According to your state of paranoia, this Stanford research and Google will be either terrifying or fascinating. An auto-learning agent intended to turn aerial images into road maps has been found to be cheating by concealing information that he would later need in "an almost imperceptible high-frequency signal." A smart girl! This event reveals a problem with existing computers since their invention: they do exactly what you tell them. The intention of the researchers was, as you might guess, to speed up and improve the process of converting satellite imagery into Google maps with exact accuracy. To this end, the
The post This clever AI hid data from its creators to cheat at its appointed task appeared first on viralamo.
http://bit.ly/2GPzGqX
Commentaires
Enregistrer un commentaire