China’s Bitmain dominates Bitcoin mining, wants to cash in on AI(qz.com)
qz.com
China’s Bitmain dominates Bitcoin mining, wants to cash in on AI
https://qz.com/1053799/chinas-bitmain-dominates-bitcoin-mining-now-it-wants-to-cash-in-on-artificial-intelligence/
58 comments
Is it that hard to believe? If you make the best X in your market, it's gonna be the most popular. 70% market share isn't so abnormal.
If you're digging for gold and there are 5 different shovels you can buy, but 1 is clearly the best, is it that hard to believe that 70% of all shovels in use will be that best one?
If you're digging for gold and there are 5 different shovels you can buy, but 1 is clearly the best, is it that hard to believe that 70% of all shovels in use will be that best one?
Can't they now take over the market and generate bitcoin until everything crashes?
In this case, no. They produce ~70% of the hardware used by other miners. If they kept the hardware inhouse it would be another story all together.
If history is an indication, selling the "shovels" to others during this digital gold rush is more lucrative for Bitmain than mining would be.
If history is an indication, selling the "shovels" to others during this digital gold rush is more lucrative for Bitmain than mining would be.
Also, even if you controlled 99.99% of the hashing power, you still couldn't "generate bitcoin until everything crashes." You can't give yourself an arbitrary sum.
with 51%, you just give yourself the entire sum of network rewards not an arbitrary sum.
At which point the exchange value of Bitcoin plummets, so you turn your $3.5 million in daily block rewards into nothing.
Not necessarily. A smart actor with massive hash power (enough to stage a 50%+ attack) would gain the most by using a Sybil variant to subtlety take control. There is no reason to be overt if you can conceal the connection between "independent" mining pools.
We might disagree on what a 51% attack means.
A single actor controlling a majority of hash power isn't a problem. Double-spends are a problem. If anyone double-spends, it will be noticed because there will be a permanent record of it in the form of orphaned blocks, and bitcoin value will decrease because Bitcoin transactions will have become less trustworthy.
Thus, even in the case where a single actor has 51% hash power, that actor has a strong financial incentive not to conduct a 51% attack.
The Sybil-style independent mining pools you describe aren't really hiding what people would be looking for, which is a coordinated preference for the fork that causes the double-spend. (Plus miners would have to accept the relayed second transaction into the mempool to begin with, which today nobody does. RBF might have changed this, but even then you should still be able to reason about whether anyone is even contemplating allowing a double-spend.)
I'm not sure if I'm talking past you -- am I correct that we disagreed on what a 51% attack looks like?
A single actor controlling a majority of hash power isn't a problem. Double-spends are a problem. If anyone double-spends, it will be noticed because there will be a permanent record of it in the form of orphaned blocks, and bitcoin value will decrease because Bitcoin transactions will have become less trustworthy.
Thus, even in the case where a single actor has 51% hash power, that actor has a strong financial incentive not to conduct a 51% attack.
The Sybil-style independent mining pools you describe aren't really hiding what people would be looking for, which is a coordinated preference for the fork that causes the double-spend. (Plus miners would have to accept the relayed second transaction into the mempool to begin with, which today nobody does. RBF might have changed this, but even then you should still be able to reason about whether anyone is even contemplating allowing a double-spend.)
I'm not sure if I'm talking past you -- am I correct that we disagreed on what a 51% attack looks like?
Double spends attacks aren't the only issue when you have access to a large portion of the network.
Malicious pools can prioritize their transactions and manipulate activity in the network to grant them as many block rewards as they want.
Malicious pools can prioritize their transactions and manipulate activity in the network to grant them as many block rewards as they want.
No one would really notice, as there's no way to prove it.
You just steer as many block rewards as you want towards your accounts, maybe give %10 to normal users.
You just steer as many block rewards as you want towards your accounts, maybe give %10 to normal users.
That's an interesting claim. A miner already receives the coinbase and transaction fees for the blocks it completes, and it can't change the contents of blocks it doesn't complete, so it can't steal those blocks' rewards. Those are the only two groups of rewards, so no, the miner can't use 51% hashpower to get any part of the other 49% of block rewards.
