6 ms·

I've worked with high level machine intelligence, it has become very obvious to me that cognitive development goes hand in hand with physical movement and manip

I've worked with high level machine intelligence, it has become very obvious to me that cognitive development goes hand in hand with physical movement and manipulation. Intelligence, real or artificial, do not fully develop without prolonged physical interaction with the subjects they proport to understand. We learn about space, objects, and people by doing regular physical activity. We learn about things by touching them, throwing them, kicking them into the woods or on the roof. Our brains develop by manipulating and getting feedback from our environment. I feel far less intelligent after a day spent in front of a computer or a book than I do after a day spent walking in the woods. I'm astonished by how brain-dead I become without physical activity, and how useless and un-creative it makes me. Kids don't need 30 min of recess -- they need an entire day of working and learning with their hands and feet along with their brains.

jostylr 11y ago
I work for a Sudbury School (Arts&Ideas in Baltimore) where this is our mantra. I've long take it as an article of faith that humans are wired this way. But I am astounded by the comment about artificial intelligence. Is there any kind of paper that explores this? Not even sure how to set it up or quantify such a thing, but that would be incredible to prove.
- hellbanner 11y ago
  http://www.psychology.emory.edu/cognition/barsalou/papers/Barsalou_BBS_1999_perceptual_symbol_systems.pdf http://www.psychology.emory.edu/cognition/barsalou/papers/Ba... "Abstract: Prior to the twentieth century, theories of knowledge were inherently perceptual. Since then, developments in logic, statistics, and programming languages have inspired amodal theories that rest on principles fundamentally different from those underlying perception. In addition, perceptual approaches have become widely viewed as untenable because they are assumed to implement recording systems, not conceptual systems. A perceptual theory of knowledge is developed here in the context of current cognitive science and neuroscience. During perceptual experience, association areas in the brain capture bottom-up patterns of activation in sensory-motor areas. Later, in a top-down manner, association areas partially reactivate sensory-motor areas to implement perceptual symbols. The storage and reactivation of perceptual symbols operates at the level of perceptual components – not at the level of holistic perceptual experiences. Through the use of selective attention, schematic representations of perceptual components are extracted from experience and stored in memory (e.g., individual memories of green, purr, hot). As memories of the same component become organized around a common frame, they implement a simulator that produces limitless simulations of the component (e.g., simulations of purr). Not only do such simulators develop for aspects of sensory experience, they also develop for aspects of proprioception (e.g., lift, run) and introspection (e.g., compare, memory, happy, hungry). Once established, these simulators implement a basic conceptual system that represents types, supports categorization, and produces categorical inferences. These simulators further support productivity, propositions, and abstract concepts, thereby implementing a fully functional conceptual system. Productivity results from integrating simulators combinatorially and recursively to produce complex simulations. Propositions result from binding simulators to perceived individuals to represent type-token relations. Abstract concepts are grounded in complex simulations of combined physical and introspective events. Thus, a perceptual theory of knowledge can implement a fully functional conceptual system while avoiding problems associated with amodal symbol systems. Implications for cognition, neuroscience, evolution, development, and artificial intelligence are explored"
  - nitrogen 11y ago
    Fascinating abstract. I think I may have accidentally downvoted your comment though, when I meant to upvote (on a phone). I downloaded the PDF; hopefully I can find time to read it eventually. There are some gaps between the layers of abstraction presented in this abstract that I hope would be addressed by the full paper.
    - hellbanner 11y ago
      Their main point was that pre-computers, cognition was theorized to be tied to physical stimulus. When humans learned how to code computers, "AI" was theorized as existing in "black boxes" of pure abstract symbol manipulation. The paper discusses how this doesn't make sense, and how cognition must be tied to biological sensing to be "smart".
- dchichkov 11y ago
  I doubt that "physical manipulation" is a requirement for machine intelligence. It may as well be any kind of manipulation in any kind of reach and consistent environment. It could potentially be rich enough virtual environment. But, in practice, the only "rich enough" environment currently available is the physical world. And virtual classroom is no substitution...
- sdg1 11y ago
  There's a somewhat relevant concept in Seymour Papert's book Mindstorms. Papert talks about "body-syntonicity" where the act of drawing a circle on the computer screen with LOGO can be seen as bridging children’s sense and knowledge about their own bodies to something abstract. Young children would often pretend to be the LOGO turtle, acting out commands on the floor, during stages of planning a LOGO program, or while debugging LOGO code. It has also been a strategy for teachers to introduce programming and procedural ways of thinking. This video shows this in action (with some followup comments from Papert): https://youtu.be/bOf4EMN6-XA?t=254 https://youtu.be/bOf4EMN6-XA?t=254
