5 UNINTENDED BEHAVIOUR
Can we trust that artificial intelligent agents, trained on data from the wild, will do what they are expected without causing collateral problems? The founder of Cybernetics, Norbert Wiener was concerned about machines taking harmful shortcuts and compared them to the talisman of an old horror story which granted wishes, except that it did so in a “literal minded” way. Today his concerns could become real: we use statistical algorithms for individual risk assessment, and base high-stake decisions on their output. Fortunately, so far, the damage has been contained, but how do we make sure that our machines do not violate fundamental social norms, by doing what they are asked to in a “literal ...
Get The Shortcut now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
