Skip to main content

AI masters 49 Atari 2600 games without instructions




Artificial intelligence, machines and software with the ability to think for themselves, can be used for a variety of applications ranging from military technology to everyday services like automated telephone systems. However, none of the systems that currently exist exhibits learning abilities that would match the human intelligence. Recently, scientists have wondered whether an artificial agent could be given a tiny bit of human-like intelligence by modeling the algorithm on aspects of the primate neural system.

Using a bio-inspired system architecture, scientists have created a single algorithm that is actually able to develop problem-solving skills when presented with challenges that can stump some humans. And then they immediately put it to use learning a set of classic video games.

Scientists developed the novel agent (they called it the Deep Q-network), one that combined reinforcement learning with what's termed a "deep convolutional network," a layered system of artificial neural networks. Deep-Q is able to understand spatial relationships between different objects in an image, such as distance from one another, in such a sophisticated way that it can actually re-envision the scene from a different viewpoint. This type of system was inspired by early work done on the visual cortex.

Scientists considered tasks in which Deep-Q was able to interact with the environment through a sequence of observations, actions, and rewards, with an ultimate goal of interacting in a way to maximize reward. Reinforcement learning systems sound like a simple approach to developing artificial intelligence—after all, we have all seen that small children are able to learn from their mistakes. Yet when it comes to designing artificial intelligence, it is much trickier to ensure all the components necessary for this type of learning are actually included. As a result, artificial reinforcement learning systems are usually quite unstable.

Here, these scientists addressed previous instability issues in creating Deep-Q. One important mechanism that they specifically added to Deep-Q was “experience replay.” This element allows the system to store visual information about experiences and transitions much like our memory works. For example, if a small child leaves home to go to a playground, he will still remember what home looks like at the playground. If he is running and he trips over a tree root, he will remember that bad outcome and try to avoid tree roots in the future.

Using these abilities, Deep-Q is able to perform reinforcement learning, using rewards to continuously establish visual relationships between objects and actions within the convolution network. Over time, it identifies visual aspects of the environment that would promote good outcomes.

This bio-inspired approach is based on evidence that rewards during perceptual learning may influence the way images and sequences of events or resulting outcomes are processed within the primate visual cortex. Additionally, evidence suggests that in the mammalian brain, the hippocampus may actually support the physical realization of the processes involved in the “experience replay” algorithm.

Read More
Labels:

Comments

Post a Comment

Popular posts from this blog

How To Hide Text In Microsoft Word 2007, Reveal It & Protect It

Sometimes what we hide is more important than what we reveal. Especially, documents with sensitive information, some things are supposed to be ‘for some eyes only’. Such scenarios are quite common, even for the more un-secretive among us. You want to show someone a letter composed in MS Word, but want to keep some of the content private; or it’s an official letter with some part of it having critical data. As important as these two are, the most common use could involve a normal printing job. Many a time we have to print different versions of a document, one copy for one set of eyes and others for other sets. Rather than creating multiple copies and therefore multiple printing jobs, what if we could just do it from the same document?  That too, without the hassle of repeated cut and paste. We can, with a simple feature in MS Word – it’s just called Hidden and let me show you how to use it to hide text in Microsoft Word 2007. It’s a simple single click process. Open the docum...
1 comment
Read more

Build Your Own Awesome Personal 3D Avatar with Avatara

Do you use social networks and want to build your own awesome 3D avatar? Maybe you want to send someone a cute cuddly image of yourself (kind of)? Or maybe you have your own ideas of what you would do with an Avatar… Well look no further than Avatara which I discovered from the MakeUseOf directory . You can create 3d avatars out of pre-set up templates or create your own from scratch. To start, visit Avatara’s homepage . You will see this screen: Click Get Started to umm, get started! That will take you to this screen: You see that you can build your own Avatar using an uploaded head shot like the Obama one above (just an example, guys). Or roll with one of their awesome avatars. I chose to start with a blank avatar by clicking Start with a blank avatar at the bottom of the screen. That takes you to here: I clicked on the filter at the top and told it to filter out everything but male characters and then I saw this: I rolled with Buck and continued. You need to click Select...
1 comment
Read more

Ex-Skypers Launch Virtual Whiteboard Deekit

Although seriously long in the tooth and being disrupted by a plethora of startups, for many years Skype has existed as an almost ubiquitous app in any remote team’s toolkit. So it seems apt that a new startup founded by a team of ex-Skype employees is set to tackle another aspect of online collaboration. Deekit, which exits private beta today, is a virtual and collaborative whiteboard to help remote teams work smarter. The Tallinn, Estonia-based startup is headed up by founder and CEO, Kaili Kleemeier, who was previously a Head of Operations at Skype. She and three colleagues quit the Internet calling giant in 2012 and spent a year researching ideas in the remote team space. They ended up focusing on creating a new virtual whiteboard, born out of Kleemeier’s experience collaborating with technical teams remotely, specifically helping Skype deal with incident management. “Working with remote teams has been a challenge in many ways – cultural differences, language differences, a...
Post a Comment
Read more