Long term follower (and reader,). This was extremely helpful. I've been struggling to produce a narrative and this was what I was looking for. Thanks again.
Appreciate your videos and what you guys are doing, very interesting. Just one comment it would really be great if you could invest in separate wireless microphones for the presenters as the reverberation is very annoying and sometimes you cannot understand what the presenter is saying Thanks again
Keep up this excellent work, guys. You are giving me deeper insights into how the brain really works and additionally how might we approach AGI from a new angle. Fantastic stuff.
This is perfect. Matt supplying the counterpart for a plato-style dialogue. How about providing a link to a forum topic so that subsequent conversation can occur there? Thoughts that arise in my brain: - I love the idea of seeing a cortical column as a superposition of separate processes, and pulling them apart. In this case orientation and location. Reminds me a little of QuakeWorld (Mouse for Changing orientation, Keyboard for Changing location). - But this only applies for visual stuff. So suppose we analyse a cortical column receiving input from the visual cortex -- And maybe it is structured in this manner. But what if we analyse a column receiving input from the auditory cortex? This separation now makes no sense -- we may find a different pattern of layers? - So maybe each column learns to dynamically Figure out its own layering And interconnection between layers Depending on its input/feedback? - How about analysing a more simple animal brain, Like an ant? And then moving upwards, Reptiles then mammals? Surely that's the intelligent logical progression... - This entire talk focused on one single column. But I think we need to see another Exposition with two columns side-by-side Showing the interconnections between the columns, Maybe with a "Thalamus" circle In the middle of the connection if it routes through the thalamus. - How about looking at Using genetic algorithms to attempt to evolve brain structure? Look at the parameters varied by various genes, Such as Dendrite/Axon Properties At various vertical locations within the column (Sideways connections, up/down connections, Thalamus connections, Density, etc). And try to generate creatures That must survive in an artificial environment. I mean the whole project is to Work from the biology, So why not work from the biology All the way down to the ground including evolutionary Dynamics?
Here is a discussion already: discourse.numenta.org/t/outline-of-jeffs-april-4-2019-presentation-what-do-we-know-about-cortical-columns/5767 (please feel free to start a new topic or add your comments to that thread)
Very nice. My understanding: 1.Two way: 1.egocentric(to navigate) 2.allocentric(to understand object). 2.these two are base on one universal algorithm: location(grid cell). 3.Motor output is for continuous feedback(also is the survival need: to act)
@Matt This fall I'm taking a robotics and spatial intelligence machine learning class. I'm not sure how it will work out, but I'm at least planning to try to implement the radial and linear system and see if I can get them to inform each other. I'll make sure to let you know if anything significant is found. If Jeff plans on publishing a paper, maybe an example of it working in machines could support the theory. If you have any advice on implementation I'd be happy to hear it.
There was a small mistake at 45:30. I guess 6b in primary cortex doesnt project to higher order thalamus (LP) , actually it is 5b. please check this paper onlinelibrary.wiley.com/doi/full/10.1002/cne.24423
I found it curious how I lost the idea whose voice I hear in the end of the video, at 1:05:29, when Matt comes closer to the camera and his voice gets lower. I think it's that my predicted scale of his voice was not coherent with that predicted by my internal representation which just forgot the location of the mic :)
This was extremely clarifying and useful. ALWAYS glad for any time that Jeff can share his thoughts.
Another awesome video, thanks Matt and Jeff!
I'm reading A Thousand Brains ....truly a deep dive.
This video is wonderful. I wish I could read a whole book with all the info that Jeff knows.
Long term follower (and reader,). This was extremely helpful. I've been struggling to produce a narrative and this was what I was looking for. Thanks again.
Appreciate your videos and what you guys are doing, very interesting. Just one comment it would really be great if you could invest in separate wireless microphones for the presenters as the reverberation is very annoying and sometimes you cannot understand what the presenter is saying
Thanks again
Matt recently passed away .. RIP to a real one
Keep up this excellent work, guys. You are giving me deeper insights into how the brain really works and additionally how might we approach AGI from a new angle. Fantastic stuff.
