Bitcoin Forum

What exactly does it do? worth it?

Link?

https://sophon.ai/product/sc1.html

Thanks for the link.  Very interesting indeed.  Will dig around further on this.

BR

Not for mining, it's intended for use in "Deep Learning" AI related software acceleration.

Deep Learning is it only for Games, Software or what??? ??? ???

Deep learning or machine learning is the new tech, you will be hearing more and more about this AI in the time to come, look now and you can find out more from MIT and other notable organisations who have been researching this for years...

Tensor ... nvidia GPU has tensor cores, so, does it mean it will work in tandem with GPU's ? like "AI" accelerator card or something? if it will, then it should be possible to utilize it for mining... in theory at least.

https://www.researchgate.net/profile/Konstantinos_Sgantzos/project/Implementing-A-Church-Turing-Deutsch-Principle-Machine-on-a-Blockchain/attachment/5a01b45a4cde26d9055d81cc/AS:558051050233856@1510061146342/download/Sgantzos_K_GA-HSCBB17.pdf?context=projectUpdatesLog

I just was this and I will be interested to know how this technology can be applied.

This is OLD news, they talked about this last year,

The ML algo is well known, so it can be placed on asic

What good is it ?

Self driving cars, spy cameras with ability to ID U built inside, red-light cameras that read license plates and email u a speeding ticket,

Big Brother SPYING IOT all on one chip,


All the algos, language translation, image analysis, big-data analysis, its all down now to one small common well understood algorithm

Have they delivered the chip? Is it working? Probably not, Google is also working on the same chip, everybody is

The thing to remember is that BITMAIN is a one trick pony, an the S9 is dead, nothing new is coming as nobody can get 14nm or smaller to work, so in order t stay in biz BITMAIN had to find new markets, so they play the same trick as 'mining' they promise something and hope that the can get the money up front and some day deliver,

The problem is the ML people ( mostly NSA, CIA, FBI MS FB ) google-is-nsa type spying level, they're NOT going to use chinese shit because the US-GOV wants control of the world,

Now I doubt that BITMAIN will give us the $1 chip that does ML, but certainly the people who make the rasberry-pi clone CPU can and will.

***

two most general and popular algo's are ...

    Artificial Neural Networks

    Deep Neural Networks (Convolutional or Recurrent Networks)

Both of the above can be coded in a page of PYTHON, so its easy to learn this stuff, but it takes a bunch of GPU ( 1080 class ) to 'train', but once you train the system with your data, then you can get 99.9% which is better than a human,

all of it works pretty much the same, huge matrixes are correlated and then a calculus descent function is applied to find the optimum

Y = F (X)

Say you have an 'x' it could be photo of a woman, and you want a Y that indicates on a scale 1-10 is she hot or not, ... the ML algo will generate you a F() function ( matrix ) that will take the image, and return you a number 1-10 of 'Y'.

Now how the F() is built is training, both super-vised and un-supervised learning, supervised is where I pass the algo 1M photos and I tell it 1-10 for each photo, eventually the algo figures it all out, un-supervised training is where the algo all on its own figures out what is a 1, and what is 10,

The best in developing this shit is 10% supervised, followed  by 90% un-supervised, once the algo gets general idea, it will always learn better than a human, but initial seed of 10% supervised gets it in the right direction.

***

Now set back and think in the general form of problems

Y = F ( x )

x can be anything, and Y can be what ever u want, about 8 years ago I wrote on ML that took any music 'live' (MP3) and generated 'music script' so that I could play guitar of any song I wish, the transcription sw at the time was expensive and didn't work so I rolled my own.

***

Now specific to MINING, say you have a bitcoin address and you want its PRIVATE-KEY, well there is a database on line of all training data, so you train an ML to learn the relation between all addresses and all keys, ... this algo actually exists

Lot's of interesting stuff to think about

It's from the 1950's, what is NEW is the GPU made it possible,
The entire reason CUDA exists and the nvidia devleopers library is-was so that scientists could code ML algo's.

All the science was developed in 1950's for ML, 1980's a guy discovered backwardization, its stalled in 1990's cuz CPUS were too slow, but now with GTX-1080 class machines anybody with a desktop can generate their own ML.

Just a few days ago a ORANGE grower somewhere 'rolled' is own AI-ML sorting machine for 'oranges' if I remember right, he trained the ML to recognize some 100's of defects and types, and BINGO he had a sorting machine that could recognize any orange, or any fruit for that matter better than a human,

This is why all these AG jobs will go away, pretty soon NO HUMAN will be involved in BIG-AG

The math/physics is all 1950's people, nothing new, its just the GPU in the last 5 years that made all this possible with a $1000 class cpu/gpu.

I want someone to buy it and try to put it to mine coins xD

But Can it run Claymore miner?

lol yeah I wonder what the hash rate is on this card, can someone report and confirm? ^_^

So there is not interest for us miners?

Did you read the paper I linked? It may for sure be interesting one day as a "miner". I can just not tell you if economically it is good to now get the foot in the door or not. We are looking long term here. If you have questions about the paper, feel free to ask. I am not the author, but have some understanding about it.

Deep learning is subset of Machine learning and they both are old technologies. DL become so popular because in nowadays we have a lot of data and computer power, which wasn't available in the past.

Yes you are 100% right. Since the 60ies. Non the less I am convinced we will see a new evolutionary phase with this technology.

There may be considerable interest if the guys at VectorDash http://vectordash.com/ get it off the ground.
Claims AI researchers pay at least double the best crypto mining rates to use your nVidia GPU's to run their ML training sets on your GPU's.

x1 risers are a bit of a drawback.  Maybe I'll put together a B8PLUS 8 https://octominer.com/shop/octominer_b8plus/ with a few Sophon SC PCIe cards https://sophon.ai/product/view.html?product_id=000201710181659398323r05Yqq8067F&template=sc1-detail if Bitmain will ever ship this product.

I have a significant QTY of Sophon SC1 TPUs in deployment for our research customers.
It’s worth noting that the device presents itself to the OS as a Xilinx memory controller with 16GB DDR4.
Bitmain has detailed FP16/32 INT functions available to the Metal API with a theoretical max throughout of 2TFLOPS.

If anyone is interested in getting schedule SSH access to a host with an SC-1, please feel free to PM me.