After repeatedly failing to gain traction in the music streaming space, Google's latest attempt is finally here. YouTube Music, first reported last month, is set to launch Tuesday and replace Google Play Music as the company's music subscription service. Some of its features might look awfully familiar to users of Spotify or Apple Music. YouTube Music, for example, borrows its pricing model from Spotify with a free tier to attract new potential subscribers to the platform and a $9.99 per month paid tier that offers ad-free listening, music downloads, and the ability to play music in the background on your mobile device. While both Spotify and Apple Music aim to intelligently learn from your listening habits to recommend songs you'll also like, Google takes this a step further.

"That job [to formalize ... the fundamental elements of human understanding] proved difficult and was never finished. A.I. researchers need to return to that project sooner rather than later, ideally enlisting the help of cognitive psychologists who study the question of how human cognition manages to be endlessly flexible. Today's dominant approach to A.I. has not worked out. If machine learning and big data can't get us any further than a restaurant reservation, even in the hands of the world's most capable A.I. company, it is time to reconsider that strategy.'

Firstly I just want to say I'm a teenager who's interested in Artificial Intelligence so my knowledge is limited/ slightly inaccurate. Sorry if this already exists/has been discussed. Do you see the possibility of deploying artificial intelligence in a simulated world in order to observe how they learn/grow/respond to events around them. Would this be a good way to teach them about the world before deploying them in the real world.

I am trying to reproduced result from the paper "DYNAMIC COATTENTION NETWORKS FOR QUESTION ANSWERING" (https://arxiv.org/abs/1611.01604). I have implemented the code in pytorch but it is overfitting. In the paper it is mention that the authors use dropout for regularization. I added dropout and it helps a bit but not too much. I am also curious if it is possible to get feedback on my model code.

I am by no means a particularly good example of study habits, but generally I tend to read what I need and go from there... Basically this in practice often means starting somewhere relevant to whatever work/assignment/project I'm trying to do, and then going backwards building a recursive stack of readings that seem important to understanding the previous thing until I reach a point where I am familiar with the material already. Then I work through the stack until I'm back to wherever I started. Essentially this is the backward chaining algorithm. I also, if I need to learn a lot from a book for some reason (i.e. a course) or have no particular goal in mind but find my self with a text that piques my interest, then I tend to skim from cover to cover everything that actually attracts my attention, occasionally flipping back to something that I realize is important for understanding later stuff. If it seems especially critical and I can't understand it, then I'll look through exercises and maybe do them if it seems worthwhile.

I am aware of several awesome Python libraries that allow me to do image manipulation, cropping, etc. to prepare my data for training input to Neural Nets. Are you aware of nice looking tools with a graphical interface specifically for the purpose of data preprocessing (resize, crop, brightness) and manipulation (say, assign classes to multiple images easily)? I know every problem is unique and everybody has its own setup, but are there tools out there that do this well?

SANTA CLARA l MUMBAI, April 27, 2018: Digitate, a software unit of Tata Consultancy Services (TCS), (BSE: 532540, NSE: TCS), today announced that its cognitive automation solution, ignio, won an award for the first use of machine learning within the workload automation (WLA) analytics industry from Enterprise Management Associates (EMA). Connecting organizations across a common automation language, WLA builds skills that reduce errors, speed development and remove highly repetitive work, which increases productivity and reliability in defining automated tasks. As WLA continues to evolve, digitization increases and IT accelerates the integration of new technologies such as machine learning. In its quarterly WLA Radar Summary Report, EMA recognized ignio for Batch, underscoring how it keeps business processes running smoothly, reducing the time required to plan and make changes with impact analysis in minutes. Workload automation for batch jobs execution is critical to the proper operations for enterprises that have large numbers of transactions within banking and financial services, retail and manufacturing industries. For services running over 100,000 batch jobs on a daily basis, any failure could lead to business downtime the next day.

Are you a .NET developer? Want to bring machine learning into your applications? ML.NET allows .NET developers to develop their own models and infuse custom ML into their applications without prior expertise in developing or tuning machine learning models, all in .NET. ML.NET was originally developed in Microsoft Research and evolved into a significant framework over the last decade and is used across many product groups in Microsoft like Windows, Bing, PowerPoint, Excel and more. ML.NET runs on Windows, Linux, and macOS - any platform where 64 bit .NET Core or later is available.

In the movies, Captain America is a white guy. But what if he were played by Korean-American actor John Cho? This was tricky to imagine a few months ago, as movie studios remain allergic to casting Asian actors in starring roles. Now, thanks to a new collection of "deepfakes," seeing is believing as a new project illustrates what Asian-Americans might actually look like as the heroes of the biggest Hollywood blockbusters. Digital strategist William Yu is the creator of #SeeAsAmStar (short for "See Asian-American Stars"), a collection of videos and GIFs published on May 4. The title is the goal, as Yu tells Inverse he wants to "break that notion and assert that we can play anything."