they called it "NORMAN"
So, if you have only saw AI villains in movies like Arya(eagle eye), WOPR(wargames) and plenty lots which is just a science fiction. Have you heard about real evil AI program ?
Bias is the most common term in ML which is the most sensitive factor of ML. suppose we have a model to predict weather and the data we train the model is all about hurricanes so what would you expect ? Can you expect this model be predicting about rainy day ? no....
This is what happend to Norman AI which was image captioning neural network which was fed wrong data(deliberately) to see how does it caption good pictures.
It's not a mistake infact neither complicated. You fed model with criminal records and train on police officers record and use it on civilian. What result it would yield ?
You can find more about Norman project on MIT site:
Thanks for reading 😊
Top comments (1)
What it sees in ink blots is kind of disturbing.