"Giving ten percent to normal users" is also a novel concept. I have never seen coinbase awarded to any address but ones having a relationship to the successful miner. Are you under the impression that normal users usually get some part of block rewards? They don't. (I'm not sure what a "normal user" is, either.)
Nor can a miner change the destination of funds in transactions included in the completed block. That would be isomorphic to calculating private ECDSA keys from addresses, which is not possible with Earth technology.
Finally, the miner can't simply change the rules (for example, including both parts of a double-spend in a single block, or substituting a different signature for others' transactions). That would cause a hard fork, and every client on the planet would choose the side that followed the rules.
According to the design and implementation of Bitcoin, your statements cannot be anything but wrong.
On an unrelated matter, why do you put the percent sign in front of numbers? I have never seen that before in any language, forum, or writing style. Not only is it stylistically uncommon, but it's incorrect inasmuch as the % symbol is short for "per hundred." It wouldn't make grammatical sense to say "give per hundred ten to normal users," which means the prefix symbol doesn't make sense, either. Where are you from, if you don't mind my asking?
"Giving ten percent to normal users" is also a novel concept. I have never seen coinbase awarded to any address but ones having a relationship to the successful miner. Are you under the impression that normal users usually get some part of block rewards? They don't. (I'm not sure what a "normal user" is, either.)
Nor can a miner change the destination of funds in transactions included in the completed block. That would be isomorphic to calculating private ECDSA keys from addresses, which is not possible with Earth technology.
Finally, the miner can't simply change the rules (for example, including both parts of a double-spend in a single block, or substituting a different signature for others' transactions). That would cause a hard fork, and every client on the planet would choose the side that followed the rules.
According to the design and implementation of Bitcoin, your statements cannot be anything but wrong.
On an unrelated matter, why do you put the percent sign in front of numbers? I have never seen that before in any language, forum, or writing style. Not only is it stylistically uncommon, but it's incorrect inasmuch as the % symbol is short for "per hundred." It wouldn't make grammatical sense to say "give per hundred ten to normal users," which means the prefix symbol doesn't make sense, either. Where are you from, if you don't mind my asking?
You are only considering what is possible with a solo miner, the attack described involves multiple miner nodes colluding in a pool to suppress, delay, and orphan transactions and blocks.
I'm from Moon.
I'm from Moon.
If you're either hoarding or spending 99.99% of all new bitcoins, that's going to cause issues in the market. But the other thing you can do with majority hashpower is double-spend your coins, then fraudulently verify the transaction as if it were valid.
They still could if they had put a backdoor in all their mining hardware.
Even if they could, why would they destroy an absolutely MASSIVE income stream, it goes completely counter to their own economic incentives.
The story talked about how they survived the 2013 crash or something.
Also IIRC, digital currency is base on how much someone is willing to pay for it. If you own 90% of bitcoin and the rest own 10%, then wouldn't it be useless? I mean I'm not force to use bit coin I can just use real money if it's so hard to get bitcoin. There aren't that many people who are deeply invested in bit coin with a 90-10. The whole thing is kinda base on greed...
Also IIRC, digital currency is base on how much someone is willing to pay for it. If you own 90% of bitcoin and the rest own 10%, then wouldn't it be useless? I mean I'm not force to use bit coin I can just use real money if it's so hard to get bitcoin. There aren't that many people who are deeply invested in bit coin with a 90-10. The whole thing is kinda base on greed...
They could mine it and still sell it, so it wouldn't be 90-10
As others have said they sell a lot of miners but their pool is the largest at 25%. If they did push 50%, even getting there, let alone the potential manipulations, could only be for negative outcomes. McDonalds could destroy the beef industry... But why?
No. They don't actually control that much hash rate. Either the article is confusing the sales of their mining units to other miners or is outright wrong. If they really controlled that much of the hash rate, they would have activated emergent consensus to increase the max block size. They would have gotten around core and all their toxic censorship and bitcoin cash would never have needed to create a fork to get reasonable throughput.
