Summary

  • Andrew Barto and Richard Sutton have won the Turing Award for their work in developing reinforcement learning, a technique that has underpinned many of the most pioneering AI programs.
  • In an interview with the Financial Times, the pair expressed concern that companies are not properly testing their AI programs before releasing them to the public.
  • Barto said that releasing software without safeguards is not standard practice in engineering and warned that the rush to be the first to market is leading to poorly tested systems.
  • Their comments echo those of other leading AI figures including Yoshua Bengio and Geoffrey Hinton, who have all warned of the potential dangers of not thoroughly testing AI programs.
  • It follows a statement from a group of senior AI executives, researchers and engineers in 2023 warning that the dangers of AI should be a global priority.

By Jess Weatherbed

Original Article