Here's why it is a good thing that AI decides your next job

By now, you’ve heard the horror stories. AI that’s racist, sexist, or both. Screening tools ranking white men’s resumés higher, language models reinforcing gender stereotypes, and artificial intelligence that seems to have a homophobic bias.

- EDMUND CUTHBERT

And should we really be surprised? Generative AI gets its training data from humans, after all, and humans are biased. If you mix their views together in a statistical soup, you get all the bad opinions, all the snap assessments we make every day based on stereotypes – and then you create systems that get to work enacting those prejudices.

But there’s a heck of a lot of nuance being lost in this debate. Like a lot of recent discussions that have become hugely polarised and politicised – raw milk, fluoride, home-schooling, electric vehicles – the truth is inconveniently complicated.

It’s easy to be suspicious of AI. In fact, it’s important that we are. It’s not just a composite of human intelligence – it makes its own, new assumptions, and businesses are keen to put it to work. It scans medical files for malpractice, advises lawyers and parses resumés, reads PDFs at lightning speed and constructs graphics out of sentences. And as we run out of authentic, human-generated data to feed the models, companies are turning to AI models themselves to generate more training data. This synthetic data can be even more prone to bias, and overreliance on it could lead to many more embarrassing moments – like Google’s Gemini AI generating images of Black Nazi soldiers in 1940s Germany.