The last invention.

  • 388 Replies
  • 53286 Views
*

ranch vermin

  • Not much time left.
  • Terminator
  • *********
  • 961
  • Its nearly time!
Re: The last invention.
« Reply #255 on: December 19, 2017, 05:50:56 pm »
Yes, You should add a motor component.

*

Korrelan

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1113
  • Look into my eyes! WOAH!
    • YouTube
Re: The last invention.
« Reply #256 on: December 20, 2017, 10:54:43 am »
The system does have the equivalent of a motor cortex and cerebellum. I’m driving speech output through good old Microsoft Mary at the moment. Speech is expressed as a temporal pattern which was causing problems, so I’m writing a module to re-combine individual phonemes back into words, sentences in real time… ear muffs are a requirement when working on that one lol. 

I can also drive any number of servos through an output module that recombines neural patterns back into a linear output.  This drives the head servos/ steppers and will also drive the arms/ hands... which I'm still working on.

The motor cortex was required because the system has to be able to hear its own speech output/ patterns and see/ feel its own movements, this is a requirement to enable the schema to build an internal model of it's own body/ reality, which of course is linked to self awareness.

 :)
It thunk... therefore it is!... my project page.

*

ranch vermin

  • Not much time left.
  • Terminator
  • *********
  • 961
  • Its nearly time!
Re: The last invention.
« Reply #257 on: December 20, 2017, 12:16:22 pm »
god that sounds impressive, what your asking of yourself.
im going for a more direct artificial approach... but im bloody stuck on an ordinary bug, ill get back to it tomorrow.

Maybe if u see some success from me, maybe u might want to shortcut yours sometime, if you dont get anywhere for a while.  its just an idea.

*

Zero

  • Trusty Member
  • ********
  • Replicant
  • *
  • 737
    • Thinkbots are free
Re: The last invention.
« Reply #258 on: December 20, 2017, 02:43:35 pm »
The system hearing and listening to its own voice is very exciting!! So, it can feel itself, wow. It must be difficult to monitor what's happening then... I mean, the vids are always really cool, but they don't show the underlying moving structures... God damn, I'd love to see those structures, with clear labels on them. I know it's impossible, becauze they're all plugged together.

ED: I see you don't resent me for suggesting minecraft, how nice of you  ;D
Thinkbots are free, as in 'free will'.

*

LOCKSUIT

  • Emerged from nothing
  • Trusty Member
  • *************
  • Transformer
  • *
  • 2165
  • First it wiggles, then it is rewarded.
Re: The last invention.
« Reply #259 on: January 18, 2018, 03:56:37 am »
I'm really excited about korrelan's project. I too would love to see more, like a tour of his lab, his Beowulf, his notes, and the AGI blueprint overview. The more I/we know about you / your AI / your knowledge, the better I/we can give feedback.

Also, I would like to cooperate on a higher level with any of yous, like a DeepMind/Google Brain team. Google does very good because they have so much money, computational power, and top experts (teams) combining different skill sets (I have a unique set: passion+unlocking). While we can't easily get supercomputers, we can easily unite and create something powerful that runs efficiently.

Korrelan, I'm reading a book called www.DeepLearningBook.org and getting better understanding of terminologies, and you/I may not understand some of my/your standard/unique terms, but, maybe the Global Thought Pattern didn't work/emerge until the frontal cortex was implemented? Because the GPT is a super-high representation at each step and represents the active representations including the GTP itself (hence mirror-mirror effect to infinity). Btw do activated representations ex. in higher layers fade in brightness slowing leaking energy? As you watch a video for example, this would show a slope of bright-to-dark lightshow moving around the brain.
« Last Edit: January 18, 2018, 04:31:27 am by LOCKSUIT »
Emergent

*

Korrelan

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1113
  • Look into my eyes! WOAH!
    • YouTube
Re: The last invention.
« Reply #260 on: January 18, 2018, 09:45:46 am »
Quote
I'm really excited about korrelan's project.

I’m just getting back up to speed after the holidays...

Quote
I'm reading a book called www.DeepLearningBook.org

Good. All information is useful… obviously the more you know the better.  Deepmind is a very cool concept/ architecture but in its present format I personally don’t think it will ever be anything but a useful tool… a narrow AI.  Although the schema is incorrect Deepmind would represent the equivalent of the retina and V1, V2 visual cortices at the moment.
 
Quote
hence mirror-mirror effect to infinity

The frontal cortex is made up of hundreds of sub areas that are self organised according to which other cortex networks/ regions they are tuned to.  So for ocular/ vision for example it basically listens to the patterns in GTP that originated in the visual cortex.  As it learns the patterns it generates its own sparse patterns which eventually feedback to the visual cortex hubs amongst many other areas. Together they eventually reach a kind of equilibrium where the visual cortex’s definitions/ understanding of visual patterns are influenced by the state of the frontal cortex and vice versa. Hundreds of other networks are all doing the same and inter mixing their information in thousands of hubs/ relay points.  This is where internal, ocular, audio, tactile sensory streams are filtered, mixed and matched.

