How it works

Our Neural Networks

How we train BIAS firewall

Solving the problems of bias in the field of Artificial Intelligence is one of the great challenges that society is facing nowadays. In our case, there is no discrimination of any kind since the Neural Networks are trained with totally anonymized texts and we do not know the age, gender, race, biological sex, height, social class, culture of the people who write or speak the texts.

Before training the Neural Network, a random sampling is done so that the research team does not know who has written that text when labeling it. Erudit AI trains its Neural Networks with thousands of texts from all over the world and in both languages Spanish and English. The only thing that is diagnosed and labeled is the language and not the person behind it.

From data to metric
Got any suggestions?
Let us know!