Black or white. Good or bad. Too much or too little. To be or not to be. Or maybe.
The human brain insists on all or none, but maybe there is maybe in between.
My random ramblings sometimes uncover an article such as one titled "Analog versus Digital: Extrapolating from Electronics to Neurobiology." (1) Its author was Rahal Sarpeshkar of the Bell Laboratories and MIT. It was published in the journal Neural Computation, which is hardly my regular read. I had had only a slight inkling of the power of this paper. It turned out to be a supreme delight.
First, he rates the performance characteristics of the human brain that he estimates at 3.6x10 to the 15th power operations per second. He estimates that the brain is at least seven orders of magnitude more efficient than the best digital like coprocessor. Despite being made out of meat. He attributes this extraordinary capacity to "clever exploitation of the physics of the medium in which the brain is built, to their local wiring strategies, and to their enormous capability to adapt and learn." He notes that biological systems typically compute constantly rather than episodically, and proceeds to analyze in great depth the defining characteristics of digital and analog physical computers.
This was of course way over my head, but his conclusions resonate. "The advantages of the analog computation rise from its exploitation of physical primitives for computation. The advantages of digital computation arise from its multiwire representation of information and information processing, and from its signal restoration properties." He describes the brain as a system that simultaneously distributes computation in a hybrid fashion which combines the best of analog and digital worlds to create a world that is more efficient than either alone.
In biological systems communication costs are relatively low compared with communication costs in silicon. The optimal signal-to-noise ratio per wire is lower than that in silcon. Nature was smart to distribute computational resources over many noisy neurons which are the dendrites and communicate information between neurons over many noisy axons. The brain is extremely information efficient in its processing and hybrid representations and massively complex system.
"Neurobiology and electronics behave similarly because physical and mathematical laws such as the laws of thermodynamics and the law of large numbers do not change with technologies." The brain is a hybrid suggesting that both analog and digital mechanisms are exploited to a supreme degree. It is therefore not an either or or organ. It is both.
Such an elegant essay is tremendously gratifying to me as it rationalizes the immense complexity of the brain in terms that are knowable.
Aristotle would approve.
Reference:
Sarpeshkar, R. Analog Versus Digital: Extrapolating from Electronics to Neurobiology; Neural Computation 1996,10:1601-1638.
from Healthy Living - The Huffington Post http://www.huffingtonpost.com/walter-m-bortz-ii-md/dare-to-be-100-either-or-_b_6064872.html?utm_hp_ref=healthy-living&ir=Healthy+Living
via IFTTT
The human brain insists on all or none, but maybe there is maybe in between.
My random ramblings sometimes uncover an article such as one titled "Analog versus Digital: Extrapolating from Electronics to Neurobiology." (1) Its author was Rahal Sarpeshkar of the Bell Laboratories and MIT. It was published in the journal Neural Computation, which is hardly my regular read. I had had only a slight inkling of the power of this paper. It turned out to be a supreme delight.
First, he rates the performance characteristics of the human brain that he estimates at 3.6x10 to the 15th power operations per second. He estimates that the brain is at least seven orders of magnitude more efficient than the best digital like coprocessor. Despite being made out of meat. He attributes this extraordinary capacity to "clever exploitation of the physics of the medium in which the brain is built, to their local wiring strategies, and to their enormous capability to adapt and learn." He notes that biological systems typically compute constantly rather than episodically, and proceeds to analyze in great depth the defining characteristics of digital and analog physical computers.
This was of course way over my head, but his conclusions resonate. "The advantages of the analog computation rise from its exploitation of physical primitives for computation. The advantages of digital computation arise from its multiwire representation of information and information processing, and from its signal restoration properties." He describes the brain as a system that simultaneously distributes computation in a hybrid fashion which combines the best of analog and digital worlds to create a world that is more efficient than either alone.
In biological systems communication costs are relatively low compared with communication costs in silicon. The optimal signal-to-noise ratio per wire is lower than that in silcon. Nature was smart to distribute computational resources over many noisy neurons which are the dendrites and communicate information between neurons over many noisy axons. The brain is extremely information efficient in its processing and hybrid representations and massively complex system.
"Neurobiology and electronics behave similarly because physical and mathematical laws such as the laws of thermodynamics and the law of large numbers do not change with technologies." The brain is a hybrid suggesting that both analog and digital mechanisms are exploited to a supreme degree. It is therefore not an either or or organ. It is both.
Such an elegant essay is tremendously gratifying to me as it rationalizes the immense complexity of the brain in terms that are knowable.
Aristotle would approve.
Reference:
Sarpeshkar, R. Analog Versus Digital: Extrapolating from Electronics to Neurobiology; Neural Computation 1996,10:1601-1638.
from Healthy Living - The Huffington Post http://www.huffingtonpost.com/walter-m-bortz-ii-md/dare-to-be-100-either-or-_b_6064872.html?utm_hp_ref=healthy-living&ir=Healthy+Living
via IFTTT
No comments:
Post a Comment