I found this very interesting. It shows the importance of diversity even when it comes to AI.
https://techcrunch.com/2021/04/19/flawed-data-is-putting-people-with-disabilities-at-risk/
Posted Using LeoFinance Beta
You are viewing a single comment's thread from:
All of it depends on the premise and good training sets. If the training model is bad, then the AI will also be bad. Unfortunately this part still requires humans to filter the information.
Posted Using LeoFinance Beta
100% correct. I was working on a project once which was related to Bank data , the exact same thing happened . Our guide was so bad , he told to just pick 100 random data out of 1 million data and do the analysis , it sucked lol . It was totally off .
That's when I learnt , training model is very important and methods to choose that should be ideal.
Posted Using LeoFinance Beta
Yes, that's where the bias kicks in. Because we don't have a lot of diverse point of views curating the data that feeds the models, they end up "flawed"
Posted Using LeoFinance Beta
Garbage in , garbage out.
AI can only proceed based upon the training data is has. If there are biases in there of any sort, it will pick up on them.
Posted Using LeoFinance Beta