  - rurounijones 11y ago
    When I was in school with the BBC micro and the LOGO turtle we had an actual physical turtle hooked up to the machine via a long cable. Needless to say this was much more popular with us than just the screen. I am sure it is partly for the robotic'ness of it but we would often walk around with the turtle as it was moving. So what you are saying and linked to I can well believe.
    - juanuys 11y ago
      This sounds great. Do you have any more information on the mechanical turtle? (consumer kit, or a side-project by an industrious teacher?)
      - obitoo 11y ago
        Here he is , as i recall it was a kit http://www.classicacorn.freeuk.com/8bit_focus/logo/logo.html http://www.classicacorn.freeuk.com/8bit_focus/logo/logo.html
      - rurounijones 11y ago
        This was a consumer kit aimed at schools. Looks like there were many different types as obitoo has another that I have never seen. http://gallery.nen.gov.uk/asset78077_1582-.html http://gallery.nen.gov.uk/asset78077_1582-.html
- rsaarelm 11y ago
  There is the theory of general causality developed by Judea Pearl, which as far as I've understood implies that you need to be able to purposefully affect the world you're observing in some way to do experiments before you are able to infer causal relationships even in theory. See eg. http://www.michaelnielsen.org/ddi/if-correlation-doesnt-imply-causation-then-what-does/ http://www.michaelnielsen.org/ddi/if-correlation-doesnt-impl...
LiweiZ 11y ago
It seems the way of our brain works have some similarities with machines(well, it perhaps among the best average person at this time is able to perceive.). The brain gets information from sensors, tries to categorize everything and makes sense by use the categories in mind. New information from the sensors may trigger the brain to recategorize. All those steps make sense. I believe that's one of important reasons screen based devices are not good for little kids. They need to get exposed to the physical world since they are still learning how to interact with the physical world they live in. Leaving kids to screen is a low cost and time/energy-saving solution, but it increases the risk of "less-developed-focusing-muscle".
grubles 11y ago
Montessori-based education revolves around learning and working with physicality[0]: "Exploration, Manipulation (of the environment), Activity, Order, Orientation" etc. Disclosure: I was put into a Montessori school when I was young. [0]https://en.wikipedia.org/wiki/Montessori_education https://en.wikipedia.org/wiki/Montessori_education
javajosh 11y ago
So is the solution to focus on simulating a body within a fairly detailed simulated reality? Forget how you model emotions (which might be emergent from patterns of subtle physical sensations) how do you do prioperception in that context? Also, do you cheat and allow the entities "imagination" to be a perfectly modifiable simulation, or do you allow "errors" to enter into the simulation? What about the ethics of maintaining the integrity of the events and reactions that compose the entity, and the possible responsibility to ensure that reactions continue, or that those events have something to do with our "real" world and are not entirely simulated?
deckiedan 11y ago
'Society of the Mind'[1] by Eric L Harry is a great sci-fi read about AI and has this as a major part of the premise. I read it a few years ago and recently bought a copy, re-read it. Still excellent. [1] http://www.amazon.co.uk/gp/product/B00SRV9UGE/ref=as_li_tl?ie=UTF8&camp=1634&creative=19450&creativeASIN=B00SRV9UGE&linkCode=as2&tag=madswor09-21 http://www.amazon.co.uk/gp/product/B00SRV9UGE/ref=as_li_tl?i...
geographomics 11y ago
Related to this, an interesting read on the philosophy of embodied cognition: http://plato.stanford.edu/entries/embodied-cognition/ http://plato.stanford.edu/entries/embodied-cognition/
arethuza 11y ago
It's been a while since I worked in AI research but although I worked in agent architecture and model based reasoning I was came to believe in the approach of "the world is its own best model" associated with Rodney Brooks: https://en.wikipedia.org/wiki/Subsumption_architecture https://en.wikipedia.org/wiki/Subsumption_architecture i.e. Why put so much effort into modelling the world to allow reasoning when you can interact with the world to answer the same questions.
Kenji 11y ago
I feel far less intelligent after a day spent in front of a computer or a book than I do after a day spent walking in the woods. Odd, for me it's the other way around. I may feel more at peace, more relaxed and better able to fall asleep when I was in the wood, but I definitely don't feel as intellectually enriched as I do after reading a book or on the PC.
thesz 11y ago
Three main stresses that human body learn to deal with during the course of evolution trigger development of brain (and nervous system in general). These stresses are endurance activity (walk long hours), strength activity (fight or flight response) and famine. The agent responsible for development of nervous system in these cases is brain-derived neurotrophic factor. It is responsible for creation of new neurons and branching of old ones and directly related to intellectual performance. So your link to high level machine intelligence is wrong, I believe. I also think I explained above why do you feel smarter after a day of physical activity.
- skj 11y ago
  > So your link to high level machine intelligence is wrong, I believe. Alternate explanations are not refutations. Both could be factors.

