2011

A LessWrong discussion about a parable of AI learning to distinguish sunny from cloudy days instead of camouflaged tanks from trees, prompting a conversation about real-world examples of similar AI failures and dataset bias. via LessWrong

TechCrunch offered detailed steps for enabling Facebook Timeline early, though the method was intended for developers and may have had bugs.

Robin Hanson, in his scenario, explores a future where immortality is achievable but costly, potentially leading to a significant wealth gap based on merit. via Overcoming Bias.

A new study finds that even being assigned to a made-up group with negative stereotypes can hurt test performance if people identify strongly with that group. via The Wall Street Journal (WSJ)

A LessWrong user believes that a 2011 review article exploring the connection between neuroscience and behavioral game theory might help understand human values. via LessWrong

Freakonomics Radio finds NFL expert predictions to be barely more accurate than random chance, favoring past performance over statistical likelihood. via @freakonomics