Can digital data be better than real data? The immediate answer would be no, but the reasoned answer is yes. The proof comes from Replica Italia, which represents the first digital twin of the Italian population. It should not be considered a standard dataset because the system reproduces the entire demographic, social and behavioural structure of the country. This means that it simulates just under 60 million individual profiles, starting from the characteristics of each individual (age, profession, interests, habits).
Synthetic data with GenAI
Accessible via the web and equipped with an intuitive interface, Replica Italia is a conversational platform where the user can select the variables that interest them based on their needs. In doing so, it only takes a few seconds to generate the virtual population in line with the chosen characteristics, from which to obtain collective data through the appropriate prompts. It’s the power of digital twins with GenAI.
Let’s take an example to better understand: if the objective is to investigate the potential of a fitness tool, the most suitable age group for the target is selected, interested in the gym and the evolution of training. In this way, replicas are obtained in text format or structured data on tastes, ideas on types and use of favourite tools.
In the era in which digital meets GenAI, synthetic data is a useful alternative for many companies. This is why Clearbox AI, the company that created Replica Italia, started working in 2020 with the ambition of overcoming the limits of real data. Which refer to structural bios and privacy-related issues.
‘We have developed a proprietary model for the generation of synthetic data, based on statistical models and machine learning, capable of producing realistic data from aggregated, anonymous and public sources’, Shalin Kurapati, CEO of Clearbox AI, told Italian Tech.
Can we trust synthetic data?
At this point, it remains to be seen if and how synthetic data is reliable and, therefore, useful for carrying out studies, research and surveys, as well as finding ideas for calibrating business plans. Clearbox AI’s solution is Sure, a validation library that includes advanced metrics of data privacy and fidelity, with which to establish the likelihood of synthetic data with respect to the reality they intend to describe or summarise.
‘It’s the tool we can use to scientifically measure data quality, guaranteeing that it faithfully reflects the dynamics of the real population, without ever exposing sensitive information,’ explained Kurapati. Regarding the risk of bias, Replica Italia reduces the threat as it expresses trends and behaviours modelled by aggregated data and not by subjective responses.
From the technical point of view, Clearbox AI’s digital twin will continue to evolve along with the real country, simulating changes in citizens’ tastes and habits. This evolution allows marketing researchers to have the necessary information at hand at any time without having to disturb people.
As for the future, Italy is the first step because the startup’s goal is to expand its range of action by creating digital twins of other populations. ‘We started with the Italy because we already had structured data, but the strength of the model lies in its scalability and adaptability to any geographical context,’ said Kurapati.
It makes little difference whether it’s the United States, Asia, South America or Africa as long as quality data is available. Clearbox Ai is already working to gradually expand into the European and US markets. With the upgrading of the system, the startup founded at the Polytechnic University of Turin is set to become a leader in the development of global digital twins from which a long list of information can be extracted—a significant advantage for the many companies active in various sectors.