jostylr 11y ago

I work for a Sudbury School (Arts&Ideas in Baltimore) where this is our mantra. I've long take it as an article of faith that humans are wired this way. But I am astounded by the comment about artificial intelligence. Is there any kind of paper that explores this? Not even sure how to set it up or quantify such a thing, but that would be incredible to prove.

hellbanner 11y ago
http://www.psychology.emory.edu/cognition/barsalou/papers/Barsalou_BBS_1999_perceptual_symbol_systems.pdf http://www.psychology.emory.edu/cognition/barsalou/papers/Ba... "Abstract: Prior to the twentieth century, theories of knowledge were inherently perceptual. Since then, developments in logic, statistics, and programming languages have inspired amodal theories that rest on principles fundamentally different from those underlying perception. In addition, perceptual approaches have become widely viewed as untenable because they are assumed to implement recording systems, not conceptual systems. A perceptual theory of knowledge is developed here in the context of current cognitive science and neuroscience. During perceptual experience, association areas in the brain capture bottom-up patterns of activation in sensory-motor areas. Later, in a top-down manner, association areas partially reactivate sensory-motor areas to implement perceptual symbols. The storage and reactivation of perceptual symbols operates at the level of perceptual components – not at the level of holistic perceptual experiences. Through the use of selective attention, schematic representations of perceptual components are extracted from experience and stored in memory (e.g., individual memories of green, purr, hot). As memories of the same component become organized around a common frame, they implement a simulator that produces limitless simulations of the component (e.g., simulations of purr). Not only do such simulators develop for aspects of sensory experience, they also develop for aspects of proprioception (e.g., lift, run) and introspection (e.g., compare, memory, happy, hungry). Once established, these simulators implement a basic conceptual system that represents types, supports categorization, and produces categorical inferences. These simulators further support productivity, propositions, and abstract concepts, thereby implementing a fully functional conceptual system. Productivity results from integrating simulators combinatorially and recursively to produce complex simulations. Propositions result from binding simulators to perceived individuals to represent type-token relations. Abstract concepts are grounded in complex simulations of combined physical and introspective events. Thus, a perceptual theory of knowledge can implement a fully functional conceptual system while avoiding problems associated with amodal symbol systems. Implications for cognition, neuroscience, evolution, development, and artificial intelligence are explored"
- nitrogen 11y ago
  Fascinating abstract. I think I may have accidentally downvoted your comment though, when I meant to upvote (on a phone). I downloaded the PDF; hopefully I can find time to read it eventually. There are some gaps between the layers of abstraction presented in this abstract that I hope would be addressed by the full paper.
  - hellbanner 11y ago
    Their main point was that pre-computers, cognition was theorized to be tied to physical stimulus. When humans learned how to code computers, "AI" was theorized as existing in "black boxes" of pure abstract symbol manipulation. The paper discusses how this doesn't make sense, and how cognition must be tied to biological sensing to be "smart".
dchichkov 11y ago
I doubt that "physical manipulation" is a requirement for machine intelligence. It may as well be any kind of manipulation in any kind of reach and consistent environment. It could potentially be rich enough virtual environment. But, in practice, the only "rich enough" environment currently available is the physical world. And virtual classroom is no substitution...
sdg1 11y ago
There's a somewhat relevant concept in Seymour Papert's book Mindstorms. Papert talks about "body-syntonicity" where the act of drawing a circle on the computer screen with LOGO can be seen as bridging children’s sense and knowledge about their own bodies to something abstract. Young children would often pretend to be the LOGO turtle, acting out commands on the floor, during stages of planning a LOGO program, or while debugging LOGO code. It has also been a strategy for teachers to introduce programming and procedural ways of thinking. This video shows this in action (with some followup comments from Papert): https://youtu.be/bOf4EMN6-XA?t=254 https://youtu.be/bOf4EMN6-XA?t=254
- rurounijones 11y ago
  When I was in school with the BBC micro and the LOGO turtle we had an actual physical turtle hooked up to the machine via a long cable. Needless to say this was much more popular with us than just the screen. I am sure it is partly for the robotic'ness of it but we would often walk around with the turtle as it was moving. So what you are saying and linked to I can well believe.
  - juanuys 11y ago
    This sounds great. Do you have any more information on the mechanical turtle? (consumer kit, or a side-project by an industrious teacher?)
    - obitoo 11y ago
      Here he is , as i recall it was a kit http://www.classicacorn.freeuk.com/8bit_focus/logo/logo.html http://www.classicacorn.freeuk.com/8bit_focus/logo/logo.html
    - rurounijones 11y ago
      This was a consumer kit aimed at schools. Looks like there were many different types as obitoo has another that I have never seen. http://gallery.nen.gov.uk/asset78077_1582-.html http://gallery.nen.gov.uk/asset78077_1582-.html
rsaarelm 11y ago
There is the theory of general causality developed by Judea Pearl, which as far as I've understood implies that you need to be able to purposefully affect the world you're observing in some way to do experiments before you are able to infer causal relationships even in theory. See eg. http://www.michaelnielsen.org/ddi/if-correlation-doesnt-imply-causation-then-what-does/ http://www.michaelnielsen.org/ddi/if-correlation-doesnt-impl...

