Backed by U.S Defense funding, IBM has recently announced the world’s first “synaptic” processor project: it’s a chip that which mimics the human brain architecture. The company calls this an unprecedented progress in building a machine that will (try to) learn, and understand – two things that computers have not been so great at so far.
In theory, the IBM chip is capable of re-wiring itself (like our brains) to adapt to certain situations or certain types of computations. At the moment, IBM has managed to perform some basic operations like solving mazes or recognizing mildly complex patterns. However we are still far from what “intelligence” commonly means to people.
Yet, imitating a small portion of the brain may help cracking how it works, says IBM. The chip is currently using 256 silicon neurons. In comparison, the human brain can have one hundred Billion neurons. This is great research, but it is my personal opinion that it is presented to the public in a slightly optimistic/inflated way.
Is it really a hardware problem?
Since the beginning, the main issue with artificial intelligence (AI) is that we just don’t know how it works, which is why building an artificial version has been so elusive. Even from the medical perspective, we only have a relatively superficial knowledge of how the brain functions. Today, we can make very narrow forms of “intelligence”, most of which are more “specialized skills” (like chess, financial analysis, etc…) than “intelligence”. Artificial intelligence is not a hardware problem, it’s a software problem.
Put simply, even with all the computing power available in the world today, we could not make AI what we want it to be, yet.