News
SCIENCE | Artificial intelligence faces reproducibility crisis
![](/newsroom/files/newsroom/styles/fullwidth_breakpoints_theme_moriarty_small_1x/public/channels/image/science.png?itok=4bZoSQK3×tamp=1519063715)
Published: 19 February 2018
At the AAAI meeting, Peter Henderson, a computer scientist at 苹果淫院 in Montreal, showed that the performance of AIs designed to learn by trial and error is highly sensitive not only to the exact code used, but also to the random numbers generated to kick off training, and to 鈥渉yperparameters鈥濃攕ettings that are not core to the algorithm but that affect how quickly it learns.