Creating A.I. out the hashing power of ASICs (Sci-fi?)

[iohen]

Hi,

I just want to float a crazy idea that occured to me a few days ago. Don't think I am the only one around here thinking about this matter (and I am lazy searching the forum for a similar thread).

What is Bitmain developing other than ASICs? SOPHON, their deep learning AI.
Now imagine the incredibly wast amount of computing power now already connected to all the mining pools in the world being suddenly somehow switched (or waking up) forming a consious AI.

Any reactions?

[monsterer2]

Hi,

I just want to float a crazy idea that occured to me a few days ago. Don't think I am the only one around here thinking about this matter (and I am lazy searching the forum for a similar thread).

What is Bitmain developing other than ASICs? SOPHON, their deep learning AI.
Now imagine the incredibly wast amount of computing power now already connected to all the mining pools in the world being suddenly somehow switched (or waking up) forming a consious AI.

Any reactions?

Totally possible. Look for a thread on here about PoW mining using conway's game of life.

edit: here you go: https://bitcointalk.org/index.php?topic=2977765.0

[HeRetiK]

What is Bitmain developing other than ASICs? SOPHON, their deep learning AI.

So I see the Three Body Problem books are popular with Bitmain

Now imagine the incredibly wast amount of computing power now already connected to all the mining pools in the world being suddenly somehow switched (or waking up) forming a consious AI.

Totally possible. Look for a thread on here about PoW mining using conway's game of life.

edit: here you go: https://bitcointalk.org/index.php?topic=2977765.0

Running endless iterations of conway's game of life !== Machine learning

Deep learning / AI in the sense that is currently being applied for eg. autonomous vehicles !== consciousness

AI has taken enormous strides but is yet to grow beyond the hyper-specialized applications we see today. We're still far off from generalized AIs that would count as conscious, intelligent beings, let alone having this all emerge from a blockchain's PoW scheme. We're barely able to create programs that pass the Turing test without relying on special tricks such as having the program pretending to be a 13-year old Ukrainian boy [1], let alone creating an algorithm that goes beyond pretending.

So while it's incredibly important to start discussing the implications of possible future developments in the AI space -- knowing that society takes decades to adjust to new ethical challenges -- we're still far away from what could one day become the singularity.

[1] https://www.theguardian.com/technology/2014/jun/08/super-computer-simulates-13-year-old-boy-passes-turing-test

[hacker1001101001]

Hi,

I just want to float a crazy idea that occured to me a few days ago. Don't think I am the only one around here thinking about this matter (and I am lazy searching the forum for a similar thread).

What is Bitmain developing other than ASICs? SOPHON, their deep learning AI.
Now imagine the incredibly wast amount of computing power now already connected to all the mining pools in the world being suddenly somehow switched (or waking up) forming a consious AI.

Any reactions?

ASCI has a is specially programmed to do just mining and solve some specific types of mathematical problems which give us the mined coins and it does not have any further use in any other computing unless the hardware changes. AI can be added to increase the hash rate of every miner and make it faster and intelligent in mining crypto.

[HeRetiK]

AI can be added to increase the hash rate of every miner and make it faster and intelligent in mining crypto.

Unless an AI is successfully trained to create ever more efficient ASIC layouts that then get cast into silicon that's not how any of these technologies work.

[pebwindkraft]

I meant to understand, that AI systems need to be trained for a long time, before they can "grasp" the idea behind an object. Example: take a cup of coffee. Human beings are able to grasp this object and it's function after two or three visual observations. AI systems need to be trained with 2000 pictures of coffee cups, and then they can decide, if it is a cup of coffee or something else. In the underlyings of this "intelligence" is  simply a very fast comparison engine, able to handle huge databases. I would even say this is a matching machine. Like partnership matching machines. In principle they "learn", how to sort their pictures efficiently. These systems are reactive, and not (yet) self deciding. Intelligence is an expression, which is based on experience, and capability to make decisions for the future on it. Every human being (well, nearly) is doing this automatically, at more or less complex levels. An AI system, if you didn't tell it (train it), how to use the cup of coffee, will never ever understand what a cup of coffee is. AI is promising since +25 years solutions, but til today hasn't advanced very much, besides gaining speed in CPU power, machine's ressource optimization, and extremly fast sort/recover algorythms (basically "if-then"s). I know that I just became a target (or victim) for all AI fans, cause I'm pessimistic on promises of AI I understand that AI is a substream of IT technology, and necessary, cause it provides additional path of managing huge data. But it is not a promise of "intelligence". So my pessimistic view is against the "intelligence" usage and understanding in this buzz word. Increadible would be a much nicer word, cause the calculations done by these machines, and the algos are really fascinating. Obviously increadible isn't a good marketing word... So (my personal) summary:

