Muazzam bir araştırma. Kendi evreninde, kendi oyununda, kendi yansımalarının tuzağından kurtulabilmenin yolu. https://t.co/Pa73uc9sdx
A Theory of How Columns in the Neocortex Enable Learning the Structure of the World.
Mentions
Extended reach
Total likes
Video views
Mentions per month
Oct 2017 – Mar 2025Where it's discussed
-
X / Twitter39 posts
-
News18 articles
-
YouTube8 videos
-
Facebook3 posts
-
Reddit2 posts
AI synopsis
What people are saying
Discussions include enthusiastic endorsements such as calling it 'super cool' and emphasizing its relevance to AI and brain understanding. Phrases like 'good weekend reading' and mentions of the potential for future AI applications show excitement and recognition of the publication’s significance. Technical references, such as the voting mechanism in cortical columns and the comparison to neural transforms, demonstrate deep engagement and a recognition of the research's potential impact.
Most discussions support or accept the significance of the research, with some referencing it as foundational or important for understanding brain functions.
Posts demonstrate high curiosity, often highlighting how the publication relates to broader topics like AI, neuroscience, and cognitive models, indicating strong engagement.
Several comments delve into detailed aspects such as the voting mechanisms between cortical columns, implications for AI, and neural modeling, reflecting deep engagement.
While many acknowledge the research's importance, few claim it has immediate practical impact, suggesting a perception of long-term influence rather than urgent change.
Abstract in brief
- The article discusses how neocortical columns, organized into layers, facilitate learning and recognizing complex objects by integrating sensory input over time and space.
- It introduces a network model where each column computes a location relative to an object, which, combined with sensory data, improves object recognition and allows understanding of objects extending beyond individual cells’ receptive fields.
- Simulations indicate that small, multi-column networks can efficiently learn to recognize many objects with minimal movements, suggesting that the columnar structure of the cortex plays a powerful role in perception and modeling beyond previous assumptions.]
Top posts
Most liked on X and BlueskyIn this meeting, Numenta intern Jack Schenkman reviewed a paper on multiscale representation in the hippocampus by @tamir_eliav et al., and our researcher Ben Cohen reviewed a paper on representational drift by Schoonover et al. https://t.co/xo4RcJI0To
In the continuation of last week’s research meeting, @SubutaiAhmad explains voting in the Thousand Brains Theory. In part two, he explains how cortical columns vote across multiple columns. TLDW below
Read more about how cortical columns in the neocortex enable learning the structure of the world here: https://t.co/CWAfiX0aLx
@IntuitMachine @Numenta Thanks! We published a detailed model in 2017 of how voting might work using long-range lateral connections that exist between cortical columns (layer 2). https://t.co/8KQqpmOvtI
For reference: "Columns" paper: https://t.co/sjN1hzZ8Oz "Frameworks" Paper: https://t.co/g7JL2KDG3h
All mentions
Neocortical Columns and Object Recognition in the Brain
This video explores how neocortical columns enable learning and recognition of complex objects by integrating sensory input over time and space, emphasizing the role of columns in perception and understanding beyond…
Role of Allocentric Location in Neocortex Function
In this talk, Jeff Hawkins describes a theory that the neocortex processes both sensory data and allocentric location representations. This dual input allows cortical columns to build comprehensive object models…
Neocortical Columns and Object Learning in the Brain
This video accompanies the paper titled A Theory of How Columns in the Neocortex Enable Learning the Structure of the World by Jeff Hawkins, Subutai Ahmad, and Yuwei Cui. It explores how cortical columns use location…
Understanding Active Dendrites in Neuron Prediction and Context Integration
Subutai Ahmad reviews the biology behind active dendrites and discusses how they influence neuronal computation. He explores various findings on dendritic functions, emphasizing their role in prediction and context…
Understanding Cortical Column Voting Mechanisms in Neural Networks
Subutai Ahmad explains the voting mechanisms in cortical columns developed by Numenta, highlighting how neocortical columns facilitate learning and object recognition by integrating sensory input over time and space.
