ฝัง
- เผยแพร่เมื่อ 24 ก.ย. 2024
- Get ready for a breakthrough new framework for intelligence based on grid cells in the neocortex! We understand more about how your brain models reality than ever before. Watch as we explain how your brain represents objects in space. More info about our theory at numenta.com/ne...
Thank you!!! I can't believe I'm alive to see this!!!
thanks rhyolight , Your contribution was very helpful and understandable, I am so grateful
impressive... only 500 people seen this Creating "Buy-in" can be tough.
Keep up the good Work I love it whenever you guys put out a video on Htm school
Hey Matt , Awsome video, I think the breakthroughs you guys have made and your project in general will prove to be one of the most impactful research projects in human history and I am so inspired and greatful. Could you please comment on how OpenWaters new MRI technology could contribute to your understanding of cortical columns , is it sufficient for you to ultimately understand the full workings of the cortical colemns , or will more powerful scanning tools be required.
Thanks for the compliment! I am not familiar with OpenWaters MRI technology.
Good explanation. Very helpful. Thanks for the video.
About how long till the 1000 brains video? That one seems really interesting. Good job on episode 15, very informative.
I'm guessing a month or so...
@@NumentaTheory Been more than 3 months now, still waiting like a fanboy!
@@NumentaTheory Don't leave us hanging.
@@argolo9512 I have been working on it once again. :)
I have Dyslexia, I have always struggled to explain what its like. I can now say that either my universe is too small, objects in my universe get clumped together or my displacement cells are not very accurate. This also might explain why dyslexia is more common in the creative industries "although I think the main reason is that its a profession that one can excel at without being outed.", creativity is all about bringing things together that you would otherwise not think to bring together. If a persons displacement cells were less accurate they may begin to link two things that were otherwise unrelated and then evaluate that link and possible create a new valid displacement were none existed before.
It would be sad if thats all creativity is though as that's the thing a lot of folks seem to think will save us from the automation revolution.
We get a lot of comments like this from people with neurological disorders, who have hope that a better understanding of intelligence will help. I think it will.
Psycologists can help if the condition is caught early.
What and how the treatment works I don't know.
I have this condition too as does my son, but that maybe coincidence. The way I've explain it is to imagine a page of words is like a natural forest, in a gentle breeze, with many different types of trees, shrubs and low ground cover. They are moving slightly so it's initially it's difficult to see detail. With concentration you pick out those differences, but it's easy to miss things out. This is what it's like to read. Reading classic hand writing, rather than printed text is a nightmare. Spelling is remembering that detail and reverse engineering it in your own handwriting. The problem is it doesn't look like the original printed text, but is a representation of it. Rather like a childs picture of a tree, a brown trunk with a fluffy green blob on the top. Mistakes are everywhere, so you read it, correcting and adding detail, repeating the process until you're satisfied. Practice, makes perfect. I've got better with age, but the problem has never gone away. Obviously, technology has helped enormously and I wouldn't have been able to go to university without it.
Finally! Thank you for all your effort Matt! Also, is there going to be a paper or something released on this? It's a really interesting concept and would love to see more details.
numenta.com/neuroscience-research/research-publications/papers/a-framework-for-intelligence-and-cortical-function-based-on-grid-cells-in-the-neocortex/
Thanks! You guys are the best, keep on doing what you're doing
Do you think the same mechanism works for abstract concepts? Like, statistics is somewhere near probability theory, they both are inside math, which is inside science?
Yes we do.
Waited so long for this!
with regard to the ‘universe’ idea, how does that fit with allowing people to first learn local features in space of certain regions, then learn to fit the regions together, and integrate all that knowledge into a map of the whole object?
Thank you!
I had a flashback of linear algebra class when he talked about displacement.
Francis crick - thinking of the brain.
Grid cells - everything can sense. All space
Grid cells in neocortex
Universe space
So like my brain broke at several times towards the end then Matt apologises for blowing our minds lmao. Like in all that confusion and marvelling.