Quote
Btw do activated representations ex. in higher layers fade in brightness slowing leaking energy?

Actual images as you understand then never actually get past the retina.  The retina re-encodes the light patterns into a sparse/ compressed temporal representation that is lastly encoded by the retinal ganglion cells before being sent down the optic nerve.  You have approx 150 million light receptors (rods/ cones) in each retina, but only about a million nerve fibres in the optic cord/ nerve that passes the information to your brain. 

When you view a scene/ image for the first time your past visual experience fills in approx 85% of the information you think you are actually seeing.  Then as you move your fovea (hi resolution center) around the scene, focusing on objects you are not certain/ curious about the finer details are filled in.  This is part of how your imagination works, you are able to mentally superimpose visual information on top of what you think you are actually seeing or ‘see’ pictures in your ‘minds eye’.  It’s the main cause of hallucinations and many other mental phenomena like pareidolia and synesthesia.

So no…the patterns that represent what you are looking at don’t actually fade/ dim in the common understanding of the terms but they do loose their initial resolution/ complexity as they are converted into the mental facets your brain requires/ understands.

 :)
It thunk... therefore it is!... my project page.

*

LOCKSUIT

  • Emerged from nothing
  • Trusty Member
  • *************
  • Transformer
  • *
  • 2165
  • First it wiggles, then it is rewarded.
Re: The last invention.
« Reply #261 on: January 18, 2018, 06:40:39 pm »
So Frontal Cortex affects Sensory Cortices, Sensory Cortices affect Frontal Cortex, repeat? What is the purpose of the Frontal Cortex affecting the ex. Visual Cortex? Better recognition accuracy? Higher level concepts/representations? And what is the purpose of the Visual Cortex speaking back to the Frontal Cortex? To improve how its doing its job? Backprop?

I know there's no images in the brain now. I wasn't talking about this though. What I meant was that the neurons that stand for/represent eyes/faces/talking/etc could dim in energy and fade. Why? > Because normally neuron selections/activations in the our brain algorithms go black after activation, where, if they slowly de-activate and get darker slowly, then it draws the search for new input, i.e. new sensory input considers what you were last thinking. Simply put, any activity in the brain should slowly fade away, but not instantly turn black. Do you/s do this?
Emergent

*

LOCKSUIT

  • Emerged from nothing
  • Trusty Member
  • *************
  • Transformer
  • *
  • 2165
  • First it wiggles, then it is rewarded.
Re: The last invention.
« Reply #262 on: January 27, 2018, 08:00:52 pm »
Hey korrelan, can you give us a somewhat thorough list of where you learnt all of your knowledge from? I want to be more like you.

Here's an example of the type of reply I'm expecting from you:

EXAMPLE:
These 3 e-books are the most important I read.
Took Computer Science at Harvard for 4 years (online course equivalent is: "n")
Read about 80 random articles.
Watched about 60 lectures/videos on YouTube.
Emergent

*

LOCKSUIT

  • Emerged from nothing
  • Trusty Member
  • *************
  • Transformer
  • *
  • 2165
  • First it wiggles, then it is rewarded.
Re: The last invention.
« Reply #263 on: January 29, 2018, 05:57:12 am »
(This makes 3 replies for korrelan to read):

https://en.wikipedia.org/wiki/Synesthesia

I seem to have the Spatial Sequence Synesthesia. Don't yous? I can imaginate in the daytime pretty powerfully. I can see a clock all around me like in weird strange views like a bit out of body, or not!, and "see" them to left, idk extending my visuual field by pure representation?

I also have the good memory. I often come up with ex. 8 things to do in bed for tomarrow (ex. wash, hike, run, and skydive will help create a new AGI) and if I make a 1 picture, I only have to remember 1 thing to get em all! -> A generated video of a guy running up a mountain that's rained on and then skydives from off the top.
Emergent

*

Zero

  • Trusty Member
  • ********
  • Replicant
  • *
  • 737
    • Thinkbots are free
Re: The last invention.
« Reply #264 on: January 29, 2018, 08:36:53 am »
It's good to see that you want to learn things, LOCKSUIT.  O0
Thinkbots are free, as in 'free will'.

*

Korrelan

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1113
  • Look into my eyes! WOAH!
    • YouTube
Re: The last invention.
« Reply #265 on: January 29, 2018, 03:53:04 pm »
Soz for the delay… I’ve been a bit busy.

Quote
What is the purpose of the Frontal Cortex affecting the ex. Visual Cortex?

The way you interpret what you are seeing at any given moment is affected by both your current state of mind and your past experiences.  As I mentioned earlier your brain fills in/ simulates most of what you experience as reality, to do this there has to be a blending of the limited visual pattern/ stream with the ‘imagination/ experience’ patterns.  Your connectome is constantly guiding/ influencing what you think you are seeing. 

