Physics, Cognition, Semantic Processing and Gymnastics
Physics, Cognition, Semantic Processing and Gymnastics
David Amerland takes the high view in a post with many ramifications.
In a way this also reflects the complexity, beauty and power of semantic search. Instead of mapping every single relationship associated with every single data point it assigns specific values to the ones it maps, creating a fluid schematic of evolving values that are retained. Instead of starting anew with each similar search query to arrive at the same or a similar answer each time, it uses those retained values to speed things up and conserve processing energy.
#gymnastics #AI #neuralpatterns #cognition #semanticprocessing
Originally shared by David Amerland
Physics, Cognition, Semantic Processing and Gymnastics
In many ways this is an answer to anyone who talks about AI (and the fear if it taking over the world), search (and how it works), cognition (and the limits of the human mind) and, of course, gymnastics (because it is the Olympics after all).
The human brain contains information about the outside world (let's call this real-world knowledge), it also contains information about a parallel, made-up world (let's call that one imagination). The real world and the imaginary one overlap at specific points. The overlaps are where something we dream of can become real, while the distance between the points of the two worlds that do not overlap determines whether something is possible (but not yet attainable) or something will remain forever in the realm of the imaginary where its existence is "real" in the way it is qualified and quantified but bounded specifically by the limitations of its world.
Artificial Intelligence (AI) of course tries to emulate the results achieved by human brains processing information while injecting in the mix its very specific capabilities of speed, tirelessness and the capacity to crunch a lot more data than any human, much faster and much more accurately.
Yet, despite the breadth of its processing bandwidth and its capacity to tirelessly process more information, faster, there are still specific areas where AI cannot venture.
The reason here is energy, or rather the tendency of organic systems to do the almost impossible: conserve energy in ways that make it seem they do not use any energy at all. The case in point in this article is Simone Biles' incredible gymnastics routine but the same principle applies to robots, Google search and intelligence. Organic systems have evolved through trial and error, over millions of years, to recycle and repurpose energy because they work in an energy-competitive environment. Non-organic systems on the other hand, despite their impressive performance rely on a constant, sustainable energy input to function properly.
This may give them an advantage at the initial stages where the extra energy they have access to allows them to boost their performance but as we get into higher and higher levels of complexity where more and more energy is required organic systems can do more with less.
The brain, for instance, has a finite storage space but infinite storing capacity, thanks to the "continuous semantic space" within the frontal cortex which allows information to be turned into knowledge and experience by the constant appraising of new relationships between objects, concepts and attributes. So, the brain is constantly learning, discovering, relearning and forgetting in a complex balance of pertinence, relevance and energy conservation. When Simone Biles performs her complex routine, for instance, she does not have to consciously work out anything. Her brain's heuristic subroutines kick in and the midbrain takes over, allowing something that is incredibly difficult to happen through hardwired muscle memory and learnt instinct. The "values" of each part of her complex aerobatics have already been predetermined. The body performs as it should, activating the blocks of memory that have been created through countless hours of practice. Were she to try and consciously guide herself through it, she would crash and burn as the processing time required would be longer than the routine itself which would make it next to impossible to actually perform.
In a way this also reflects the complexity, beauty and power of semantic search. Instead of mapping every single relationship associated with every single data point it assigns specific values to the ones it maps, creating a fluid schematic of evolving values that are retained. Instead of starting anew with each similar search query to arrive at the same or a similar answer each time, it uses those retained values to speed things up and conserve processing energy.
The defining characteristic of quality in anything is "elegance" as in: "the quality of being pleasingly ingenious and simple."
http://qz.com/754760/simone-biles-vs-a-robot-rio-2016/
David Amerland takes the high view in a post with many ramifications.
In a way this also reflects the complexity, beauty and power of semantic search. Instead of mapping every single relationship associated with every single data point it assigns specific values to the ones it maps, creating a fluid schematic of evolving values that are retained. Instead of starting anew with each similar search query to arrive at the same or a similar answer each time, it uses those retained values to speed things up and conserve processing energy.
#gymnastics #AI #neuralpatterns #cognition #semanticprocessing
Originally shared by David Amerland
Physics, Cognition, Semantic Processing and Gymnastics
In many ways this is an answer to anyone who talks about AI (and the fear if it taking over the world), search (and how it works), cognition (and the limits of the human mind) and, of course, gymnastics (because it is the Olympics after all).
The human brain contains information about the outside world (let's call this real-world knowledge), it also contains information about a parallel, made-up world (let's call that one imagination). The real world and the imaginary one overlap at specific points. The overlaps are where something we dream of can become real, while the distance between the points of the two worlds that do not overlap determines whether something is possible (but not yet attainable) or something will remain forever in the realm of the imaginary where its existence is "real" in the way it is qualified and quantified but bounded specifically by the limitations of its world.
Artificial Intelligence (AI) of course tries to emulate the results achieved by human brains processing information while injecting in the mix its very specific capabilities of speed, tirelessness and the capacity to crunch a lot more data than any human, much faster and much more accurately.
Yet, despite the breadth of its processing bandwidth and its capacity to tirelessly process more information, faster, there are still specific areas where AI cannot venture.
The reason here is energy, or rather the tendency of organic systems to do the almost impossible: conserve energy in ways that make it seem they do not use any energy at all. The case in point in this article is Simone Biles' incredible gymnastics routine but the same principle applies to robots, Google search and intelligence. Organic systems have evolved through trial and error, over millions of years, to recycle and repurpose energy because they work in an energy-competitive environment. Non-organic systems on the other hand, despite their impressive performance rely on a constant, sustainable energy input to function properly.
This may give them an advantage at the initial stages where the extra energy they have access to allows them to boost their performance but as we get into higher and higher levels of complexity where more and more energy is required organic systems can do more with less.
The brain, for instance, has a finite storage space but infinite storing capacity, thanks to the "continuous semantic space" within the frontal cortex which allows information to be turned into knowledge and experience by the constant appraising of new relationships between objects, concepts and attributes. So, the brain is constantly learning, discovering, relearning and forgetting in a complex balance of pertinence, relevance and energy conservation. When Simone Biles performs her complex routine, for instance, she does not have to consciously work out anything. Her brain's heuristic subroutines kick in and the midbrain takes over, allowing something that is incredibly difficult to happen through hardwired muscle memory and learnt instinct. The "values" of each part of her complex aerobatics have already been predetermined. The body performs as it should, activating the blocks of memory that have been created through countless hours of practice. Were she to try and consciously guide herself through it, she would crash and burn as the processing time required would be longer than the routine itself which would make it next to impossible to actually perform.
In a way this also reflects the complexity, beauty and power of semantic search. Instead of mapping every single relationship associated with every single data point it assigns specific values to the ones it maps, creating a fluid schematic of evolving values that are retained. Instead of starting anew with each similar search query to arrive at the same or a similar answer each time, it uses those retained values to speed things up and conserve processing energy.
The defining characteristic of quality in anything is "elegance" as in: "the quality of being pleasingly ingenious and simple."
http://qz.com/754760/simone-biles-vs-a-robot-rio-2016/
She is absolutely incredible. I'm watching the replay of the women's all-around right now.
ReplyDeleteShe won gold. Her teammate Ali Raisman got the silver
ReplyDelete