In college, it wasn’t rare to hear a verbal battle regarding artificial intelligence erupt between my friends studying neuroscience and my friends studying computer science.
One rather outrageous fellow would mention the possibility of a computer takeover, and off they went. The neuroscience-savvy would awe at the potential of such hybrid technology as the CS majors argued we have nothing to fear, as computers will always need a programmer to tell them what to do.
Today’s news brings us to the Neural Turing Machine, a computer that will combine the way ordinary computers work with the way the human brain learns, enabling it to actually program itself. Perhaps my CS friends should reevaluate their position?
The computer is currently being developed by the London-based DeepMind Technologies, an artificial intelligence firm that was acquired by Google earlier this year. Neural networks — which will enable the computer to invent programs for situations it has not seen before — will make up half of the computer’s architecture. Experts at the firm hope this will equip the machine with the means to create like a human, but still with the number-crunching power of a computer, New Scientist reports.
In two different tests, the NTM was asked to 1) learn to copy blocks of binary data and 2) 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. Additionally, the computer’s methods were found to be very similar to the code a human programmer would’ve written to make the computer complete such a task.
These are extremely simple tasks for a computer to accomplish when being told to do so, but computers’ abilities to learn them on their own could mean a lot for the future of AI.
Elon Musk is not going to be happy about this.