Voting Mechanisms in the Thousand Brains Model of Cortical Columns
Subutai Ahmad explains how cortical columns use voting mechanisms to learn predictive models of sensorimotor sequences, laying the foundation for understanding how the neocortex recognizes complex objects through…
Thousand Brains Theory: Hierarchical Voting for Rapid Object Recognition
In this video, Subutai Ahmad explains how the neocortex uses hierarchy to vote across levels and sensory modalities, enabling rapid consensus on objects through multiple cortical columns and sparse connections, leading…
Multiscale Neural Representation and Drift in the Neocortex
This video reviews key research on neuronal encoding of spatial perception in the hippocampus of flying bats and stability of odor responses in the olfactory cortex. It also explores how neocortical columns enable…
RT @vedattozyilmaz: Muazzam bir araştırma. Kendi evreninde, kendi oyununda, kendi yansımalarının tuzağından kurtulabilmenin yolu. https://…
Muazzam bir araştırma. Kendi evreninde, kendi oyununda, kendi yansımalarının tuzağından kurtulabilmenin yolu. https://t.co/Pa73uc9sdx
The solution is already quite clear: https://t.co/mKYMcNBCqN https://t.co/eWyRKqDQ7G https://t.co/ikdMjts9TC
Hawkins, J., Ahmad, S., & Cui, Y. (2017). A theory of how columns in the neocortex enable learning the structure of the world. Frontiers in Neural Circuits, 11, 81. https://t.co/AJd49ITYdE
https://t.co/vpUP9SwONd
@goldstein_aa These are the two most important papers IMHO: https://t.co/5y0UJk79T1 https://t.co/Qn4eA3wvzq
@HansaMD_ It’s not the five fingers who vote. It’s the cortical columns associated to the fingers. I suggest you read this paper for a full description of how the system works: https://t.co/Qn4eA3wvzq
RT @hwingo: A Theory of How Columns in the #Neocortex Enable Learning the Structure of the World https://t.co/TG7kd2Ojj6 #ICYMI #neurosci…
A Theory of How Columns in the #Neocortex Enable Learning the Structure of the World https://t.co/TG7kd2wHUw #ICYMI #neuroscience #AI #ArtificialIntelligence #brain @Numenta
A Theory of How Columns in the #Neocortex Enable Learning the Structure of the World https://t.co/TG7kd2Ojj6 #ICYMI #neuroscience #AI #ArtificialIntelligence #brain @Numenta https://t.co/FXZcCEsWuJ
https://t.co/7imINNVMFg
RT @Numenta: Read more about how cortical columns in the neocortex enable learning the structure of the world here: https://t.co/CWAfiX0aLx
RT @Numenta: Read more about how cortical columns in the neocortex enable learning the structure of the world here: https://t.co/CWAfiX0aLx
Read more about how cortical columns in the neocortex enable learning the structure of the world here: https://t.co/CWAfiX0aLx
“A Theory of How Columns in the Neocortex Enable Learning the Structure of the World.”
2/4 https://t.co/WZPnuAPYIX
For reference: "Columns" paper: https://t.co/sjN1hzZ8Oz "Frameworks" Paper: https://t.co/g7JL2KDG3h
RT @Numenta: In the next research meeting, he will review voting across multiple columns. Read our "A Theory of How Columns in the Neocor…
RT @Numenta: In the continuation of last week’s research meeting, @SubutaiAhmad explains voting in the Thousand Brains Theory. In part two,…
RT @Numenta: In the continuation of last week’s research meeting, @SubutaiAhmad explains voting in the Thousand Brains Theory. In part two,…
RT @Numenta: In the continuation of last week’s research meeting, @SubutaiAhmad explains voting in the Thousand Brains Theory. In part two,…
In the continuation of last week’s research meeting, @SubutaiAhmad explains voting in the Thousand Brains Theory. In part two, he explains how cortical columns vote across multiple columns. TLDW below
miles de modelados del mundo de forma simultanea, que se complementan y compiten entre sí, hasta llegar a un estado final de decisión a través de un mecanismo de "votación neuronal"
This is a good weekend reading, to understand how the future of AI might work by understanding how human’s brain works: A Theory of How Columns in the Neocortex Enable Learning the Structure of the World https://t.co/04XGZJEPJ4
RT @Numenta: In this meeting, Numenta intern Jack Schenkman reviewed a paper on multiscale representation in the hippocampus by @tamir_elia…
RT @Numenta: In this meeting, Numenta intern Jack Schenkman reviewed a paper on multiscale representation in the hippocampus by @tamir_elia…
RT @Numenta: In this meeting, Numenta intern Jack Schenkman reviewed a paper on multiscale representation in the hippocampus by @tamir_elia…
RT @Numenta: In this meeting, Numenta intern Jack Schenkman reviewed a paper on multiscale representation in the hippocampus by @tamir_elia…
In this meeting, Numenta intern Jack Schenkman reviewed a paper on multiscale representation in the hippocampus by @tamir_eliav et al., and our researcher Ben Cohen reviewed a paper on representational drift by Schoonover et al. https://t.co/xo4RcJI0To
@IntuitMachine @Numenta Thanks! We published a detailed model in 2017 of how voting might work using long-range lateral connections that exist between cortical columns (layer 2). https://t.co/8KQqpmOvtI
@connectedregio1 @RebelScience @markcannon5 For example this is a very complicated paper https://t.co/gvmJDUNHko
@bbc1183 @mrcslws We have experimented with union pooling during sensorimotor inference. In what we call the "output layer" of this paper we are pooling objects in a union. https://t.co/lpvL3zKPFN
A Theory of How Columns in the Neocortex Enable Learning the Structure of the World [Reading this through I think the example of vision provides a stronger argument than sensorimotor sensing, e.g. considering the constant changing information on the fovea] https://t.co/V7tajeiax9
I also mentioned @Numenta ‘s research on reverse-engineering the neocortex, when @jfcloutier asked about universal function... here’s a short video that explains the magic of allocentric location. Super cool stuff: https://t.co/9Q4wfvgMC1
https://t.co/wzmrTEnLyn
HTM neurons in the output layer seem to doing hough transform. https://t.co/gvmJDUNHko
@el33th4xor Any cross-over potential in how Avalanche protocol works in the same way as neocortex in this view? https://t.co/aPrPtRLuZD
RT @Numenta: .@CDMaver debunks the belief that the brain fills in the gap in our vision to compensate for our blind spot using our sensorim…
If this is the case and everything springs from loci, wouldn’t it mean tests of intelligence could be boiled down to a person’s aptitude at primitive spatial awareness, memory, and composition? https://t.co/HcNFLq9yhL
V
N
SA
A
Q
YK
HG
LD
H
H
J
I
JH
DP
PS
RA
N
CB
R
BH
S
P
M
MA