play challenge had been done in a way that only required a "first order" chain (I suspect so!) but this probably would have made for a very different game, and it is "too late now" anyway.

(EDIT: I feel that the importance of this 3-way time-stamping synchronization mechanism can't be overstated.  If we can successfully have a second-order chain we can inductively assume any order of chain.  If we can synchronize 3 "inter-related logs" than we could synchronize tens, or hundreds, or hundreds of thousands.  This mechanism might allow for things like a generic meta-chain structured to include a synchronization against any arbitrary chain for which it had a definition of the work curve characteristic. We can go way beyond just "wall clock time" "main chain time" "in game time."   The possibilities around such structures quickly even become a little mind boggling.  The practical applications directly to MOTO are awesome, like the possibility for tournaments that not only escrow MOTO and/or BTC for entry, but could offer claim of a future BTC network coin-base as prize.  It would take a lot of work to get there, and some improvements to related "p2pool like" technology, but such things become possible.

The fact that MOTO is the first mover in both the proof-of-play and "second-order chain" attempts is a big part of why I've jumped into the project in the first place, and why I'm still personally buy-and-hold on every coin I can get.  (EDIT2: Despite the particular and peculiar growing pains that the first attempted combination of the two resulted in!)   When people wake up to what MOTO itself will be able to do with this novel sort of chain, and what other "conscious" networks will be able to draw on from the work that MOTO is already pioneering.......)


I'm trying my hardest.  Will has also been quite a bit more "above board" than most devs I've observed as well, even despite some of the early drama.  If nothing else, at least we not only don't make vague and bogus claims, but actually go out of our way to do the opposite and offer sound reasoning and justification behind our opinions and corresponding decisions.  We've also been active in reaching out to other coin communities to try to point out others' vauge, bogus, unreasonable or unjustified claims when those claims overlap with the goals of MOTO.

Also, I think that just the fact that we have two independent devs, who are even "at odds" with one another and have entirely different backgrounds, perspectives, and motivations (and that we're seeking more involvement!) should be a clear sign that this is not your average alt.  I'm quite sure that any of these now-thousands of scam coins would not ever have taken the step that Will did in bringing an independent developer into the project.  (I suspect most would not even entertain the notion if proposed to them, and many would even go so far as to moderate the suggestion out of existence to keep the idea out of the heads of others.)


I think any "human mined" coin both always inevitably will be and always should be a bot-coding competition.  While this might evoke a negative knee-jerk reaction by many, it rationally shouldn't.  This is no different from the fact that bitcoin has always been and will always be a "more efficient SHA coding competition" and this is a good thing for bitcoin, not a deficiency.  The only real unique aspect here is that it could never be feasible for SHA to be done by hand, let alone made enjoyable to do so.  With MOTO human mining can (and will) be made enjoyable.  (Again.)  It is important to keep separated these two distinct concerns, and many on both sides of the debate seem all too eager to blur the lines.

In the long run, I see "cyborg mining" as likely being the eventual outcome.  Just "running the bot" will under-perform relative to human-guided bots, at least for a very long while.  Since bot guidance is itself a sort of game (this has become particularly clear to me since I've starting toying with Foldit after my conversations in the other thread with Gatra about proof-of-play for scientific games) it may very well be that nobody even has interest in "pure" human mining in the end, anyway.  It will likely be that MOTO, HUC, and any other "human mined" coin in the future, will always be destined to operate much like Foldit, where success depends on a healthy balance of AI and human intuition working together on a single challenge.  HUC has already been like this since about 2 weeks after launch.


It is increasingly surprising how many people share this delusion.  These days it is becoming quite apparent how people are not aware of what is happening with contemporary AI research.  The world is on the fast track to sentient agents, and it is looking like people outside of the industry are going to get blind-sided, specifically because they are stuck in this particular and peculiar delusional state.


I mostly agree, and it already seems to be doing just this.

Where I disagree a little is in that we don't entirely have to take MOTO for just what it is, as it is.  I think more important (at this juncture) than attracting devs to the mining side is the goal of attracting devs to the dev side!  There is a lot that we can do in terms of making the coin more interesting and enjoyable, but as we've concluded earlier in the thread this is a task that is bigger than the combined efforts of only one official dev (Will) and one community volunteer dev (myself) can handle.  Between everyone in the community we have a nice laundry list of ideas that are good and could be implemented (and an even longer list of other ideas, heh) but we're already "under staffed" just on addressing the low level infrastructural work to be done!  We need more volunteer contribution of developer time here than on the mining side, for the moment.  The mining side is largely "taking care of itself" naturally through competition.  (We already have a tx processing network that is quite a bit faster than most, without (so far) any sacrifice in security.  This is a hugely understated accomplishment!)


Unfortunately, there will always be those who desire the network to function differently than how it does.  This seems to be universal among crypto offerings, as well, and not particular to MOTO.

What I find most surprising is the common combination of a desire for a network to behave differently and the lack of any initiative or effort to make it do so.  If you really believe that a specific network should behave differently then put your money (in the form of energy) where your mouth is and hash on a new fork.  A block chain is a weighted range-voting process, so if you want to see some change all you need do is propose and champion the change, and then go "get out the vote!"  If your change is good and the other participants agree that it is good, your change will be enacted.  If your change is no good you'll be hashing alone and unable to interact with the network.

