>>10863492Artificial Intelligence.
Deep Learning is irreproducible. The scientific method is violated. Data are cherry-picked. Improvement are mainly due to more accessibility to data by multinational corporations.
Deep Reinforcement Learning is basically a scam at this point. Most methods pre-2014 was published without even an error bar. A slight change of hyper-parameters or even random seed make everything crash. Researchers purposely implement other methods wrong and pick the best run a few hundred to publish.
Don't fall for the AI meme like I did /sci/, just don't.