Dubbed ADEPT, the system is able to, like a human being, understand some laws of physics intuitively. It can look at an object in a video, predict how it should act based on what it knows of the laws of physics and then register surprise if what it was looking at subsequently vanishes or teleports. The team behind ADEPT say their model will allow other researchers to create smarter AIs in the future, as well give us a better understanding of how infants understand the world around them. "By the time infants are three months old, they have some notion that objects don't wink in and out of existence, and can't move through each other or teleport," said Kevin A. Smith, one of the researchers that created ADEPT. "We wanted to capture and formalize that knowledge to build infant cognition into artificial-intelligence agents. We're now getting near human-like in the way models can pick apart basic implausible or plausible scenes."

Children learn language by observing the things around them and connecting the dots between what they are seeing and hearing. By learning language like this, they are able to establish a language's word order, such as where subjects and verbs belong in a sentence. In machine learning, languages are learned by training systems on sentences annotated by humans that describe the structure and meaning of words. Gathering that annotation data can be time-consuming and practically impossible for less common languages. Furthermore, not all humans agree on annotations, and annotations might not accurately reflect how people naturally speak.

In one of the big musical numbers from The Life Of Brian, Eric Idle reminds us to "always look on the bright side of life." Norman, a new artificial intelligence project from MIT, doesn't know how to do that. That's because Norman is a psychopath, just like the Hitchcock character that inspired the research team to create him. Like so many of these projects do, the MIT researchers started out by training Norman on freely available data found on the Web. Instead of looking at the usual family-friendly Google Images fare, however, they pointed Norman toward darker imagery.

Facebook volunteers and work-at-home moms might be making city planning decisions, thanks to AI research conducted by MIT scientists. Researchers from MIT's media lab have been feeding computers a steady stream of data for the last four years to build an AI capable of determining why some cities grow and others decay. The data the researchers are using has been compiled from people, regular Joes and Janes, who choose between two randomly selected pictures to determine which one seems less dangerous or more appealing. Currently it's all common-sense driven: most of us would agree a typically beautiful environment will foster growth better than a landscape of derelict buildings. Finally, with enough data, the AI has been returning results -- which have been compared with human responses to the same image pairings.

Swarms of drones flying in terrifyingly perfect formation could be one step closer, thanks to a control algorithm being developed at MIT. The complexities involved in controlling teams of moving robots so they don't crash into each other, or indeed wipe out other objects/entities that cross their path, is a hard problem that continues to keep roboticists busy. But the team of researchers at MIT reckon they have made a breakthrough that could make perfect complex drone formations easier to pull off. They say their decentralized planning algorithm can handle both stationary and moving obstacles, and do so with reduced computational overheads. Why are decentralized control algorithms better than centralized control algorithms?