Computers have been beating chess masters for several years now, and just last month a computer dominated the tv quiz show Jeopardy. What's to stop a computer from achieving human-like consciousness?
Let's back up a bit. Computers can do some kinds of things that the human brain can do, like store information and perform rapid calculations, albeit much better. So it is tempting to see the computer as a kind of brain and, reciprocally, the brain as a kind of computer, but ultimately the analogy is misleading. This essay reviews a few elements of brain biology, to point out some key differences between brains and machines.
The brain does indeed perform computations, like a computer. The brain has units (neurons, instead of semiconductors) that, in networks, take in and compile data, operate on the data, and generate output. But brains are far more than fancy calculators. It is no small matter of detail that their circuits are made of biological stuff, not metallic conductors. Their information content is defined not only by the binary logic of a circuit diagram, but also by the dynamic physiologic state of the conductors and the context in which the system operates.
Computers typically perform the same operation repeatedly, exactly the same way, ad infinitum, and whether the end result is the calculation of the fuel consumption of a smart bomb or of the trajectory of a ball in a Wii tennis match. In contrast, the repeated use of a brain circuit alters the performance of the circuit, so the same input on two separate occasions will likely yield a different outcome. This quality is inherent in the biological makeup of the components of the circuit.
Individual neurons in the brain comprise the elements of a vast network. Indeed, the brain is more than a network--it is a network of networks. Some neurons synapse locally with other neurons in the same region; others send out long projections to other parts of the cortex (the outer layers of the brain) or to subcortical structures (clusters of neurons deep within the brain that send projections to and receive projections from the cortex and from other subcortical structures).
The wiring in a simple electrical circuit tends to have fixed conductive properties, under standard environmental conditions. However, the conductive properties of neurons change markedly with use. Without getting too much into the biological detail, heavy activity through a network can induce a physical change in the neurons that comprise that network: after heavy use, the same network may be activated with lighter stimulation than was needed to activate it before the heavy usage.
Computer circuits communicate via electrical charges, or electrons. An electron is an electron is an electron, whether it transmits charge in a computer or a toaster. Neurons in a network communicate via a variety of chemicals called neurotransmitters. There are several neurotransmitter systems, each with a unique distribution through the brain and nervous system, and each serves a unique set of functions.
Some neurons primarily encode information. These include neurons within the cortex and others that connect the cortex to subcortical structures like the hippocampus and basal ganglia. Information-encoding neurons tend to employ neurotransmitters like glutamate that promote depolarization in the receiving neuron and gamma-amino-butyric acid (GABA) that inhibit depolarization. The integration of activating and inhibiting influences on neurons and networks enriches the potential for finely tuned responses. Networks that include glutamate-mediated neurotransmission can alter their properties as described above; that is, they are capable of plasticity.
Other sets of neurons emanate primarily from arcanely named subcortical structures like the substantia nigra, raphe nucleus, and locus coeruleus, and have activating or inhibiting effects on populations of neurons over regions of cortex. These neurons secrete neurotransmitters like serotonin, dopamine, and norepinephrine. These three systems differ in how they radiate to the cortex. Behaviorally, the end result from arousal of dopamine neurons tends to be heightened mobilization for action, whereas norepinephrine sharpens thinking and perception, and serotonin modulates action.
Altered function in these neurotransmitters is associated with symptoms that reflect their influential role in the regulation of cognition, emotions, and behavior. Suicide and aggression, specifically, have been associated with perturbations in serotonin, psychosis and addiction with dopamine, anxiety and mood disorders with norepinephrine.
In contrast to the arousing effects on specific regions of cortex generated by dopamine, serotonin, and norepinephrine, neurons that secrete acetylcholine serve as the messengers to induce arousal in general deep in the brain and across the cortex.
A host of other molecules serves to translate the physiologic state of the body into adaptively useful information. The hypothalamus sits at the center of this chemical network. Its neurons radiate to emit a variety of neurotransmitters to points throughout the brain, in response to changes in hormone levels, nutritional state, stress, and so on.
So one could argue that, although computers can already do repetitive things better than we can, a human brain has the capacity for a rich variety of nuanced responses to the environment, as well as the un-machine-like ability to alter its function in response to experience.