Luckily I spend all my crypto profits on new hardware so there isn’t anything for anyone to steal 
2 Likes
Never heard of this defensive strategy 
2 Likes
im sure its one Khadas like 
2 Likes
Did you try to run some OpenCL bench/miner in the Vim3’s GPU? Was working flawlessly?
Intriguing! Personally, I myself envision a day when a CPU-based coin, switching algorithms constantly, allows ARM-based mobile devices to mine it - this way there is no longer any miner-concentration-risk like with Bitcoin ASICs.
This is very likely the future! But the mining has to be energy efficient enough that your mobile device could mine it without draining the battery…if there is some industry hardware support…this could go a long long way.
Maybe HTC would eventually develop a phone that could do this…then ultimately, the only risk becomes “developer concentration risk”…which maybe in 10-20 years could be solved with an A.I-based self-adapting coin…
So what does the A.I. “earn or gain” from helping us? Well lets say it’s objective would be to “exist” or “be mineable” on many devices as possible - spreading its code…like animals do…so the most successful A.I coin would be one that could spread widely. It would itself discourage miners from specialising in it, by varying the algorithm - since miner concentration reduces incentive for new miners to join, reducing the coin’s ability to self-propagate.
Also, the A.I coin could evolve it’s own monetary policy…and same as before…the coin with the most suitable policy would benefit the most users…and hence have the largest user base. The coin could “fork itself” and its children would have different mutations…okay I am rambling.
3 Likes
There are coins that are trying to target IoT and phones about already - like Uplexa (https://uplexa.com/) but they certainly drain the battery - well depending on how many cores you use. An interesting problem to solve is what stops someone coming along and concentrating hashing power (with multiple phones and or ARM based SBC) that ultimately makes the phones, not unprofitable, but with such small profits per device that no one bothers to mine with them.
AI / big data may well be the solution to that by detecting that concentrated hash power and somehow limiting it on the network. But it would be an arms race as if it was profitable enough that people wanted to mine on a single phone then the incentive to concentrate hashing power would be immense! Really interesting ideas - got me thinking - thank you!
2 Likes
I can think of 2 solutions:
-
Reward per device is limited. This could mean that in the future people would not think of mining for financial incentive, their devices would just do it as a background task to support the network.
-
One miner is limited to one account per one device. If you had 10 phones, you’d need 10 unique accounts to mine.
This all assumes that mining switches from a “financial incentive model”, to a “monetary policy model”. Meaning people no longer mine for profit (since the profit is so tiny), rather they mine because they like the way the coin is run (inflationary, deflationary, micro-lending, etc).
And the A.I’s incentive is to propagate itself over as many devices as possible, or maybe phrased more accurately - over as many unique user accounts (tied to some biometrics), as possible. 
2 Likes
If a person was going to have 10 Accounts, I guarantee they would swear on the person whoever brought such an idea of “one account one device”
2 Likes
perhaps its coming from a capitalist culture but i dont think you would get mass adoption unless you got some kind of tangible reward from the process though - its part of where the whole crypto space is - people doing for idealistic reasons of decentralisation and attacking those that dont really care about the philosophy but want to use the technology / make profit from it. Will be really interesting to see how all this pans out!
Your AI is stating to sound more like a virus!!
1 Like
it may be Capitalist culture but communist culture has many tricks up its sleeve
Yes I see where you’re coming from. At the same time, could it be that the reward is that our financial system is no longer controlled by a centralized group of people / entities who may not always have everyone’s best interests at heart?
The concept in my head is that it runs as a background task, yes virus like, that provides the essential service of maintaining a ledger for every person on the planet. And just like a virus, I anticipate there will be a few dominant variants.
Just as the human body has thousands of guest viruses inhabiting it all the time, maybe in the future, phones would have “good guest viruses” too. The only risk of course, is some evil fella creates a “super virus” that is unnaturally successful, and infects everyone’s phones, killing the “good virus”.
So until such time that A.I is both good and smarter than humans…world could become better. But if humans continue to be emotional, bounded rationality, and smarter than A.I, we will continue to have turmoil in our lives.
This is all just speculation of course.
3 Likes
This is becoming like wannacry all over again, are you speculating the same thing @tsangyoujun ?
or I guess its like a botnet virus
1 Like
Could you, please, share the opencl settings for xmrig? I tried several intensities and worksizes, but xmirg does not seem to work.
The Ratio is for 16MB of RAM, I’m not sure if this is the memory allocated by the drb file but any more just hurts performance intensity 8 or 4, worksize 16 or 32, you can use a work size of 8 as well just increase the intensity but 8 or 4 seem best, I would have thought we could have gotten more intensity going with the G52 since has an additional compute unit but doesn’t seem to workout but does give more hash now
Thanks wll1rah,
I was able to run cn/r, cn-pico and cn-heavy. The other algos, especially all the rx…, end with OOM killer (not enough RAM), even though CPU miner runs well.
The fastest settings are:
"intensity": 64,
"worksize": 16,
And I’m getting these hashrates:
cn/r 3.1H/s
cn-heavy 3.8H/s
cn-pico 43.9H/s
Though, it is not much compared to the hashrates of the CPU part:
cn/r 22.8H/s
cn-heavy 32.2H/s
cn-pico 1409.3H/sYour correct, But what I’ve been doing is mining UPX which has a really small scratchpad size of 128KB so it can fit in the CPU’s Cache memory freeing up the rest of the system to use the GPU, for something like the other CN algo’s. Also your not really out of memory the share size with the system ram is the problem and I’ve got no idea how to adjust that with the O.S./ kernel and then what the limits are I think being able to have a larger pool of memory would help get the hashrate higher but this is a really low powered GPU to begin with, so I’d not expect great results. It’s more of how it gets utilized and something like chukwa would be great but no miner that works with the Mali/ Intel GPU’s
Yes it is, this is my old discord handle and my github repo, but that’s most instructables
1 Like
Ah - thought i recognised the handle! Neednt have banged on about my miner when youve seen it here already!
It’s okay, It’s good to brag. I could have said something before, but no need, others in the UPX discord might not have seen it.