[DPRating] Deepbrain: Decentralized supply of computation resource for AI

Published in

DPRating

15 min readJun 5, 2018

Deepbrain (DBC)

68 | B- (investment grade)
Risk: medium
Popularity: high
Potential: medium (five grades: extremely low, low, medium, high, very high)
Website: www.deepbrainchain.org

Project overview

Deep Brain is currently in the middle stages of its development. With the release of mining machines in May, and the deployment of open source code and AI test network in Q3, DeepBrain has made some decent progress. However, it should be noted that there may not be a positive correlation between the token’s room for appreciation and the mining machine’s rate of return. It is, of course, not possible for miners to know if their ROI will remain positive.

Concerning DeepBrain’s market performance, the price of the DBC token rose, at its height, to more than ten times its crowdfunding price. The price has since dropped back to around five times the crowdfunding price. Aside from the influence of the overall market, the significant price fluctuation during this period may have been caused by overhyped promotion in the early stages of the project followed by a decline in popularity and publicity.

According to the roadmap included on their whitepaper, DeepBrain will focus on the integration and deployment of deep learning engines in the AI layer during the Q3 period in 2018. Its recent milestone events are the release of open source code and mining software. The AI training test network will come online in June.

Project Advantages

DeepBrain has clear strategic positioning and a good track record of meeting their stated goals; important upcoming nodes goals? (e.g. release of mining machines in May and deployment of AI training network in June) are steadily progressed toward completion.

Project Disadvantages

It is not as popular in China as it is overseas, considering that it is a Chinese project.
The code is currently not open source (It will be open source in the Q3 period of 2018, according to the roadmap).

If any of the following occurs, projects may receive a higher score:

Noticeable improvements in community construction and popularity.
Participation of heavyweight partners in the AI field or emergence popular sub-projects based on its ecosystem.

In the following occasions, projects may receive a lower score:

Delay of important milestones (e.g. open source code plan in the Q3 period and AI test network).

Brief:

Project: Deepbrain is positioned as an AI computing platform based on blockchain technologies + data trading. it uses computing resource and data privacy as entry points to provide computing services and support to projects in its ecosystem. In the long term, such positioning will improve the efficiency and liquidity of the original industry along with the development of DeepBrain.

Team: The founder He Yong (founder) is one of China’s early participants in the AI field. He has former experience in entrepreneurship and operating companies. The team is well-experienced in the field of AI, and has been through a long development period. Its accumulation of resources in this industry is beneficial for the future ecosystem construction of the project. As for staff expansion, the AI epert Wang Dongyan joined the team on April 12th.

Milestone completion and code updating: Q1 2018 goals have been met according to schedule, but DeepBrain’s engineering page on Github is not open source. After contacting them, we entered their private library to verify the relative datas.

Market: There are 7 exchanges that support trading of DBC. These exchanges include Huobi, gate, Lbank, Kucoin etc. In terms of the stability of market cap, the exchange price of BTC to the Benchmark bounced back after a dip, which meant that DBC slightly outperformed BTC and the Benchmark during that period; however, compared with its competitors, it has no advantages regarding the Sharpe Ratio.

Popularity: With regards to social media, DeepBrain has good publicity and activity on Twitter and high activity on its English Telegram group. Due to its promotion after release, DeepBrain gained a lot of recognition and popularity from overseas communities; but in China, it is not as popular. Its comprehensive popularity score is nevertheless rated high.

Risk: No significant changes have occurred within the team other than the participation of the primary members. DeepBrain has some first-mover advantages; however, the competitiveness of its peers and new competitors should not be neglected.

Therefore, its score in risk is medium.

Independent Dimension: potential

Evaluations of potential will be based on three aspects — current market cap, liquidity, and project phase.

Liquidity:

DeepBrain uses computing resource and data privacy as the selling points for providing computing services and supporting projects based on its ecosystem. In the long term, such positioning will improve the efficiency and liquidity of the original industry with the development of DeepBrain.

