Amazon tried to roll out some algorithms to do resume screening internally and whatever they did they couldn't get rid of a gender bias that the algorithm was picking up and they eventually decided not to roll this out.

Now Amazon is a large enough company has a lot at stake

and did all the tests and realize that there are these biases

and they did have decided not to roll it out.

But who knows there might be companies that are using these advanced AI that have these biases and so they are making mistakes as well.

So I think we also have to think about, you know, not only in terms of retraining

humans, but also these AI have their limitations as well, and so how do we best

use them? And when we think about where, you know, these biases

and where they come from at the end of the day,

you know, we talked about machine learning

and the fact that this AI are mostly learning from data.

So they're picking up biases that are in the data.

So a recruiting algorithm is trying to figure out

which people to invite for a job interview. And it's looking at decisions made by others in the past and if there have been biases that humans have had it picks up those biases and now it institutionalizes the biases and it almost scales their impact because it's

making those decisions systematically for millions. And so we have to worry about those issues as

well. So I think there's a number of issues. There's a short term issue around algorithm bias. There's an

issue about workforce training. So I think the ramifications are huge.

If we select one technology that's going to have a huge impact on society over the next 50 years,

that we need to prepare for as individuals as regulators today, then it would be AI.

I'm Sreni Rao and this is the unmistakable creative

podcast where you get a window into the stories and insights of the

most innovative and creative minds who started movements, built thriving businesses,

...