Yes, those AI systems surely create garbage results right now, and worse, once they are trained perfectly, they will provide perfect cover to codify existing hiring biases:
1. Take existing employee base, hired through biased process.
2. Use resumes to train multi-million parameter machine learning model.
3. System will replicate biases, while the process is completely opaque to review by even the people who created it.
But then again, existing hiring practices are just as bad, e.g. hiring European engineers only if they graduated from Oxford, Cambridge, or the University of Bucharest: https://twitter.com/shaft/status/1355696154990628864
Or the still widespread practice in Switzerland to demand hand written job applications for executive positions, in order to have the applications evaluated by graphologists…
1. Take existing employee base, hired through biased process.
2. Use resumes to train multi-million parameter machine learning model.
3. System will replicate biases, while the process is completely opaque to review by even the people who created it.
But then again, existing hiring practices are just as bad, e.g. hiring European engineers only if they graduated from Oxford, Cambridge, or the University of Bucharest: https://twitter.com/shaft/status/1355696154990628864
Or the still widespread practice in Switzerland to demand hand written job applications for executive positions, in order to have the applications evaluated by graphologists…