Permalink
Branch: master
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
82 lines (46 sloc) 7.07 KB

GpuHub schema

Machine learning is a high cost product.

And of course, we want it to be much cheaper. To get the result not in 1 day or 1 week, but in just 10 minutes, while paying as for renting an extra 10 gigabyte of a hard disk.

Those who are engaged in machine learning with 3D objects understand this issue even better.

There is an acute question of ML of 3D objects and processing of mixed reality, processing of parallel multimodal streams of the virtual, augmented and providing reality (VR, AR, PR), containing visual, audio, tactile and other channels. Namely, how to provide the necessary computing power? Both in terms of costs and of an actual availability to involve a wide audience.

Can we say that the GPU computing capacity today is less than what is required on the market?

We suppose that the market is looking for more automation, as well as expanding the boundaries of perception and feelings. Companies are ready to create and promote it in a greater capacity. And it’s needed more GPU power, a lot, and at an affordable price.

However, over approximately the past 3 years, thanks to the blockchain boom, in different parts of the world a wide group of people has developed, who have mastered the placement and maintenance of GPU-rigs provided for a fee, as part of mining.

The more capacity is involved, the higher a profit.

In order to switch their capacities to ML, they often have some limitations:

  1. it is necessary to upgrade the equipment: motherboards with sufficient bandwidth of each video card slot (PCIe), upgrade the processors to the server level, increase the RAM and the size of the hard disk several times.
  2. have software and mechanisms that are guaranteed to protect sensitive data of customers (ML, MR) from the possibility of illegal access to them

However, in general, when there are reasonable grounds to believe that PoW mining will end as part of evolution of the principles of consensus, we can say that there is a computing capacity provider on the market that can reduce the cost of ML. Or in other words, that there is a significant vector of providing decentralized GPU-computing.

It is in interests of miners to have steady demand, constant loading with a stable payments. And the market, in which the percentage of machine learning is growing systematically, will be ready to give it. Just because automation is more effective with ML, ML implementations will be growth further.

Just take a look at relevant statistics.

ML market forecasts

Courtesy of Forbes

At the same time, a huge number of people would like to have some very simple device, a box at home. Something like a gas stove, boiler, or even an electric kettle. I pressed the button — it worked. Entered in the web interface the address of the wallet — get paid.

Automatic income, with full-fledged constant load of this box. Also such providers (owners or holders of Isolated GPU Workstation, IGW) do not have access to internal elements and aggregates. A person can not without violating the integrity of the body to connect and read sensitive data. And if breaking — continue to be a legitimate part of the system and read the data.

Thus, when necessary, sensitive data (transmitted by a consumer of ML, MR) are transferred to the insensitive set within the IGW. And already in this form, the data goes to rigs.

Then, after processing, they are again delivered to the IGW nodes at which decoding occurs.

IGW solves several problems. First of all:

  1. increasing a degree of the network decentralization
  2. sensitive data protection
  3. wide audience access to full-fledged VR and MR spaces and experience

IGW owners receive payment for the safety of the device placement and its isolation. It is in their interest to provide it in order to earn income. At the same time, the main method of proving isolation is the analysis by the software agent of the internal states of the system for access and modification, when interacting with the regulatory nodes via a secure channel. Taking into account the rules for updating the status of the device when a request is made for a forced shutdown for a period of time.

Also, today on the market there are many persons and companies that consider investment as their main business. They are interested in being able to safely invest in a sustainable market and get significant margins. Consequently, if there were mechanisms for the secure provision of investment funds for those who want to accommodate the rig, IGW, as well as for the producers of IGW, then investors would be able to systematically extract profits through this mechanism. Having forms of protecting their investment.

And if we talk about the producers of IGW, then we would be able to see a huge number of not only major players, but also many small and medium-sized enterprises, workshops that are capable of collecting and servicing IGWs according to specification. With appropriate warranty obligations to a buyer. As well as to an investor in case investment funds were used.

Verification of compliance with the specifications in this case is carried out by the software agent when connected to the system, for compliance with the state and layers of software, as well as analysis of a stochastically generated video stream whose object is the IGW.

Thus, in this situation, each participant makes money.

IGW manufacturer — in production, sales and service.

Investor — on receiving interest as part of return on investment.

Owner/holder of IGW — by providing workstation for computations. Which does not require any special skills.

Owner of the rig — also by leasing of computing power, having the ability to predict the load and more confidently expand its business.

Consumers ML, MR — getting the desired result for less money. But most importantly, by increasing their sales due to the greater inclusion of machine learning in their activities. Increasing the percentage of machine learning in their products. GpuHub schema

Have you thought about what is the most important factor expressing the growth of the wealth of society?

It seems to us that if it were necessary to use only one word, this word would be automation.

Other things being equal from a political and cultural point of view, the higher the level of automation of the production of goods and services, the higher the level of wealth available to members of society.

Thus, an increase in the number of ML will lead to an increase in the automation of the production of goods and services. Including with help of more and more robotization.

The growth of automation will lead to an increase in the well-being of the participants of the described ecosystem.

The growth of well-being, in turn, will allow ecosystem participants to become more and more involved in the process. In the role of consumers, providers, contributors.