Market cap:

Current market cap is around 1.5 billion dollars, market cap ranking 115. Current price is around 5 times the crowd funding price.

Project phase:

In the midterm . Expectations for mining machines have been met. The next step is to release their AI test network in June. DeepBrain does not aim to be the leader in all subfields of AI, but positions itself with a specialized focus. The ambition of a project is matched with the ability of its team. We evaluate DeepBrain’s potential as medium.

Dimensions:

Project

Strategic Positioning

full score: 60 actual score: 38

Objectives of DBC are

1. Helping enterprises to reduce 70% of hardware costs;

2. Constructing a good privacy protection mechanism between data sellers and buyers (separation of right to use and ownership)

Members of DeepBrain Ecosystem:

1. Mining nodes

2. AI suppliers

3. Data suppliers

Evaluations:

1. Positioning: AI computing platform + blockchain

1. AI computing platform + decentralized data trading platform, it aims to provide complete solutions to solve problems in a comprehensive field (AI).

2. Three elements of AI are algorithms, computing power, and data. DeepBrain mainly relies on computing power and partly on data (data privacy) as their competitive advantages and selling points.

3. Popular concepts such as AI, big data, and the Internet of Things are already enticing without the addition of blockchain. Though most people do not have a deep level of understanding for such technology, they have already noticed a lot of buzz and news concerning it. Popular subject in AI mainly include machine learning, AlphaGo, autopilot, face recognition, and speech recognition.

4. In the Q4 period of 2017, DeepBrain had already gained impressive publicity overseas, because its concept was popular and it was one of only a few projects based on NEO smart contracts at that time.

Innovations:

According to their whitepaper, DeepBrain’s innovations include:

1. Algorithm Optimization: it optimizes operations on CUDA GPU and docks with dominant deep learning framework (TensorFlow, Caffe and CNTK) at present.

2. Anonymous Node Containers: In the processes of data training and model using, data buyers cannot copy data from nodes. DeepBrain uses encryption algorithms and separation mechanisms to realize privacy protection.

3. Elastic Supply: Elastic expansion technology allows containers to deploy automatically and to rapidly copy such deployment to multiple idle nodes during high flows of data traffic, which can effectively respond to bursts of data traffic.

Difficulties: How to cope with high concurrency scenarios? According to the whitepaper, DeepBrain adopts load balancing technology to have every node container cooperate together in order to share concurrent pressure; however, technical details are not mentioned.

Competitors

1. Since DeepBrain was launched a fairly long time ago, and boasts abundant industry resources and a long development period, it has some first-mover advantages in the AI competitive field. However, it does not have any technical breakthroughs and efficient means of competition.

2. After release, DeepBrain acted and promoted frequently in overseas communities, enjoying good recognition and popularity. It still remains less popular in China than overseas.

Therefore, our evaluations on DeepBrain in this dimension are as follows:

Project necessities

Full score 40 Actual score 28

In its whitepaper, DeepBrain mentions that between 2012 and 2016, 5,145 new AI companies were established in the world, with a total investment of 22.4 billion USD. The bulk of this financing was used for the purchase of hardware and computing power. Such high costs are a heavy burden on small and medium-sized enterprises. On the other hand, the right to use and ownership are not separate under current technological conditions, which means data suppliers cannot ensure that data sold will not be duplicated again.

DeepBrain combines full-function nodes (permanent nodes) on large-scale GPU/FPGA server clusters, idle computing nodes on GPU servers of small and medium-sized enterprises or private servers to perform the computing tasks of AI. Mining nodes will be rewarded according to an incentive system designed on the basis of a smart contract. Such a solution, while lowering suppliers’ ability to compute on neural networks, benefits participants via mining.

Based on this description, such an application is readably similar in concept to other resource sharing projects. For example, while projects that share idle bandwidth and storage space expand CDN at a low cost, involved nodes will be rewarded for providing equipment and computing power. The major difference between the two is their different targets. DeepBrain mainly targets business between enterprises, aiming to solve problems in large-scale computing and storage, while the latter mainly targets common users, aiming to solve problems in small-scale computing and storage.