Well. Displacement as a vector.
Hi Matt,
Thanks for this videos,it is really help. Next question arias where+how to start coding it on NuPIC. in the Github there is lot of code but I didnt find a sequential tutorial (I mean coding) to start with. Could you please provide a link.
Thanks in Advance
nupic.docs.numenta.org/stable/quick-start/index.html
Looking for my broom so I can sweep up the my blown mind bits! :-) New office? So where's the new office?
First of all, thank you Matt for sharing all this wonderful information on videos ...really appreciate it. I've been very interested in this topic for many years and I believe Numenta is really onto something very vital.
My question: What do you think about this -- could it be that instead of our neocortex modelling the 'physical' world, it 'creates' a simulation of it? Seems to me that simulation of the physical world is probably more accurate, considering that we feel/think/see/move/hear/talk/etc even during our dreams (in a pitch dark/no sound rooms). I think we are still somewhat stuck in a 'flat earth' mentality believing that this HUGE physical world, expanding, with zillion of galaxies is all 'real' -- that seems so improbable and unlikely, whereas a simulated computational world seems far more probable, doesn't it? Especially considering how as Jeff mentioned in one of his videos, 'knowledge is information/facts arranged in a useful way, which our neocortex assigns in a reference frame (via Grid cells), and then navigates through it for a particular outcome' -- this seems to point that everything we see/hear/think is most likely based on a computationally based abstract internal reference frame based simulation and maybe the entire physical world with things/space/time is a derivation of this simulation. I don't mean to imply that some whiz kid has created all this; I believe the Universe is probably some type of a computation, within which sub-components exist in some areas that also compute (like our brains), kinda like Steven Wolfram's point on how some regions of complexities can develop out of very simple programs (eg cellular automaton Rule #30) and Joscha Bach's ideas. Please share your thoughts on this. Thank You.
Tony Hashim, MD.
You can call it a simulation if you like. :)
@@NumentaTheory -- Wow! Very short, abrupt, not helpful and in all honesty very surprisingly disappointing and a rude reply from an organization that I had high regard for. My well thought out and thorough question deserved more than a 9 word answer - and certainly did not deserve a 'smiley face' ending! Very, Very disappointed indeed!!! Sincerely, Tony Hashim MD.
@@NumentaTheory One last thing - I see where you're coming from by stating, "you can call it a simulation if you like" -- basically suggesting, 'who cares, you can see it as a simulation, others can see it as a physically based reality...makes no big difference'. Wrong! This type of thinking is reminiscent of those in Copernicus/Galileo's days, suggesting who cares if you think earth is just a tiny speck in the universe of zillion galaxies, we can still go on believing in the geocentric model of the universe OR when Einstein suggested gravity should be looked at more as a curvature in space/time, others said who cares we can still live the same type of life thinking it's an attractive force. Hope you can see that such rhetoric that just b/c reality seems to be physically based (though evidence suggests it's more of a computational reality), one can just go on 'thinking or calling it as one likes' -- that is not what we expect from those such as you all who have such a high platform to express facts via evidence/science to the general public.
T. Hash I am not disagreeing with you!
Sorry I did not mean to hurt your feelings! We get comments like this all the time. The ideas you are presenting are not novel. You are thinking of it in the right way. - Matt
Time to get some soundproofing!
Yayyyy HTM School is back :)
Share it with your friends!
Trying to google the word "holonaligy" at 9:07. Just got curious... Did I spell it wrong? Mindblowing stuff btw. 👏🏼👏🏼👏🏼
Analogy 👍
😂
[5:27] "Each [cortical] column has its own unique experience of every object it has ever taken a part in learning"; An issue with storing modality specific representations of an object in different cortical columns is that it would make the task of imagining the representation of a 3D object in a new modality difficult, whereas humans can do this quite easily.
