Summary

  • SHADES is a data set aimed at helping developers to spot harmful stereotypes that emerge in AI chatbot responses in a bid to combat bias in AI.
  • Existing tools that spot stereotypes in models only work with models trained in English, relying on machine translations to localise stereotypes in other languages, which are often inaccurate.
  • SHADES was built using 16 languages from 37 geopolitical regions to counter this issue.
  • Startups are building models that they hope will produce better software and ultimately achieve AGI (artificial general intelligence).
  • The Gates foundation is under threat due to the Trump administration’s massive cuts to foreign aid.
  • Interpolation, rather than generation, could be used to improve AI and produce more accurate results that are not forged.
  • A new era of deepfake fraud is emerging where fraudulent parties manipulate video calls in real time.

By Rhiannon Williams

Original Article