From the department of “the AI made me do it”
a deep learning system called RankBrain now plays a role in “a very large fraction
” of the millions of queries that go through Google each second of each day. Problem is, humans rarely completely understand
how neural nets make their decisions…
“…it’s unclear how the move will affect Google’s ability to defend its search results against claims of unfairness or change the results in the face of complaints.”
A slightly terrifying article
and Google Research paper
on technical debt in machine learning
. And one about MIT using machine learning to fix bugs
in code. All we need now is machine learning to fix the errors in machine learning?
A mysterious bot on MakerBot’s Thingiverse seems to be smashing 3D models
together and re-uploading them. The end-result is both weird and wonderful
An AI has beaten humans at sketch-recognition
. Its creators suggest their program could be adapted to help police match drawings of suspects to mug shots.
6.0 Amazon’s Alexa seems to be striking a chord in a way that Siri herself may start to envy. We’ve been collecting examples…
“My wife said Alexa had to go. She was making the kids rude. More specifically, they were being impolite by not using ‘Please’ and ‘Thank You’ when addressing her. My wife didn’t want that sort of behaviour encouraged.”
“…(my mom) asks Alexa for jokes, to play soft music, wake her up…the last time she traveled out of town to see my sisters, she packed Alexa up and took her too”
Research suggests robots could learn human values
by reading ‘stories’
about often mundane tasks like going to the pharmacy. When faced with a wait, will he know the right
thing to do?
“Under ideal circumstances, Quixote [the prototype] never performs actions that would be considered psychotic, harmful, or antisocial. This is significant because we never told Quixote what is right or wrong. We can make the system ‘fail’ by scrambling its understanding of the stories, in which case sometimes it will do antisocial behaviours like stealing.”
Always sobering to think that we’re no doubt already actively training the future brains of these things on a daily basis.