Features of the SYPWAI project
SYPWAI is a decentralized AI collaboration platform. It unites a business that needs to get monitoring using neural networks with the performers of this work. The neural network development involves a large number of highly qualified special-field experts and the data platform development for their training, which not only takes a lot of time, but is also expensive.
Today we were able to interview one of the key people in this area – Abdigani Diriye, SYPWAI CEO Specialist.
Abdigani Diriye answered our questions about the project
Earlier, you said that your team knows how to improve the neural network accuracy. How did you manage to do it?
The development of any artificial intelligence project consists of two different components. First, you need to take the data and mark it up. Marking up is a very simple process, but the work is monotonous and time-consuming. You need to explore from about a hundred thousand to a million different types of data to train a neural network.
Second, you need to write a neural network. This is, in fact, a simulation of a small piece of the human brain, which then, similarly to a person, will be able to solve some problem. The neural network is “punished” for incorrect answers and “rewarded” for correct ones during the training process.
This simplifies the creation of any solution. The markers do their job, we check its correctness, train the model on the received data and create an API.
What is the main idea of the SYPWAI project?
The main idea of SYPWAI – anyone can participate in the neural network training. The process seems similar to children educational game. You look at the photo of the cat and answer who is this: a fox, a cat, a dog or a heron. Thus, everyone can earn and benefit society. The main task of our artificial intelligence is to predict the best way to achieve this goal by enumerating all possible variants of events.
What is more important in the SYPWAI ideology: training neural network or a social function like providing workplaces for people with disabilities, and so on?
We have been working with artificial intelligence for a long time. It became clear that there is a big problem: business often cannot allocate a separate section for data science, which will create a solution based on neural networks. Markup is time consuming. The AI market has not yet been developed: there are no large companies and large-scale players. We would like to develop an infrastructure that will be beneficial to everyone. In the process of platform development, it became clear that new occupation could appear – a data marker. Therefore, it is impossible to choose a priority, everything is important.
And what about working process?
First, you need to complete the test task. Since it is elementary, the team will take only people who have chosen 99% of the correct answers to full-fledged work. Tokens at this stage are symbolic – hundredths of a cent. In the future, the work will be as follows: first you do your part of work, then other participants check it.
When the answers of most people coincide, the data is sent to further processing: if seven people out of ten gave one answer, and three gave another, then the number here outweighs. This is how the users’ confidence rating is developed: if they deliberately choose the wrong option, their rating is decreases and they receive less funds.
What about corporate clients? What requests do they have?
We receive a request from a company, a public organization, a research group that needs the help of our artificial intelligence. Also, our company also makes a request for solving global problems. They talk about their main problems, their goals.
If a company develops prostheses, we will tell how to make them better and cheaper. If a public organization is trying to identify how to reduce crime, we will train artificial intelligence for this option.
Based on customer surveys, we prepare research studies and questions to train artificial intelligence for this company. Scientific research help to set the vector for artificial intelligence training. Answers to questions from people help to identify their real needs in order to improve products according to people’s demand.
Tell us about the main algorithms of system operation.
For each company order, we have a large amount of data including: problems or tasks from the company, the desired goal, the resources that the company owns, scientific research, people’s answers that help the company.
This data is encrypted using the XOR algorithm, and data transfer between servers and devices is carried out using public key encryption technology, so that hackers do not have access to intermediate data.
All data is mixed and sent to completely different places. It makes this process decentralized and independent of systematic dependencies.
The data is divided into small blocks so that small devices with low power can learn. All devices receive a block of data every hour and the algorithm using which the device will learn. Small devices receive 1 block, and a large device or cluster receives an array of blocks relative to its capacity.
Within an hour, the devices train their neural network using this data. After an hour of work, they send the final model of their neural network to our servers.
How much will a student earn who, on the way to study, makes a markup of a neural network using a smartphone?
On average, you can earn $ 5 per hour. It turns out $ 300-400 per month. The important thing is that a wide range of people can do the markup.
Are earnings in cryptocurrency? In light of the latest news, this may cause skepticism, don’t you think?
A person can easily transfer cryptocurrency to a bank card or other payment system. What is interesting about the payment in cryptocurrency: if you do not want to take risks, you can withdraw money immediately at a rate fixed in dollars. If you earned a few dollars in an hour, you can transfer them to your account, any problems. And if the platform develops correctly, then the value of its token grows and it is possible to leave part or all cryptocurrency in the hope that in a month $ 5 will turn into $ 7. However, no one can guarantee this, it all depends on the desire of the person.