search

daviddao / awful-ai

😈Awful AI is a curated list to track current scary usages of AI - hoping to raise awareness
https://twitter.com/dwddao
6.94k stars 229 forks source link

Pitfalls of AI #15

Closed Shashi456 closed 5 years ago

Shashi456 commented 5 years ago

Can we add biases that are made within ai ? I think that should be added because these models are trained on data curated by humans which on some level just portray the biases we have inherently.

And on the other hand, I think one section of ai research that belongs here is adversarial examples, which is very relevant in the case of autonomous car safety. Since these adversarial examples lead the neural net to misclassify and in some cases also distort the input image enough to not detect an object at all .

daviddao commented 5 years ago

I try to list those points in the first sentence of the list: Artificial intelligence in its current state is unfair (biased) and easily susceptible to attacks (adversarial attacks). Also algorithmic bias is mentioned in possible contestational research.

Regarding adversarial examples, you are right! I think adversarial examples and especially adversarial defenses is actually a great point to the list!! Do you want to submit a PR? :)

Shashi456 commented 5 years ago

Yeah I don't mind, but how do you think I should structure it, because purely adding papers would result into a huge list.

daviddao commented 5 years ago

I think it makes sense to just add a paragraph (like algorithm bias & fairness). If you submit the PR I can adjust it :)

daviddao commented 5 years ago

Closing with #16