Artificial Intelligence
| DeepMind has built a what they are calling a "Neural Turing Machine," a neural network that can access an external memory like a conventional Turing machine. The result is a new type of computer that mimics the short-term memory of the human brain. |
| |
DeepMind Technologies, the London-based artificial-intelligence firm acquired by Google earlier this year, has announced that it is designing computers that combine the way ordinary computers work with the way the human brain works. The intention is the machine will not need to be programmed They call the device a Neural Turing Machine (NTM).
In a published study, the authors representing the company, Alex Graves, Greg Wayne and Ivo Danihelka write, "Our experiments demonstrate that it is capable of learning simple algorithms from example data and of using these algorithms to generalize well outside its training regime."
The Neural Turing Machine learns like a conventional neural network using the inputs it receives from the external world but it also learns how to store this information and when to retrieve it.
| Related articles |
Neural networks, have been studied for decades but are receiving renewed attention as computers are now becoming powerful enough to use them effectively. By splitting the processing across a network of artificial "neurons", simple units that process an input and pass it on, these networks are good at learning to recognize pieces of data and classify them into categories.
"Unlike a Turing machine, an NTM is a differentiable computer that can be trained by gradient descent, yielding a practical mechanism for learning programs." |
DeepMind ran two different tests on the Neural Turing Machine. The system was first asked to learn to copy blocks of binary data and then learn to remember and sort lists of data. The results were compared with a more basic neural network, and it was found that the computer learned faster and produced longer blocks of data with fewer errors.
Also, the computer’s methods were found to be very similar to the code a human programmer might have written to perform the same task.
While these are extremely simple tasks for a computer to accomplish when being told to do so, the Neural Turing Machine's abilities to learn them on their own could mean a lot for the future of AI, and might help explain why Google was so willing to acquire DeepMind.
Bridging that gap could give you a computer that does both, and can therefore invent programs for situations it has not seen before. The ultimate goal is a machine with the number-crunching power of a conventional computer that can also learn and adapt like a human.
SOURCE New Scientist
| By 33rd Square | Embed |
0 comments:
Post a Comment