Page 63 - CT_AI_Class-8

Consider the following example:
ethical minds
Hindi Sentence English Translation from AI
Data manipulation to fit AI models वह डॉ�टर है। He is a doctor.
can cause biased predictions and
affect fairness. To address this, In this case, the AI assumes the doctor is male, even though
accurate, unbiased data should the Hindi sentence does not specify gender. This happens
be used, data processing must because the training data often contained more examples
be transparent and AI systems of male doctors. As a result, the AI learns these stereotypes
should be regularly audited for from the data and might apply them when translating. This
fairness.
shows how crucial it is to ensure that training data is balanced
and diverse to avoid reinforcing gender stereotypes.

21 st
Century #Creativity
Skills
concept capsule

A machine is trained with the following dataset:

Training Data Label

Apple

Mango

Now, test the above model for this new data:

Will the trained model identify the fruit correctly? Why?

Will the trained model identify the fruit correctly? Why?

Will the trained model identify the fruit correctly? Why?

Will the trained model identify the fruit correctly? Why?

In the sequence 100, 90, 85, 75, 70, 60, ..., what is the rule?
a) Subtract 5, then subtract 10 b) Subtract 10, then subtract 5

c) Subtract 10 each time d) Divide by 1.1

Data and Fairness in AI 61