This talk makes sense now that I've read A Thousand Brains
Beautiful work!
This is perfect. Matt supplying the counterpart for a plato-style dialogue.
How about providing a link to a forum topic so that subsequent conversation can occur there?
Thoughts that arise in my brain:
- I love the idea of seeing a cortical column as a superposition of separate processes, and pulling them apart. In this case orientation and location. Reminds me a little of QuakeWorld (Mouse for Changing orientation, Keyboard for Changing location).
- But this only applies for visual stuff. So suppose we analyse a cortical column receiving input from the visual cortex -- And maybe it is structured in this manner. But what if we analyse a column receiving input from the auditory cortex? This separation now makes no sense -- we may find a different pattern of layers?
- So maybe each column learns to dynamically Figure out its own layering And interconnection between layers Depending on its input/feedback?
- How about analysing a more simple animal brain, Like an ant? And then moving upwards, Reptiles then mammals? Surely that's the intelligent logical progression...
- This entire talk focused on one single column. But I think we need to see another Exposition with two columns side-by-side Showing the interconnections between the columns, Maybe with a "Thalamus" circle In the middle of the connection if it routes through the thalamus.
- How about looking at Using genetic algorithms to attempt to evolve brain structure? Look at the parameters varied by various genes, Such as Dendrite/Axon Properties At various vertical locations within the column (Sideways connections, up/down connections, Thalamus connections, Density, etc). And try to generate creatures That must survive in an artificial environment. I mean the whole project is to Work from the biology, So why not work from the biology All the way down to the ground including evolutionary Dynamics?
Here is a discussion already: discourse.numenta.org/t/outline-of-jeffs-april-4-2019-presentation-what-do-we-know-about-cortical-columns/5767 (please feel free to start a new topic or add your comments to that thread)
This video was super useful - thank you!!
Thank you, very interesting theory to think of
Very nice.
My understanding:
1.Two way: 1.egocentric(to navigate) 2.allocentric(to understand object).
2.these two are base on one universal algorithm: location(grid cell).
3.Motor output is for continuous feedback(also is the survival need: to act)
Now I absolutely believe that the Numenta cup is the key.
Without it, you go nowhere.
Where can I buy this cup???
It would be useful if there were some references on this. I like to read things from multiple sources.
Greetings from Ukraine! Very impressive! Thank you.
@Matt
This fall I'm taking a robotics and spatial intelligence machine learning class. I'm not sure how it will work out, but I'm at least planning to try to implement the radial and linear system and see if I can get them to inform each other. I'll make sure to let you know if anything significant is found. If Jeff plans on publishing a paper, maybe an example of it working in machines could support the theory.
If you have any advice on implementation I'd be happy to hear it.
Be sure to join HTM Forum and post about your work there!
@@NumentaTheory Will do!
@@NumentaTheory can you give link to forum?
@@ValeraVlasyuk discourse.numenta.org/
There was a small mistake at 45:30. I guess 6b in primary cortex doesnt project to higher order thalamus (LP) , actually it is 5b. please check this paper
onlinelibrary.wiley.com/doi/full/10.1002/cne.24423
Input layer - layer 4. Ears.
Layer 6 - grid cells
I found it curious how I lost the idea whose voice I hear in the end of the video, at 1:05:29, when Matt comes closer to the camera and his voice gets lower. I think it's that my predicted scale of his voice was not coherent with that predicted by my internal representation which just forgot the location of the mic :)
Geoffrey Hinton and Jeff Hawkins should definitely have a chat together and livestream it! They are after the same thing
I think that one dislike was from Matt .😁 Just for fun.
Why are some birds so smart if the cortical columns are the key to intelligence? And why aren't dolphins smarter than they already are?
like #97
5 минут начала - разговоры ниочем и тщательное вытирание доски...очень жаль...отключаюсь...
Please clean your whiteboard and do your prep before you start the broadcast. It's a waste of time to watch 5m of people straightening up...
I thought I had it clean enough, but Jeff decided otherwise **shrug**
Deeply constructive criticism of the algorithm