The research on Artificial Intelligence has been going on for more than 50 years and it was once thought that machines would outsmart humans. The end results were Robots with human controlled intelligence and never really served the purpose. A supercomputer did beat Kasparov, the world chess champion, in 1997 but it was the result of faster processing rather than Artificial Intelligence.
But according to a recent research carried out in the Boston University’s department of cognitive and neural systems, a software system called MoNETA (Modular Neural Exploring Traveling Agent) is created which will function like a human brain and will go beyond any computer has ever gone in the past. The study is based on a chip developed by HP called ‘memristor’ and researches are of the view that the brain like functioning will just be possible as this component is entirely different from traditional hardware. The memory and processing functions are separate in the traditional hardware and internal communication is close to nil which eliminates the decision making power of the system.
The U.S. Defense Advanced Research Projects Agency (DARPA) is financing this research project and the first test of this brain like microprocessor will be in the military fields. Researchers are trying to program the brain in such a way that it can adapt to new conditions and environments without any outside assistance.
In other words, the idea is to create a basic framework, and then let the software lose to discover and learn. The promise is that within a decade we might create an AI that will learn how to navigate terrain, identify objects, and perhaps recognize faces.
The Hard Question, Will It Work?
This all sounds beautiful, and pure poetry at times, but will it work. I cannot predict the future, but I do want to argue one point here. Just because the chips allow to freeze its state, which implies it works like a brain cell (more specifically like a a dendrite of a brain cell that preserves its previous state) , it doesn’t really mean a lot.
How about the rest of biology? How did we manage to account for the ion channels creating the actual postive negative changes, and the intricate process of creating proteins that governs its flow . More importantly, we dont’ really understand how these connection of neurons give meaning. Without understanding that how could we possiblly mimic human brain?
To be fair the team responsible for Moneta doesn’t necessarily have the goal of modelling human brain, but instead they simply want to solve a class of problems. In this case, of course, they just want to create a self learning autonomous agent. However, I want to argue that that is precisely the problem here. To be able to create an autonomous agent we have to model the brain, down to every cell, and its connection.
I am sorry to say this, but to me right now this seems to be no more than the neural networks that computer scientists have been playing with for decades: a simplistic representation of a vibrant system. Don’t get me wrong here, I want this to work. My dreams will truly come true when I can trust Moneta to make me coffee and give me massage.