Our audio channel aims to spread the knowledge shared in all of Step's events and reach everyone looking to learn more about the latest advances in technology, innovation, and investments from accomplished leaders.
Our audio channel aims to spread the knowledge shared in all of Step's events and reach everyone looking to learn more about the latest advances in technology, innovation, and investments from accomplished leaders.
In a discussion on the ethical implications of AI algorithms, Alnura Belyalova, Karim Tout, Sandra Helou, Maryam Al Ansari, and Andrea Ronchi deliberate on the potential biases inherent in automated decision-making systems and strategies to mitigate them.
Sign up for Step events https://stepconference.com/
Subscribe now to the Step Feed newsletter for a daily roundup of the top tech business news https://subscribe.stepfeed.com/
_____________________
HIGHLIGHTS
(0:20) Privacy Concerns in AI Ethics
(4:39) Ethical Considerations of AI Bias
(8:35) Bias in Data Collection for AI
(17:39) Tech Industry: AI Ethics and Bias
(23:55) Impact of AI Bias on Businesses
(29:34) AI Purpose and Bias
_____________________
QUOTES
"So realizing that, first of all, it's not as safe as you think it is. You don't have a say in this, you can't really say no, you can't use why information to make the process better, or use it to make the processes better."
"What we need to have is awareness and spread this awareness amongst each other who don't know much about the fact that startups or other companies can access your information."
"Bias is not always bad depends on the application itself and the objective."
"It's impossible to not have any bias, it's impossible to be fully objective, there's always subjectivity, and we cannot be random."
"If I use predictive models to make a prescription that are making the real discriminations between potential usual users, that's where you don't allow the freewill of the user to enter into places."
"It actually depends because there's two different types of bias, right? There's bias, which is very kind of totally right."
"But there are different levels of bias and how it actually gets applied, and what happens to it."
"The whole model AI model design and dataset design and feature engineering should fit well with the objective that I'm trying to achieve."
"You need to safeguard the other people that are around there, you know, when you're building an online world, a virtual world that's accessible by multiple people."
"So focus on the dataset, make sure the dataset is good quality, not biased, and you'll get a good AI system."
_____________________
If you’re enjoying this podcast, please subscribe to Step Podcast and leave us a review!
Spotify → https://spoti.fi/3LIG8wj
Apple Podcasts → https://apple.co/3PxIXU5
Anghami → https://bit.ly/3lDKxG4