Heed the business – understand its pain
Working with businesses we understood that almost everyone realizes the opportunities of artificial intelligence but does not have sufficient funds for its implementation. The process of assembling a department, setting a full-fledged task, implementing a solution is so long and expensive that many companies gave up at the stage of searching for data scientists. Today a good data scientist costs a huge sum of money (in Russia, on average, from 170 thousand to 600 thousand rubles per month, in the US – from $ 100 thousand to $ 250 thousand per year); and the best professionals are already working in the big companies such as Yandex, Google, and Facebook.
We understood that the client needs the ready-made solution on the base of neuronet. You upload data to us and we prepare your data set for neuronet training. Then we select the developers in the competition format, control their work, select the best solution and you just have to use it.
How does Dbrain work?
Markup of the data set. The client loads his data (e.g. a set of photos). People register on the site, pass a test (noting which animal is shown in the picture – to weed out those who click randomly), and then start to work: they look at photos, put a short description about it and thereby train the neuronet. For each click, the “markup specialists” get the Dbrain (DBC) cryptocurrency on their accounts, and then withdraw it to the Ethereum e-purse.
Neuronet algorithm. When the data is marked, the platform finds a developer through an open competition. This developer writes the algorithm of the neuronet according to the terms of reference from the client. The developer receives a fixed payment (from $ 1 thousand to $ 300 thousand) for a neuronet of private use or a percentage of the cost of accesses if the network is public. Dbrain checks the ready solution, and the business connects it through the API. Today, most of the cases are associated with image recognition, but there are also cases with sound and text.
Companies do not want to put their data into the public domain. Banks and other large businesses are afraid of downloading data to the cloud – they want to do everything within their own network that is not connected to the Internet. Therefore, we came up with a solution based on the blockchain, in which we can guarantee to clients the safety of the data: the client will be able to download data independently to train the neuronet, and purchase from Dbrain the number of requests to this API (for example 100 thousand) and to buy additional requests as needed.
We came up with a plan of solving the security problem: we lay out the data in parts; give only 0.1% of the data set to the developer for writing neuronet algorithms, and when the solution is ready, we substitute a large data set already on its servers.
To avoid the mistrust (inevitable in projects involving a large number of users), we use two blockchain protocols. One controls the quality of data markup (at least 3 people independently confirm the same image), the other controls the transparency of interaction between the parties. Trying to put the blockchain everywhere is useless, but in our case, it performs specific tasks.
There is also a bounty program. The obligatory condition is registration on the bountyhive.io website. Then you need to open the «active campaigns» tab and find Dbrain.
Click on the pop-up «Translation» menu, choose the most interesting for your campaign.
For example, you chose Twitter, then you see that 59 250 DBR was devoted.
The main rules of a Twitter campaign are to subscribe to Dbrain page on Twitter, do minimum 5 retweets and your original tweets (be sure to specify @dbrainio). The week starts on Monday and ends on Sunday. The page should be real and be created at least 1 month ago. After reading the whole rules you need to put a check and click «Participate» button.