Decentralized Computations. The Most Promising Projects

Decentralized Computations. The Most Promising Projects

AI Business Technology
October 12, 2018 by Leo Webb
310
Decentralized Computations

A major part of our everyday activities is based on computations. Modern people use computers and smartphones all the time. We need computing power at work and at home. It is used in science, entertainment, media, and stock markets. Today, the market of computing technologies is dominated by cloud solutions. Cloud technologies take on a large part of our everyday computations but they also have a flip side. Companies like Facebook, Microsoft, Amazon, and Google own data of millions of users, and people become dependent on them, which makes purchasing their services a necessity.

On one hand, everyday tasks of average users become increasingly demanding, and cloud computing solves this problem by providing people with the necessary resources. On the other hand, all the information is stored in centralized databases. Thus, if the central database is hacked, the whole network may be in danger. Fortunately, now we have a solution, which is the blockchain technology. Blockchain can use the power of thousands of computers from all over the world, distributing any task among them.

The main principle is simple. Users can rent the computing power of other users or data centers to perform any tasks, from simple calculations or video rendering to deep learning tests. In return, owners of the computing power get rewarded. This approach allows users to solve any task much faster. Processes that used to take weeks to complete can be done in hours or minutes using blockchain.

We decided to take a look at the most promising projects of Decentralized computations

Golem.Network

Golem.Network is a good example of a distributed supercomputer. Unlike the traditional computing systems, it doesn’t have a central server farm. Instead, there are many Providers who offer the computing power of their computers and laptops for a reward. Other users can rent it, paying GNT tokens. The thing is that usually, we don’t use 100% of our computers’ capabilities. Golem can use the spare computing power and pay for it.

On one hand, we’ve already heard of mining so this process may seem similar to how people make Bitcoins and other cryptocurrencies. On the other hand, it has nothing to do with cryptocurrencies. Mining is a “proof of work.” Mining algorithms work with cryptographic tasks, producing new Bitcoins, and this is their only function. Golem uses the computing power with a much more useful purpose.

For example, most movies would never be so beautiful without CGI (Computer Generated Imagery), which is also used in video games. CGI includes tons of graphic details, lighting, and textures, which means very high requirements for the computing power. It would take weeks or even months to render a short scene from Star Wars on a regular desktop computer. Golem divides any task into many smaller tasks and distributes them among computers of the network. Golem has already been used to render CGI scenes, but creators of this project are planning to realize other applications as well. For example, Golem can be also used for machine learning or DNA sequencing.

Sonm

In 2016, a group of Polish and Russian miners decided to find a new, practical way to use their hardware. As a result, they created a project that may become a Golem’s competitor. Sonm is a fog computing platform. Fog architecture uses a network of computers instead of cloud data centers to store information, as well as for communication, configuration, control, and management (instead of relying on network gateways, like LTE networks).

Just like Golem, Sonm has a wide range of possible applications. It can be used as a site hosting or as a powerful tool for scientific calculations. Creators of this network state that they want to replace the traditional hash-based mining. The company received $42 million in ICO and joined the OpenFog Consortium, which also includes such members as Dell, Intel, Microsoft, and Princeton University.

iExec

iExec is somewhat similar to the previous two examples, however, it also has some different features. This network was built on the Ethereum blockchain and provides the computing power. This system also gives its users access to off-chain services and data. Unlike Golem or Sonm, this system doesn’t recruit desktop providers. Instead, it uses the already existing resources, such as Amazon, Kaiko, and Stimergy.

The main difference between iExec and its main competitor, Golem, is that iExec is mostly intended not for a personal use but for decentralized apps (dApps) — blockchain-based applications that use P2P networks of computers. Thus, the market strategy is to use the limited pool of providers from the existing market for resources, offering services for a new, developing market of dApps. dApps are the main priority of this system, which is a reason why this company doesn’t promise to make computation cheaper.

PandoraBoxchain

PandoraBoxchain’s ambitions go far beyond the blockchain technology, as it’s focused on providing resources for machine learning. Today, miners and other users with powerful computers cannot sell their resources to AI-based projects. Companies that develop machine learning solutions also cannot use decentralized computation networks so they need to use data centers owned by Amazon, IBM, Facebook, Microsoft, or Google. Pandora plans to create an open market where providers and AI developers can work together. Pandora also is not going to rely on traditional consensus models (PoW and PoS), introducing its own solution — Proof of Cognitive Work (PoCW). Pandora’s creators claim that PoCW is much more effective than the traditional Proof-of-Work consensus model, as all the computations are performed by neural networks.

This system will reward people who provide their hardware, and its creators also plan to reward developers. This is a reason why they introduced a Research Work Reward Algorithm — 10% of the mined tokens (PAN) will be distributed among creators of machine learning kernels. Pandora’s ecosystem, in general, will include big data providers, kernel authors, hardware owners, validators, and customers.

GraphGrail

This system is also based on the Ethereum blockchain. Authors of this project describe their mission as a creation of an AI available for everyone and trained by developers from all over the world. Inspired by Elon Musk’s OpenAI, this system will focus on the analysis of text data, semantic classification, deep learning, and information retrieval, as well as on others issues associated with the machine learning industry. Developers emphasize the role of natural language processing, which is why they plan to create a dApp marketplace for linguistic applications.

Anryze

This AI-powered service is focused on speech recognition. It records call data and turns it into a text transcription. The main purpose of this system is risk and compliance management. Eliminating central controls, this system ensures the better quality of speech recognition, compared to Google’s or Amazon’s solutions. 70% of the transaction fee goes to the miners, and 30% is distributed among token holders and the platform.

 

Conclusion

Early computers could work only on basic arithmetic tasks, while today we have machines capable of analyzing big data, which use the deep learning technology. We can model almost any problem, using countless variables and expressing virtually any problem mathematically. There’s no surprise that disruptive technologies are associated with high technological requirements. Even production of video content has constantly growing requirements regarding the computing power.

Blockchain offers lots of opportunity for computations, allowing people to use decentralized networks of computers from all over the globe, dividing cumbersome tasks into smaller ones and distributing them among thousands of servers, desktops, and laptops. In return, hardware providers can get rewarded in tokens. Of course, not all the decentralized solutions have the same chances for success. Many of them are somewhat risky for investors, however, there are a few projects that have proved to be viable. Thus, centralized giants like Google or Amazon are going to face a strong competition in the nearest future, as blockchain-based solutions are cheaper and more effective than traditional cloud services.