It's not wrong, look at the wording:
> effectively providing 70% of all the processing power on the network.
They are providing the processing power by selling mining chips. They are not operating those mining chips, so they don't control the processing power. They only provide it.
> effectively providing 70% of all the processing power on the network.
They are providing the processing power by selling mining chips. They are not operating those mining chips, so they don't control the processing power. They only provide it.
Which means that they can't do a 51% attack, which was what the comment above me was implying.
They can't generate infinite bitcoin units anyway but if they controlled 70% of the mining power they could conceivably execute double spends, although even this would be unlikely.
In any event the post above mine asked a yes or no question and the answer is no.
The article implies that providing mining cpus is much more of a problem than it is since they don't actually control that much mining hash power.
They can't generate infinite bitcoin units anyway but if they controlled 70% of the mining power they could conceivably execute double spends, although even this would be unlikely.
In any event the post above mine asked a yes or no question and the answer is no.
The article implies that providing mining cpus is much more of a problem than it is since they don't actually control that much mining hash power.
I wonder what's the effect of the Big Firewall on connecting to the btc network
There was a block propagation study. Not much. The latency itself is the issue, but propagation time should/could be amortized.
https://bitcoin.stackexchange.com/questions/42378/how-releva...
https://bitcoin.stackexchange.com/questions/42378/how-releva...
Zipf's Law
The issue with ASIC is that when there is new AI algorithms that gets developed, they may not be even able to use those. General purpose hardware will be able to. You are stuck and need to upgrade.
Depends what the ASIC does. The bulk of deep learning is still matrix multiplication and scalar operations on matrices. GPUs are, as the name implies, general-purpose. They aren't as optimized as they could be for optimal performance on matrix ops. This is part of what Google is capitalizing on with the TPU.
Further speedups can be gained by going with lower precision ops (eg: float16). A chip specialized for deep learning could support f16 only, which would allow it to have more compute power using less area and less power.
So, in short, there are ways to make ASICs that are more efficient for deep learning while still giving the ASICs flexibility. ASIC doesn't mean there is no programmability.
Further speedups can be gained by going with lower precision ops (eg: float16). A chip specialized for deep learning could support f16 only, which would allow it to have more compute power using less area and less power.
So, in short, there are ways to make ASICs that are more efficient for deep learning while still giving the ASICs flexibility. ASIC doesn't mean there is no programmability.
Just last week I was saying that the AI gold rush needed its own version of "Bitboys Oy!" to add comic relief and occasionally good ideas. After all, you can't spell Bitmain without A I.
https://en.m.wikipedia.org/wiki/BitBoys
https://en.m.wikipedia.org/wiki/BitBoys
Aren't machine learning workloads more variable than crypto mining? How would ASIC users cope with that?
The big advances in neural networks stem from the use of GPUs that are themselves less flexible than CPUs, but allow extremely good parallelisation of (some) linear algebra (mostly matrix multiplication). This seems like a logical next step, and it's probably similar to those "tensor processing units" of the newer Tesla chips and whatever Google's custom hardware was called.
Google indeed calls them Tensor Processing Units and NVIDIA used the name as well iirc, thus the birth of the TPU.
Is that a spoiler for "The Three-Body Problem" in the second paragraph? Thanks, QZ!...
Kinda.
It's not a major plot point that's being given away
It's more like "there's an Alien in Nostromo"
It's not a major plot point that's being given away
It's more like "there's an Alien in Nostromo"
Errrr... I think it's a bigger give away than that. I came to the three body knowing almost nothing of what it was about... the information in the article would have removed much of the surprise.
Yes, (without giving too much) but you know about it before knowing how it's called
Had never heard of it, but it looks really interesting:
https://www.amazon.com/Three-Body-Problem-Cixin-Liu/dp/07653...
https://www.amazon.com/Three-Body-Problem-Cixin-Liu/dp/07653...
I know everyone says it's good, etc. I just want to reinforce the point. It is really really good. Definitely one of the best scifi I have read.
