Check out our new blog post from @robusthq on AI failure modes! Turns out lots of things can go wrong, whether it's your panda image or pandas dataframe: https://t.co/JOlkcsiDq2
An analysis by Twitter out today finds that Twitter's automatic image cropping algorithm is more likely to choose to display the photos of women over men and white faces over Black faces https://t.co/rzshBTgqnF
Governments, companies, and people are making increasingly important decisions based solely on the output of ML models.
But these ML models are subject to deep operational risks and biases.
So what do operational risks look like in practice?
Goldman Sachs and Apple made headlines for failing to address the operational risks associated with their credit decision-making algorithm behind the Apple Card.
This resulted in an investigation by the New York Department of Financial Services.
@robusthq feels like the exact right place to be working in 2021
Securing AI in a decade where AI/ML will explode into widespread adoption?
Count me in 🚀