Will Artificial Intelligence Ever Surpass the Human Brain?

Will Artificial Intelligence Ever Surpass the Human Brain?

By Giorgi Kharebava

A number of artificial intelligence (AI) enthusiasts have been predicting that machines will eventually surpass human intelligence in the foreseeable future. In some cases, when tasks are easily broken down into sequential, smaller arithmetical components, computers outperform the human brain, reflecting the lack of adaptation of the later to such problems. Despite this, the current overall consensus in the field is that computers still need to evolve significantly to match the efficiency and complexity of the brain. Unlike computer power, which is very well defined, we are still far from understanding how to express the processing power of the human brain.

At the fundamental level, the intelligence of machines, as well as the cognition of the brain is a based on the storage of information and electrical currents conducted through highly specialized and sophisticated architecture of wires. In machines, these signals flow at the speed of light while in the brain, the velocity of impulse conduction in the axons is between 0.2-120 m/s. Supercomputers have approached roughly the estimated computing power of the brain, several petaflops, but at an enormous cost: about 10 million watts of power consumption versus 20 watts in the brain.

Why the Brain Has the Upper Hand

The brain easily keeps the current lead in intelligence over machines for a number of reasons. First, it has the ability to store and process the information within the same units, neurons, and their synapses. Second, apart from the superior architectural design, the brain clearly holds the advantage in the numbers of the cores if neurons are assumed for the comparative role. Advanced supercomputers have up to 10 million cores, while the brain features nearly 100 billion neurons.

From Neuroscience To Neuromorphic Engineering

In computer science, significant research is directed to creating new computing units modeled according to neuronal function. This direction is referred to as neuromorphic engineering. In neuroscience, most efforts are directed towards understanding, as well as preventing age and disease-induced deterioration of brain function. Relatively small efforts are put to research for enhancing overall processing power and functioning of the normal human nervous system. Enhancing human brain power by interfering with the basic functional parameters, may provide the sufficient counterweight to the “existential risks” posed by the rise of AI.

In the developed brain, significant improvements to architecture will be nearly impossible to implement in the near future. However, temporary or even permanent improvement to the brain’s processing speed could be a much easier reach for current neuroscience. The cognitive power of the brain, in its significant parts, is a reflection of two processes: impulse conduction in the axon and synaptic transmission. The speed of these functions is the key to a better processing power and is highly variable in the brain. Maximizing or even enhancing these parameters through molecular manipulations may significantly boost overall processing speeds, hence cognitive function.

The Link Between Cognitive Function and Polyunsaturated Fats

Regulation of the impulse conduction velocity or the speed of synapse function in the brain has been intensively studied but is still not well understood. Nevertheless, it has been clearly established that the axonal conduction velocity is proportional to the axonal diameter. In the cortex, axonal diameters vary from 0.3 to 20 microns but maintain a target speed of <5ms.

At the molecular level, the axonal diameter is regulated by neurofilaments and their modifications. The expression of these cytoskeletal proteins is precisely regulated during development, but the mechanisms and triggers of this process have not been described yet. It is very well possible that understanding on how neurofilaments control the axonal diameter and how their expression can be manipulated will hold the key to the regulation of axonal diameter, hence axonal conduction velocity. Recently, it has been shown that dietary polyunsaturated fats, such as docosahexaenoic acid and its derivative ethanolamine regulate axonal growth and neurofilament expression, as well as modifications of these proteins in the axon. Polyunsaturated fatty acids are well known to be critical for the brain development, to enhance cognitive function and thwart decline in elderly brains. Nevertheless, there is very little knowledge how these dietary components mediate their effects. The same group has shown previously that polyunsaturated fats of the omega-3 class enhance synapse formation and function.

Based on all these findings it is evident that polyunsaturates are used by the brain to regulate most basic components that determine processing speed and computing power of the brain. It is also very clear that dietary ingestion of the large amounts of the omega-3 class of fatty acids will not immediately enhance the processing power of the brain to the desirable levels. The reason for such situation is the numerous feedback regulatory mechanisms keep tight control and re-purpose all the signals if necessary, for establishing optimal function (including speed) in the brain. This “optimal function” determined by the evolution is one of the options out of many possible states. Hence, understanding the related mechanisms would open up the possibilities for the intervention, modification, and improvements for the brain.

In conclusion, the message in this article is that designing intelligent machines to make our lives easier is a very legitimate direction but designing enhancements to our own intelligence is one of the most important directions that is currently overlooked in research. This direction, if successful, has the potential to accelerate progress in all other areas of life, including the design of artificial intelligence.

References:

1. How powerful is the human brain compared to a computer?

2. China builds world’s most powerful computer

3. Reducing Long-Term Catastrophic Risks from Artificial Intelligence – Machine Intelligence Research Institute.

4. Stephen Hawking, Elon Musk, and Bill Gates Warn About Artificial Intelligence.

5. Waxman, H.A.S. and G. Stephen, Axonal conduction delays. Scholarpedia, 2012. 7(6): p. 1451.

6. Cao, D., et al., Docosahexaenoic acid promotes hippocampal neuronal development and synaptic function. J Neurochem, 2009. 111(2): p. 510-21.

Image courtesy of pixabay.com.

Learn more about PreScouter at www.prescouter.com.

 

More of Our Insights & Work