The story just gets bigger and bigger as the books progress as well as the imagined tech and physics.
I know it's cliche but it definitely caused me to think slightly differently when looking up at a night sky.
Highly recommended. Sequel is quite good as well.
Really, really good. I would rank it up in top scifi I've ever read.
Great book, somewhat unusual sci fi.
It's a minor plot point, and it's not even described accurately.
How great would it be if AI advances ended up controlled by an authoritarian Chinese government built off of the profits of libertarian Bitcoin users?
It's kind of fascinating just how dependent the value of cryptos and interest in them has been dependent upon China.
It's kind of fascinating just how dependent the value of cryptos and interest in them has been dependent upon China.
>How great would it be if AI advances ended up controlled by an authoritarian Chinese government built off of the profits of libertarian Bitcoin users?
Bitmain is a private company. And even if you can describe bitcoin as a libertarian technology, I wouldnt think that most of Bitmain's profits come from libertarians based on who uses their hardware.
>It's kind of fascinating just how dependent the value of cryptos and interest in them has been dependent upon China.
Nothing special in the biggest part of the market dominating the price the most, really.
Bitmain is a private company. And even if you can describe bitcoin as a libertarian technology, I wouldnt think that most of Bitmain's profits come from libertarians based on who uses their hardware.
>It's kind of fascinating just how dependent the value of cryptos and interest in them has been dependent upon China.
Nothing special in the biggest part of the market dominating the price the most, really.
I think most libertarian see it as libertarian because it is not issued by any government and so it isn't affected by gov when they change interest rate.
It is purely dictated by the market of supply and demand.
This is just what I think why Libertarian see it as libertarian; I'm not one in any real sense nor do I subscribe to any of Ayn Rand's reasoning.
It is purely dictated by the market of supply and demand.
This is just what I think why Libertarian see it as libertarian; I'm not one in any real sense nor do I subscribe to any of Ayn Rand's reasoning.
are you also fascinated that many internet applications are dependent on data centers in America? I think it is more fascinating that I can build a crypto application on top of infrastructure built in another continent.
Good.
Nvidia stuff 1080ti cost too much for a graduated student.
AMD is working on their own stuff supposely later this quarter we'll get tensorflow and other support on their card.
Google's TPU is not for consumer, only for cloud.
If bitmain can sell ASIC for consumer with reasonable price and support tensorflow, pytorch, etc... then it'll be good for the deep learning community.
Nvidia stuff 1080ti cost too much for a graduated student.
AMD is working on their own stuff supposely later this quarter we'll get tensorflow and other support on their card.
Google's TPU is not for consumer, only for cloud.
If bitmain can sell ASIC for consumer with reasonable price and support tensorflow, pytorch, etc... then it'll be good for the deep learning community.
I suspect they'd be designing for datacenters rather than consumers. That's essentially what they already do with bitcoin rigs too. It's a rented mining cloud.
>rented mining cloud
Would you like to rephrase this to something more accurate?
Would you like to rephrase this to something more accurate?
What part do you consider inaccurate? You pay for hashes/s in their hardware pool. The phrasing seems pretty sufficient.
e.g. https://pool.viabtc.com/contract/
e.g. https://pool.viabtc.com/contract/
Hashnest is a similar service run by bitmain.
> Hashnest is a cloud hashing platform. This means you can participate in bitcoin mining without maintaining the hardware yourself.
> Hashnest is a cloud hashing platform. This means you can participate in bitcoin mining without maintaining the hardware yourself.
There are still a lot of software side of things we as a community can do in accelerating compute speed - for example, add OpenCL support for popular AI framework/libraries, which will leverage existing hardware and non-Nivida GPUs' power.
[1]: https://github.com/tensorflow/tensorflow/issues/22
[1]: https://github.com/tensorflow/tensorflow/issues/22
A lot of the frameworks(keras, Theanos) uses Cuda close to the metal, so they would need to write alternative interfaces and the to be supported by these libraries
Incredible story, thanks for sharing!
It's incredible that one company accounts for the vast majority of the entire Bitcoin network's hash rate.