How to de-bias AI

How to de-bias AI

Screenshot 2019-10-01 at 14.45.12.png

In two years’ time, the world will have algorithmic and anti-bias data auditors to tackle the ‘pale, male and stale’ artificial intelligence sector. That’s because of the need to resolve data and compliance challenges with high-quality, diverse and unbiased training data, says CCS Insight in its latest Predictions for 2020 and Beyond report.

Things are already in motion here. The firm points to an algorithmic accountability bill that was introduced into the US Congress this year. It would require certain algorithms to be regularly reviewed for signs of bias.

And by 2023, a lack of diversity in data sets will push a wearable device maker to pay users for their data as a means to incentivise underrepresented demographic groups. The move aims to improve the quality of research and data for training of artificial intelligence models, says CCS Insight.

Also, environmental pressure will virtual reality displace 20% of business travel by 2029.

“The use of the technology for enterprise collaboration is prompted by the spiralling cost of business travel and the advent of more environmentally aware young people in the workforce,” according to the report.

Facebook's 'reckless indifference'

Facebook's 'reckless indifference'

Commodification drives deepfakes growth