People platform to discuss and share, research, ideas, #deeplearning methods and applications.

Once the stuff of science fiction novels, machine learning--where computers improve automatically through experience--is now attracting the attention of a wide range of industries. As with many recent advances in tech, machine learning's growth has been largely fueled by the development of new learning algorithms and theory, and by the ongoing explosion in the availability of online data and low-cost computation. Machines are better equipped than ever to capture and analyze large quantities of multisourced, ever-changing data. But analyzing data is not the same thing as using it to make decisions--and here is where humans come in. Humans are still needed to innovate, to put ideas in an appropriate context, and to suss out an action's wide-ranging implications.

Spoiler alert: If you are yet to watch the Season 5 finale of "Game of Thrones" and have also managed to avoid a year of headlines about the fate of one particular character, read on at your own peril. For the rest of us, we've been subject to months of speculation that Jon Snow may not be completely dead, despite being run through with cold, hard steel by most of the Night's Watch when we saw him last. Now, a machine learning algorithm designed by a team at the Technical University of Munich has analyzed data on all the characters in Westeros, both dead and alive, and concluded that it is very likely that Snow is actually a survivor. The project, dubbed "A Song of Ice and Data", basically scrapes info from the online Wiki of Ice and Fire encyclopedia, which focuses largely on the series of books by George R.R. Martin, but also covers the HBO show they inspired. Using this data source, two dozen features of each character are statistically compared to try and figure out which features make a character most likely to die.

Smack in the center of Palo Alto, Calif., sits a huge warehouse featuring three-story ceilings and at least 15,000 square feet of open space. Standing as it does in the very zip code that gave birth to Google, Facebook, and HP, this building represents some of the most valuable real estate on the planet. Inside, a platoon of workers bend metal and install soundproof glass, readying the structure for its rebirth. If Andy Rubin and his backers have their way, this former apricot canning facility will become ground zero for a massive shift in how society and business understand not just data, computing, and the Internet, but the very workings of the world around us. Back in 2016, we'll tell our kids, we didn't have the actual Internet. Call it the actuated Internet -- a virtuous circle of real world objects, at-scale artificial intelligence, and command and control that animates everything of value in our lives.

Note: This is the 6th part of a short essay series aiming to condense knowledge on the Artificial Intelligence Revolution. Feel free to start reading here or go to Part 1. The project is based on the two-part essay AI Revolution by Tim Urban of Wait But Why. I recreated all images, shortened it x3 and added a couple of new perspectives. Read more on why/how I wrote it here.

Artificial intelligence has had its share of ups and downs recently. In what was widely seen as a key milestone for artificial intelligence (AI) researchers, one system beat a former world champion at a mind-bendingly intricate board game. But then, just a week later, a "chatbot" that was designed to learn from its interactions with humans on Twitter had a highly public racist meltdown on the social networking site. How did this happen, and what does it mean for the dynamic field of AI? In early March, a Google-made artificial intelligence system beat former world champ Lee Sedol four matches to one at an ancient Chinese game, called Go, that is considered more complex than chess, which was previously used as a benchmark to assess progress in machine intelligence.

In a new paper, researchers from CSAIL and the machine-learning start-up PatternEx have demonstrated an artificial-intelligence platform called "AI²" that can predict 85% of cyber-attacks, by continuously incorporating input from human experts. To predict attacks, AI² combs through data and detects suspicious activity by clustering the data into meaningful patterns using unsupervised machine-learning. It then presents this activity to human analysts who confirm which events are actual attacks, and incorporates that feedback into its models for the next set of data.

A new artificial intelligence (AI) system being developed at MIT's Computer Science and Artificial Intelligence Laboratory is being trained by researchers to aid humans in identifying potential cyber-attacks. Typically, when trying to pinpoint possible attacks, analysts are required to sift through massive amounts of data to find abnormalities and discrepancies--a method that is time-consuming and tedious. Anchored on the idea that AI never gets tired, the new computer based method means that humans can identify cyber-attacks more efficiently. AI2 for instance--MIT's new system, which honed its ability to identify threats after reviewing three months worth of log data from an unidentified ecommerce platform--can review millions of log lines every day. Once it spots something suspicious, a human can then take over and promptly check for possible signs of a security breach.

We are currently experiencing a fundamental change in the way we live and work. It is definitely possible, that in the near future, millions of jobs are replaced by technology. This holds also true for the insurance industry. Low-level processing of claims and some standardized underwriting is automated, and it is expected that more will follow. With a significant part of an insurer's cost structure coming from human resources, there is an increasing need to shift to automation in order to deliver significant savings.

I've been interested in artificial intelligence (AI) – the concept that humans can design a machine that thinks for itself without explicitly provide instructions (or a "program") -- since the early 1970s. As an undergraduate math major, I wrote a program designed to learn to play the game of Monopoly (checkers was too easy, chess was too hard). The program "knew" the objective, the layout of the board, the rules of play, and the content of the three stacks of cards that drive the game. It could handle up to 8 players. It wasn't very smart, but it had one great advantage -- it could play thousands of games against itself, remember the outcomes, and analyze why it had won (someone always wins) or lost using some simple algorithms. After a few weeks it was virtually unbeatable against one or more humans.