The current state of DeepBrain is an AI computing platform; but in the future, it plans to expand the functions of its data trading platform. At present, it has no properties of smart contracts and underlying public blockchain platforms. In terms of applications, DeepBrain officially announced the release of its first project on blockchain on the 15th of May. Though other information remains undisclosed, according to the latest on Medium, this project is onegame. See details below:

Reference: https://medium.com/@deepbrainchain_74263/onegame-a-project-based-on-deepbrain-chains-public -chain-1f7427495e6f

Team

Technical Team

Full score:40 Actual score:35

Information on DeepBrain’s core members is disclosed in the whitepaper and on their official website. Information on the CEO, CTO, and CMO is as follows:

We notice that He Yong and LEE Chuanfeng all have strong backgrounds in the AI field. Feng Lizeng (architect) once was the chief architect of open platforms in Huawei Technologies, and was also independently in charge of architect designs and core code development for several distributed software systems.

In terms of blockchain technicians, information on Wang LinBin and Yi Changjun is as follows: Wang once was the R&D engineer at SOHU and Giant Interactive Group. Familiar with Solidity, he mainly uses Java and Go to develop. However, his specific experience on blockchain projects is not mentioned. Yi once was the development engineer at Bytom. As for staff expansion, Wang Yandong joined the team on April 12th.

Operation Team

Full score:40 Actual score: 35

Among all the information about its core members disclosed in the whitepaper, Christine Chang (director of North American market) was once the global marketing director at Cheetah Mobile and co-founder of the RobotX Incubator.

Feng HE(founder) is one of China’s early practitioners in the AI field. He is an entrepreneur with experience in company operations and has achieved some degree of success.

Investors and Consultants:

Full score 20 Actual score 16

According to disclosed information, DeepBrain’s investment institutions include GSR Ventures, Gobi Found, GBIC and Bite Holding. Its advisors are numerous, including Ding Jian (partner of GSR Ventures), Yang Zhiwei (partner of GSR Ventures), Jiang Tao (partner of Gobi Found), Hu Tangjun (general manager of Gobi Found), Qian Haijie (partner of Qianshi Investment) and Kong Huawei (director of Shanghai Institute of Computing Science of Chinese Academy of Sciences). Therefore, our evaluations on DeepBrain in this dimension are as follows:

Therefore, the overall score in terms of team is as follow:

Development Progress

Roadmap disclosed in whitepaper is as follows:

Fulfilment of goals in the Q1 period of 2018 is as follows:

Reservations of AI mining machines are expected to start in mid-May at miner.deepbrainchain.org. The test network for the AI training network will be brought online in June. The AI training test network is on course for its planned Q2 release, and is expected to be completed on schedule. It is important to note that the project’s code is currently still not open source — but according to their roadmap, such work will be done in the Q3 period.

Given the above, our evaluations on DeepBrain’s goal fulfilment of the whitepaper are as follows:

DeepBrain’s code library on Github is not open source. After contacting them, we entered their private library to perform basic auditing and verification.

1. The project director is now working on the development of its dev branch.

2. Currently, there are 9 people (auditors of the evaluation team not included) in total who have submitted code to the private library.

3. Under the dev branch, the number of releases is 7 and indicators of Watch|Star|Fork are 1|2|1 respectively. Since it’s a private library, the three values will not affect its final score.

Given the above, based on assessment of its code base quality, our evaluations on DeepBrain’s updating quality are as follows:

Market

Support of Exchanges

Full score: 50, Actual score: 26

We will classify current exchanges into 3 grades — first-class exchanges, quasi-first-class exchanges and regular exchanges. Projects will be divided by the number of exchanges that support trading in their tokens.

According to data on CoinmarketCap, there are 7 exchanges that support trading in DBC, among which there are 1 first-class exchange (Huobi), 3 quasi-first-class exchanges (gate, kucoin and lbank) and 3 regular exchanges. DBC is not listed on Binance, Okex or other international exchanges. Therefore, we determine that the liquidity of DBC is medium. Full socre: 50, actual score: 26.

