Why the most dangerous AI may be the one we trust too much
Machines don’t have to become superintelligent to cause catastrophic harm. As artificial intelligence spreads through military decision-making, they just have to be wrong – and trusted by the humans in charge.
The Conversation ·