It seems as though artificial intelligence creates a hype wherever it goes. Its first adoption in the blockchain community was seen when DeepBrain Chain decided to use blockchain to make artificial intelligence available at a big scale on low prices. This happened way back in December 2017.
Their integration of Artificial Intelligence neural networks in a blockchain infrastructure is complicated; however, the service that the startup provides is promising.
According to the DeepBrain Chain’s whitepaper:
The vision of DeepBrain Chain is to provide a low-cost, private, flexible, safe, and decentralized artificial intelligence computing platform to artificial intelligence products.”
What does that mean, and how does it all come into play?
For any artificial intelligence product, the developer needs to go through a rigorous training session by neural network calculations for the model. This often requires putting an astronomically large amount of data into a series of complex algorithms as input as algorithm runs on the data to give a certain output. Every time the algorithm runs, some processing power is used (just like your computer needs power to run); this processing often requires a hardware such as a GPU, a cloud based GPU or an FPGA. More the data that needs to be analyzed, more the processing power would be required, resulting in a better and more expensive computing resources. The amount of computing that a neural network may need can be phenomenal.
Therefore, it is not surprising that much of the expenditure of one artificial intelligence product is dedicated towards buying and maintaining the computing resources. DeepBrain Chain wishes to provide AI developers around the world an alternative to the traditional expensive processing of a neural network. It will do this by making the artificial intelligence neural network operation decentralized and distributed. The computing resources required is spread over mass nodes through out whole world by the blockchain technology.
The cost to run a neural network is thus reduced; the company estimates that it will be less than 30% of the cost of the user’s self-built neural network server and less than 50% of the traditional artificial intelligence centralization cloud computing platform.
As mentioned earlier, data is an integral need to run any neural network; often, companies purchase high-quality data or annonate the low-quality one. However, a lot of the data involves user privacy issues. Data is sold on the basis of the right to use, rather than being sold on ownership. However, this is not respected and the data is exploited (duplicated for use). With the use of smart contracts in blockchain, DeepBrain Chain is able to maintain the separation between data usage and data ownership.
Transactions in DeepBrain Chain are based on smart contracts which are carried out by the use of the token, DBC.
DeepBrain Chain will consist of a network of nodes that are composed of hardware operators. They can take multiple forms including:
- full-function nodes (permanent nodes) composed of large GPU or FPGA server clusters;
- independent nodes composed of idle GPU servers owned by small and medium-sized companies, and
- idle GPUs owned by individuals.
Mining nodes earn their income from two sources:
- AI processing fees paid by AI companies in exchange for computing power, and
- mining rewards from the system
The mining rewards are calculated based on DeepBrain Chain’s reward algorithm.
Thus, enterprises can buy AI services which can take the form of:
- AI trading platform;
- AI data exchange platform;
- AI calculating data trading platform;
- AI model exchange platform;
- AI container exchange platform, and
- AI application exchange platform.
The Token, BDC
The token, DeepBrain Coin or BDC is based on NEO. It i s run on the DeepBrain Coin issuing algorithm on the smart contract of NEO.
There are a total of 10 billion DBC. 50% of it is set to be mined, while 1.5 billion tokens were kept for sale.
The token was sold on 15th of December 2017, in just one day.
It is currently valued at $0.015526 (US), and ranks at 185 (at the time of writing). Further details can be checked here.
DeepBrain Chain is based in China, however it is also working in Silicon Valley. The developers have around 15 or more years of experience and worked for Baidu, Alibaba, Tencent, IBM, Cisco, Supermicro, Huawei, Midea, Netease, China Mobile and Ericsson.
DeepBrain chain AI training net went live on 8th August 2018. The training net is “a trading platform that consolidates computing resources and carries out matchmaking for computing power provider and computing power requester.”
In their semi-annual report, the company published its road map which can be seen below.
Till now, the company has successfully ran multiple AI and machine learning real-life training scenarios which include: MNIST, Natural Language Processing, Voice recognition, driver less cars, machine arm grabbing, smart production AI detection, medical tumor detection, planet inspection among others.
In June, the company also started the Skynet project. Through this the company is recruiting AI computing power from around the world. The purpose of the project is to “build a worldwide DeepBrain Chain Testnet to make preparations for the launch of the DeepBrain Chain Mainnet at the end of this year.”