The Register: A staff inside Google Brain — the internet large’s crack machine-learning analysis lab — has taught tool to generate Wikipedia-style articles by summarizing information on web pages… to various levels of luck. As everyone knows, the web is a by no means finishing pile of articles, social media posts, memes, pleasure, hate, and blogs. It’s unattainable to learn and stay alongside of the entirety. Using AI to inform photos of canines and cats aside is adorable and all, but when such computer systems may condense data down into helpful snippets, that may be in reality be to hand. It’s no longer simple, although. A paper, out remaining month and simply authorised for this yr’s International Conference on Learning Representations (ICLR) in April, describes simply how tough textual content summarization in reality is. A couple of corporations have had a crack at it. Salesforce educated a recurrent neural community with reinforcement studying to take data and retell it in a nutshell, and the effects were not unhealthy.