It seems many people don't understand this most fundamental aspect of these currency networks, and it saddens me.  Much progress that should have been made by now in the crypto space has likely been delayed simply by many refusals to actually champion their cause and cast their votes.

If you aren't happy with some PoW coin's miners' claim to subsidy then your only sensible and useful recourse is to bring up more hashing power and reduce their subsidy claim!

If you aren't happy with the MOTO bots' operators' claim to subsidy then your only sensible and useful recourse is to operate a bot and reduce their subsidy claim!

Complaining about bot operators without also running your own bot is a bit of "pissing into the wind."  Either turn around, or stop screaming into that wind about your pants being wet.  No other course of action accomplishes anything.


Again, I personally always try to defer to education before exclusion.  I find it best to briefly try to get across to people where their rationale has gone astray and/or what they are "missing" about the peer consensus process, as I find that in most cases people do quickly get to the "A HA moments."  If they don't, they'll usually just eventually opt to exclude themselves on their own, so explicitly excluding from the start usually accomplishes nothing that wouldn't happen naturally anyway!  

You should fact-check against network logs.  The original devs claim to have never bot mined, and there is not real indication on the network to the contrary.  I was mining, but am now only solving about 1 block per hour.  Mining production is currently dominated by what appears to be either a new participant, or some new nodes from one of the "silent" participants.  This miner has a new bot, which is something between the "second generation" bots most were using, and what I would call a third gen learner.  (I have confirmed that it is neither of the other miners who have "come forward" who are running this new bot.)  This new bot has an excellent map filter that finds very "easy" maps very quickly, and has a very rapid solver.  I've seen this new bot very rapidly solve 3-5 consecutive blocks within seconds.

Further, even without deference to the network block submission statistics your statement doesn't stand on its own.  Anyone can run minim1ner's bot and mine.  "As is" you will find about 10 blocks per day, and with my map filter changes back-ported into the bot you'll have about the same production rate that I have currently.  With some tweaks to the search parameters one could probably get this up to about 2-3 blocks per hour.


MOTO faces some unique challenges, particularly as it is the first chain (AFAIK) to attempt to synchronize three time-scales.  Where I see a major distinction from many other coins is that we are openly discussing both the issues of the coin and their resolutions, and are collaborating directly with the community (both users and miners) on ensuring a smooth solution.  We are very clearly not "pretending" on any front.


It seems like I've had to repeat this time and again, but I have neither any proposal nor interest in any "solution against bots."  The bots are necessary to ensure the security of transaction selection.  The goal is not, and will not be, to find any solution against bots, only to balance the game to allow human mining to resume in some form.  It is unlikely that there even is a "solution" to stop the bots in any case, so it wouldn't make sense to spin our wheels on it anyway.

It does appear that coinader has closed and absconded with any funds.  Seems like someone went through an awful lot of work to steal a few MOTO.  (It was the only coin that ever had anything on their books, really!)

I hope nobody lost too much.  I think I'm out about $100 in combined btc and moto value.  All well.

I am attempting to contact them through a couple of means, but I am not holding my breath.  We'll see about removing their posts if they do not come back with an explanation quickly.

2-3 days is not really long.  This secondary warp issue Will brought up is a big one, but we're closing in on a solution.  (More details to follow.)


Will has given me the same access rights that he has on all project resources, so the project can have continuity independent of him if necessary.  As other developers come on board we can also further replicate that access, so that the project can not end with the loss of any subset of developers.  Now we just need more developers! XD


No, since this is not something anyone is working on yet.  While I am all for the idea of a non-fixed map size, actually implementing such a change would be somewhat difficult.  For the time being, I think that we are both going to be focused on the more critical matter of closing both warp attack vectors.


You are not wrong, but this is only a stop-gap solution to restoring human mining. I'm sure we would all like to see solutions that can work to restore human mining perpetually over time.

To me it looks like solving the issues is complicated, there is no precedent for this in crypto world, Motocoin is the first to attempt pure proof-of-play implementation. HunterMinerCrafter has been more transparent about his motives than 99% of the developers I've seen around. This coin is far from over. Yes, it will probably always be a bot-coding competition. Anyone who thinks computers are ever gonna be long-term inferior to humans in solving of any deterministic problems is, well, delusional. There is no permanent solution, so you gotta take Moto for what it is - and improve its current premise to attract more talented AI engineers... Instead of crying over what it cannot realistically ever be. But you seem to have zero understanding about how cryptocurrencies and blockchains work, so it's probably futile to try and explain the rationale.

Well MOTO is dying anyways, for 5 dollar its not an expensive lesson. There's more reports of Coinader being scam, don't use it.
MOTO is failing simply because the devs are one of the few or may be the only ones that can mine this coin with bots, in simple words it now belongs in catagory 98% premine-dump scheme, pretending to try and solve the issues. Another scenario is they might have had a solution against bots ready from the get go but simply wait to implement it to create another lifecycle. Be carefull to buy into this.

Ouch! That has got to suck so much.
We better remove it. Dev?