Pages

2nd P.S of Nea Erythraia

  • Our Activities:

     

    Quality Data & Biases Activity

    • Last week, our students ran a very interesting activity with their class teacher Sotiria Marolia. She asked them to draw images of children in order to teach a Learning Machine program to recognize human children from a drawing. Through the past activities we had in the ICT class ( machinelearning4kids "make me smile", Quick Draw from AI experiments with Google ,Accenture Intelligent Space Exploration game) the students should have now a better understanding about how Machine Learning works. Let's see what they drew.

    • Base on the output the students drew in our previous activity, we will reflect this week on those drawings and talk about the importance of the Data and the need for Data Quality. We will also talk about biases and check if some of those biases have been reproduced in the children's drawings.

    We asked our partners to join and help us by completing this survey with their students!

     

    • Students from St1 and St2 class also filled the Survey we had created with the pictures of the Student of the St2 class. As they had to evaluate there own drawings, they realized finally how important it was to create and choose quality data when it comes to use them in a Machine Learning process. When they completed the second part of the questionnaire, they realized that there were unconscious biases in all of us. We tried to discover some of these biases and find out even more of them. We will continue this discussion when we will have all the survey results from the partners' schools.

     

    • Results for the "Quality Data & Biases".

    Based on the results of the questionnaire, students have realized how important are Data (well defined, quality, diversity, without misunderstandings) to train a Machine Learning program. They also realized that they have reproduced some  Unconscious Biases (girls have long hair and boys short hair) and have taken into account children with their own characteristics (European, no glasses, no fat children, no or very little children with handicaps, no African nor Asian children...). 

     

    Illusions with  NVIDIA AI PLAYGROUND (NVIDIA AI Research in Action): meet GauGAN, an AI that creates fake landscapes that look real.

    Students played this week with Gaugan and realized how easy it was to create fake photos with the help of an A.I. What can we trust?

    (see our all A.I resources Here)