LiweiZ 11y ago

It seems the way of our brain works have some similarities with machines(well, it perhaps among the best average person at this time is able to perceive.). The brain gets information from sensors, tries to categorize everything and makes sense by use the categories in mind. New information from the sensors may trigger the brain to recategorize. All those steps make sense. I believe that's one of important reasons screen based devices are not good for little kids. They need to get exposed to the physical world since they are still learning how to interact with the physical world they live in. Leaving kids to screen is a low cost and time/energy-saving solution, but it increases the risk of "less-developed-focusing-muscle".

grubles 11y ago

Montessori-based education revolves around learning and working with physicality[0]: "Exploration, Manipulation (of the environment), Activity, Order, Orientation" etc. Disclosure: I was put into a Montessori school when I was young. [0]https://en.wikipedia.org/wiki/Montessori_education https://en.wikipedia.org/wiki/Montessori_education

javajosh 11y ago

So is the solution to focus on simulating a body within a fairly detailed simulated reality? Forget how you model emotions (which might be emergent from patterns of subtle physical sensations) how do you do prioperception in that context? Also, do you cheat and allow the entities "imagination" to be a perfectly modifiable simulation, or do you allow "errors" to enter into the simulation? What about the ethics of maintaining the integrity of the events and reactions that compose the entity, and the possible responsibility to ensure that reactions continue, or that those events have something to do with our "real" world and are not entirely simulated?

deckiedan 11y ago

'Society of the Mind'[1] by Eric L Harry is a great sci-fi read about AI and has this as a major part of the premise. I read it a few years ago and recently bought a copy, re-read it. Still excellent. [1] http://www.amazon.co.uk/gp/product/B00SRV9UGE/ref=as_li_tl?ie=UTF8&camp=1634&creative=19450&creativeASIN=B00SRV9UGE&linkCode=as2&tag=madswor09-21 http://www.amazon.co.uk/gp/product/B00SRV9UGE/ref=as_li_tl?i...

geographomics 11y ago

Related to this, an interesting read on the philosophy of embodied cognition: http://plato.stanford.edu/entries/embodied-cognition/ http://plato.stanford.edu/entries/embodied-cognition/

arethuza 11y ago

It's been a while since I worked in AI research but although I worked in agent architecture and model based reasoning I was came to believe in the approach of "the world is its own best model" associated with Rodney Brooks: https://en.wikipedia.org/wiki/Subsumption_architecture https://en.wikipedia.org/wiki/Subsumption_architecture i.e. Why put so much effort into modelling the world to allow reasoning when you can interact with the world to answer the same questions.

Kenji 11y ago

I feel far less intelligent after a day spent in front of a computer or a book than I do after a day spent walking in the woods. Odd, for me it's the other way around. I may feel more at peace, more relaxed and better able to fall asleep when I was in the wood, but I definitely don't feel as intellectually enriched as I do after reading a book or on the PC.

thesz 11y ago

Three main stresses that human body learn to deal with during the course of evolution trigger development of brain (and nervous system in general). These stresses are endurance activity (walk long hours), strength activity (fight or flight response) and famine. The agent responsible for development of nervous system in these cases is brain-derived neurotrophic factor. It is responsible for creation of new neurons and branching of old ones and directly related to intellectual performance. So your link to high level machine intelligence is wrong, I believe. I also think I explained above why do you feel smarter after a day of physical activity.

skj 11y ago
> So your link to high level machine intelligence is wrong, I believe. Alternate explanations are not refutations. Both could be factors.