The really big advantage is the speed, at which the systems work, once they have been trained to sort the data correctly. And the second advantage is, that you can use all marketing language driving cars, neuronal networks, system learning, deep learning, ...), to sell it to the customers 
The promises of intelligence are yet to be delivered.

To come back to bitcoin:
Imagine, equivalent of 2-3 cups for human beings to understand the blockchain. And AI systems times 1000... What is the current size of the blockchain?
OP seems to be correct, to put Sci-fi in the headline...

[Spendulus]

I meant to understand, that AI systems need to be trained for a long time, before they can "grasp" the idea behind an object. Example: take a cup of coffee. Human beings are able to grasp this object and it's function after two or three visual observations. AI systems need to be trained with 2000 pictures of coffee cups, and then they can decide, if it is a cup of coffee or something else. In the underlyings of this "intelligence" is  simply a very fast comparison engine, able to handle huge databases. I would even say this is a matching machine. Like partnership matching machines. In principle they "learn", how to sort their pictures efficiently. These systems are reactive, and not (yet) self deciding. Intelligence is an expression, which is based on experience, and capability to make decisions for the future on it. Every human being (well, nearly) is doing this automatically, at more or less complex levels. An AI system, if you didn't tell it (train it), how to use the cup of coffee, will never ever understand what a cup of coffee is. AI is promising since +25 years solutions, but til today hasn't advanced very much, besides gaining speed in CPU power, machine's ressource optimization, and extremly fast sort/recover algorythms (basically "if-then"s). I know that I just became a target (or victim) for all AI fans, cause I'm pessimistic on promises of AI I understand that AI is a substream of IT technology, and necessary, cause it provides additional path of managing huge data. But it is not a promise of "intelligence". So my pessimistic view is against the "intelligence" usage and understanding in this buzz word. Increadible would be a much nicer word, cause the calculations done by these machines, and the algos are really fascinating. Obviously increadible isn't a good marketing word... So (my personal) summary:

The really big advantage is the speed, at which the systems work, once they have been trained to sort the data correctly. And the second advantage is, that you can use all marketing language driving cars, neuronal networks, system learning, deep learning, ...), to sell it to the customers 
The promises of intelligence are yet to be delivered.

To come back to bitcoin:
Imagine, equivalent of 2-3 cups for human beings to understand the blockchain. And AI systems times 1000... What is the current size of the blockchain?
OP seems to be correct, to put Sci-fi in the headline...

We need ....

Coffee 2.0....

[RNC]

Hi,

I just want to float a crazy idea that occured to me a few days ago. Don't think I am the only one around here thinking about this matter (and I am lazy searching the forum for a similar thread).

What is Bitmain developing other than ASICs? SOPHON, their deep learning AI.
Now imagine the incredibly wast amount of computing power now already connected to all the mining pools in the world being suddenly somehow switched (or waking up) forming a consious AI.

Any reactions?

You are bang on right, you are not alone with your thinking and should look at "Cog-Net" because they
are not keeping a cage around it (AI) and it's free to come and go as it pleases.

I would give you a merit but people like me and you here don't have any to give away do we now

[RNC]

To come back to bitcoin:
Imagine, equivalent of 2-3 cups for human beings to understand the blockchain. And AI systems times 1000... What is the current size of the blockchain?
OP seems to be correct, to put Sci-fi in the headline...

I very good read until we got to the bottom and no it does not come back to bitcoin at all apart
from the mining machines becoming a host for the AI as it escapes the lab so maybe you felt the needed
to add that bit at the bottom to simply stay "on topic" in the forum.

Talking of "Chains" we are being pushed off the top of the food-chain and maybe we deserve it anyway.