Researcher Creates $20.000 Kit For Artificial Brain
Researcher Andrew Ng has created a building kit costing 20.000 dollar with which someone can built a neural network that functions like the human brain.
Andrew Ng, a researcher at the University of Stanford, worked on the Google Brain project last year. During the project researchers were able to create an artificial neural network using a thousand computers containing 16.000 processors.
The artificial brain network was able to learn to recognize cats in videos. In this way the system mimicked the workings of the human brain. The computer network would be able to help many scientists with, but the findings of the project were not being used by many scientists.
Ng explains in a paper that he thinks this is mainly caused by the high costs of artificial neural network. The Google Brain project would have costed around one million dollar, something the average scientists doesn’t have in its budget. Ng therefor created a similar artificial neural network for the price of just 20.000 dollars.
The building kit Ng created contains cheaper, but more powerful graphical processors, using 16 computers containing 16 Nvidia based processors. Ng acknowledges that building the neural system is not something everyone can do, but he hopes that with the use of his cheaper system more researchers will be able to create their own artificial neural network for research.
To Wired Ng explained that he hopes the less expensive hardware will open new doors. He also said that the possibility to create a network of more than one billion connections for 20.000 dollars can open a new world to science, for instance to improve speech recognition in computer devices.