Stability

Full score: 50, Actual score: 21

Based on Bitcoin, we assessed the changes in the DBC-Bitcoin exchange rate. According to the change in value, our evaluations on DeepBrain in this dimension are as follows:

From February 5th to March 4th there was no significant fluctuation of DBC’s price (blue line) throughout the period, but the DBC-BTC exchange rate (yellow lone) decreased dramatically. DBC-Benchmark exchange rate (green line) kept almost stable. Conclusion: while DBC underperformed BTC, it performed almost the same as the Benchmark.

Between March 4th and April 4th, the price of DBC (blue line) declined, and DBC-BTC exchange rate (yellow line) slightly rose, while DBC-Benchmark exchange rate (green line) almost coincided with the yellow line. Conclusion: DBC outperformed BTC and Benchmark with a similar percentage increase.

Between April 4th and May 4th, the price of DBC (blue line) increased, and DBC-BTC exchange rate declined after a rise, while DBC-Benchmark exchange rate (green line) almost coincided with the blue line. Conclusion: DBC underperformed BTC during this period with a similar increment to Benchmark.

Overall, between February 5th and May 4th, the price of DBC (blue line) and DBC-BTC exchange rate (yellow line) slightly decreased, while DBC-Benchmark exchange rate (green line) rose. We can then conclude that DBC outperformed Benchmark, but slight underperformed BTC during the period in question. According to the 3-month performance, both the price of DBC and DBC-Benchmark exchange rate rose. DBC-BTC exchange rate bounced back after a dip, but declined on the whole. Since DBC-BTC exchange rate declined in the first two months, but noticeably bounced back in the last month, we then determined that DBC was favorable in exchange for Benchmark in the period in question. Full score: 25, actual score: 18.

After DBC was listed, the comparison of its price index with competitors’ (Bottos, golem and BTC) and that of less known currency (Benchmark) is as follows:

Measuring method:

profit/price volatility = sharpe ratio

This ratio shows the return rate of investment target per unit of risk. The higher the ratio is, the higher the return rate becomes when investment takes the same risk.

According to the fluctuation during the past three months, the price volatility of DBC was at 39.84%, which was ranked second among the four.

Between February 5th and April 4th, the price of DBC dramatically decreased, but tended to bounce back from April 4th to May 4th. In the 3-month statistics, DBC’s rate of return was 36.47%, which outperformed that of BTC. However, DBC showed no advantages compared with its competitors. In terms of risk-benefit ratio, it was ranked №4. We then determine that DBC has no advantages compared with its competitors, and its per unit risk to benefit ratio is poor. Full score: 25, actual score: 5.

Independent Dimension: risk

As for upgraded projects, we focus on these several aspects while evaluating risks:

Policy risk: low

Team change risk: low

No significant changes have occurred in the DeepBrain team; its staff expansion was stable.

Development progress risk: medium

Market manipulation risk: low. The price of DBC has been through significant fluctuation, so such risk existed.

Competition risk: medium

Given the above, our evaluations on DeepBrain in this dimension are medium.

Independent dimension: popularity

When evaluating DeepBrain’s popularity, we mainly looked at its official social media, such as the activity of its telegram group and slack group. At the same time, we also took into account the number of followers it has on Twitter and Facebook.

Facebook

430 followers, 4 tweets in April, 25 likes (6.25 on average), 21 reposts (5.25 on average), 1 comment (0.25 on average).

Twitter

30,400 followers, 45 tweets in April, 799 messages in total (17.75 on average), 13568 likes in total (301.51 on average), 28301 reposts in total (628.91 on average).

12,280 people, 5337 people chatting within 5 days (1067.4 on average per day, percentage of active users: 8.69%). Since its active percentage is more than 4%, we determine that its activity is good.

To sum up, DeepBrain’s current popularity is high.

Here are our final evaluations after summarizing the results from all dimensions: