2015 : WHAT DO YOU THINK ABOUT MACHINES THAT THINK?

leo_m_chalupa's picture
Neurobiologist; Professor of Pharmacology and Physiology, George Washington University
No Machine Has Ever Thought About The Eternal Questions

Recent demonstrations of the prowess of high performance computers are remarkable, but unsurprising. With proper programming machines are far superior to humans in storing and assessing vast quantities of data and in making virtually instantaneous decisions. These are machines that think because similar processes are involved in much of human thought.

But in a broader sense, the term thinking machine is a misnomer. No machine has ever thought about the eternal questions: where did I come from, why am I here and where am I going? Machines do not think about their future, ultimate demise or their legacy. To ponder such questions requires consciousness and a sense of self. Thinking machines do not have these attributes, and given the current state of our knowledge it's unlikely that they will attain them in the foreseeable future.

The only viable approach to construct a machine that has the attributes of the human brain is to copy the neuronal circuits underlying thinking. Indeed, research programs now ongoing at UC Berkeley, MIT and several other universities are focused on achieving this precise objective. These programs are striving to build computers that function like the cerebral cortex.
Recent advances in our understanding of cortical micro circuitry have propelled this work, and it is likely that the recent White House brain initiative will provide a wealth of valuable additional information. In the coming decades we will know how the billions of neurons in each of the 6 layers of the cerebral cortex are interconnected as well as the types of functional circuits that these connections form.

This is a much-needed first step in designing machines capable of thinking in a manner equivalent to the human brain. But understanding the cortical micro circuitry is not sufficient in constructing a machine that thinks. What is required is an understanding of the neuronal activity underlying the thinking process. Imaging studies have revealed much new information of the brain regions involved in processes functions, such as vision, hearing, touch, fear, pleasure and many others.

But as yet we don't have even a preliminary understanding of what takes place when we are in thought. There are many reasons for this, not the least of which is our inability to isolate the thinking process from other bodily states. Moreover, it may well be the case that different brain circuits are engaged in different modes of thinking. Thinking about an upcoming lecture would be expected to activate the brain differently than thinking about unpaid bills.

In the near term, we can expect computers will do more and more things better than humans. But a far better understanding of the workings of the human brain is needed to create a machine that thinks in a way equivalent to human thought. For now, we don't need to be concerned with civil or any other rights of machines that think; nor do we have to be concerned with thinking machines taking over society. If things should get out of hand, just pull the plug.