I don't see how it would make modality imagination difficult. In fact you would imagine that new things you've never heard before would sound similar to things you've seen / heard in the past. Modalities inform each other.
Thanks for your response. Further to my previous comment, I think this method would be an inefficient use of cortical space (m.th-cam.com/video/fhnMUc36opI/w-d-xo.html 'Why Does the Neocortex Have Layers and Columns - video demonstration'). It would be more efficient if the brain was able to derive the 3D coordinates of any sensory input (eg finger, visual area, etc), and build up a modality independent map (mesh) of its surrounds. Spatial/temporal object recognition would operate on this mesh. Local (wrt agent) visual coordinates should be calibrated to tactile coordinates (if need be by visualising where one is touching). Likewise, we need to be able to imagine (predict) what this stimuli will feel/look like in a new modality.
Where's the next episode on the 'thousand brains theory'?
I've been working on it again, see www.twitch.tv/videos/483536324
@@NumentaTheory Please do and post episode 16 onward soon. We truly appreciate them.
Cool shirt!
I hope we won't have to wait six months again for that next video. I thought you had moved on and that was it. By the way, great job. Keep them coming. ::: Separate issue. Why don't you have merchandise as suggested by Nick King in the comments? While at uni some friends and I had some tee shirts made up with a picture angry looking young man with his fist in the air, sort of soviet style. The text around it said "No Use Without Instantiation". A programmers joke that everyone here will understand, but it confused the hell out of everyone else.
I'm working on the next two videos now, so it will not be another 6 months. The only merch I have available is www.redbubble.com/people/rhyolight/works/27261237-single-order-memory and www.redbubble.com/people/rhyolight/works/27261362-high-order-memory :)
@@NumentaTheory glad to hear that, although it is not that far :)
Although starting by the release date almost 6 months have already passed (2 days remaining)
@@climatechangedoesntbargain9140 I'm sorry Andreas, but my priorities have recently changed. I am no longer working on videos for TH-cam. I plan on continuing at some point in the future, but as a Community Manager, I am getting a lot more interaction on my twitch channel, so I am focusing my efforts on community engagement there. www.twitch.tv/rhyolight_
3:10 Just turn on the light! xD
4:20 what's up with "sensory input" mess up on your hand? after the gratuitous robot voice, i doubt this is an unintended artifact. both look like simple inside jokes, but as interesting as all these sounds, and how excited you are on describing each little thing, i don't think i'm coming back and still, i'm curious about those fringe jokes. anyone?
Nope, these were both unintended artifacts, not inside jokes. I tried to accommodate them as best I could.
"In the next video..."
I have a script! (But then I had a script 6 months ago...)
@@NumentaTheory Oh cool! In the meantime I just started reading On Intelligence.
You were so inspiring to me 1,5 years ago, when I started learning about AI, but on the way it feels like you lost the moto you started with.
I tried contacting you guys but someone there doesn't share the vibe.
Hint try vision, it uses 25-50% of our brain power, how do your models work with vision, what lacks?
Thank you for your guidence I have moved on past it now.
The 25-30% of your neocortex that processes vision works the same as the rest of your neocortex, that is the point.
@@NumentaTheory exactly my point.
In ~15 days (presentation is ready) I will give my lecture on vision and how I think the neocortex works not physically but functionally thus proving the thesis through vision.
Then I will go and plot a theory about General AI in a system I call deep-wide NN, that might even become AGI.
I do think HTM plays a large part in it but it is only half the picture, especialy if we want to go towards AGI.
I agree that HTM will play a big role an AGI, but we never said it was the only theory necessary to achieve that goal. We are all working on the same problem. Good luck with your work!
@@NumentaTheory I wrote an invitation letter to a mail I think belongs to mr. Hawkins to my presentation. Looking forward in greeting my heroes to my presentation.
@@SaveAsss Hi I'm still new to SDR based systems and would like to learn more about the different systems out there. Do you have a link to your presentation or paper?