Via blog.statsbot.co
Spine-Tingling True Stories in AI/ML
Medium’s Stats & Bots blog has compiled a list of “creepy” AI behavior in recent years, from Facebook’s AI systems that infamously developed their own shorthand language to communicate, to the recent facial-cognition study in which a neural network was capable of identifying homosexual individuals’ faces.
While fears of AI often concern far-off apocalyptic dystopias of “killer robots” or god-like superintelligent AI, these anecdotes of less-than-stellar –and yes, even “creepy”– AI behavior prove the point that perhaps AI’s most worrying behavior is not omnipotent god-like malevolence, but instead simple design flaws or unforseen consequences.
Could the greatest danger of AI in the coming years ultimately be that AI systems are too dumb?
