Elena Litovchenko: At Ruwiki, We Are Building a High-Quality Knowledge Base for AI Training
In an era dominated by social networks and privately owned news and analytics platforms that generate vast amounts of information, the quality of data circulating online has become a pressing concern. Neural networks are trained on this data, making its reliability especially important.
Developing robust verification mechanisms has become a strategic priority for every country. Errors that accumulate and spread can undermine confidence in the results produced by major platforms. Experts discussed this challenge during the Second Global Digital Forum in Moscow, at the panel discussion “National Code Without Borders: Exporting Art, Technology and Media.”
Elena Litovchenko, deputy general director of the Ruwiki Internet Encyclopedia, a partner of the Second Global Digital Forum, noted that Russia already offers a tool to address this problem.
“Every AI system, even the most advanced one, is trained on some kind of datasets. If the data is chaotic, drawn from blogs and whatever else is available online, errors start to multiply. The result is a digital echo. This is a global challenge, including for the scientific community.
How can we solve this problem in a way that works universally? Here in Russia, at Ruwiki, we have found a solution in a hybrid encyclopedia model. What does that mean? We are not abandoning technology. These tools are excellent at gathering, updating and processing information. But final verification always remains in the hands of experts and subject-matter researchers. Our know-how lies in combining the speed of machine technology with human critical thinking. In this way, we are building a high-quality, verified and comprehensive digital knowledge base, a digital foundation for AI training,” the expert said.
For international partners, Ruwiki offers the Digital Ark (Tsifrovoy Kovcheg) project, a solution for creating a national digital encyclopedia in each country. It enables information to be verified while taking national context, values and local laws into account. Such databases can help ensure that neural networks produce accurate, objective and reliable results in each country.