For example… if you where to drop a small screw on floor, you are able to mentally tune your visual system to heighten the detection/ recognition of the visual properties composing the screw.

Quote
Simply put, any activity in the brain should slowly fade away, but not instantly turn black. Do you/s do this?

Oh I get what you mean, like the neurons should stay primed after firing so the next flush of information is more likely to activate them, so a chain/ pattern can build up from the stream of incoming information… similar to the persistence of vision effect?

https://en.wikipedia.org/wiki/Persistence_of_vision

If you think about the firing schema of any type of artificial neuron it relies on a set firing methodology, adding a slowly decaying activation threshold would negate any kind of inbuilt ‘logic’ that tests for a firing level/ threshold.  The essence of an artificial neuron whether it be a classic weighted/ sigmoid or a spike derivative relies on the accuracy of the firing mechanism, adding a slowly decreasing activation function would totally mess up the logic/ pattern matching abilities of the system.

Quote
list of where you learnt all of your knowledge from?

Harvard… haha… they would have thrown me out lol. I’ve never taken any computer science courses; I did write computer course syllabuses and teach them though so I guess that counts...

Ok seriously?

If I remember correctly you’re in your twenties… I’m 52 this year… so obviously age/ time has a huge bearing on my knowledge/ insights (for what they are worth).

I really don’t know what to advise, except look how far you have come, how much you have learned in the past few years.  Even I will admit the current Locksuit is a vast improvement on Advancessss or the King lol… this shows you are very capable of learning and improving you knowledge/ insights… just keep improving… if my generation fails to produce a AGI humanity is going to need a new fresh outlook.

It takes time for a human to integrate knowledge and skills, I suppose the single most important piece of advice I can give is… make sure you understand what you are reading.  If something doesn’t make sense then re-read, research the base topics, keep going until you have a thorough understanding of the topic.

I’ve always been interested in anything to do with engineering/ science/ psychology; I read hundreds of articles every week on a very wide range of subjects, neuroscience, robotics, electronics, etc. I even read sites that deal with religion, UFO’s and ghosts; not because I believe in them but because I need to understand what makes these people tick, why they believe what they do… its all important.

We are all unique, we all have our own particular skill sets… that’s good.  That’s why the human race has come this far, we each approach a problem space form a different angle.

Quote
I seem to have the Spatial Sequence Synesthesia.

We all have schizophrenia, synesthesia, illusions, paranoia, etc its part of what makes us unique and human. Some people have more than their share of a particular trait though, and this means they don’t fit into our societies ‘normal’ accepted scope/ range.

 :)
« Last Edit: January 31, 2018, 09:15:16 pm by korrelan »
It thunk... therefore it is!... my project page.

*

LOCKSUIT

  • Emerged from nothing
  • Trusty Member
  • *************
  • Transformer
  • *
  • 2165
  • First it wiggles, then it is rewarded.
Re: The last invention.
« Reply #266 on: January 30, 2018, 05:50:17 am »
"Simply put, any activity in the brain should slowly fade away, but not instantly turn black. Do you/s do this?"

It's exciting that you understood me, and I understood your reply back today! I see! It would mess up the thresholds. Well, after I stare at lots and lots of anime girls, or pikachu, random swirls and things yellow will look like I just seen pikachu or anime girls, just like swirls in carpet/etc *naturally* look like faces being seen. After playing Pikmin I also almost see pikmin running along things in my house. Also, notice how when you look around, you are kinda like a video? Your life/sensory is like a 8 second video of the last 8 seconds. I call this effect VideoSense. What may happen is multiple things get selected and higher representations are selected (or fuller energizing of something not easily selected) and then "seen" by "getting a look around the hallway for 8 seconds". Each feature has a short-term memory neuron that strengthens/weakens faster than long-term memories.

So you're saying if you recognize a screw, and drop something, this selects frontal cortex representation "n", which send signals to the visual cortex's screw representation features? Ya but, the very thing I brought up above (yellow blobs looking like I just seen pikachu) is how, no? -> You see the screw (or pick one up without seeing it but seen it internally as a memory selected, or even ask yourself what you dropped), and then when you drop it, you feel this, and do "look at ground actions", and the features of the screw are partially activated...
« Last Edit: January 30, 2018, 06:12:37 am by LOCKSUIT »
Emergent

*

raegoen

  • Roomba
  • *
  • 5
Re: The last invention.
« Reply #267 on: January 31, 2018, 06:24:41 pm »
Hey LOCKSUIT, I think what is happening in the situation you described when you look at something yellow and it looks like Pikachu is that your brain is constantly trying to match your sensory input (in this case sight) to a pattern it has already encountered before, in this case it reminds you of Pikachu especially since you said you watched it prior to this experience, that memory/pattern is closer to your consciousness so your awareness is drawn to that particular memory.

When you drop a screw and try to find it, you're priming your brain to focus on and respond strongly to the pattern of the screw in your sensory receptors (again this being sight) by recalling what the screw looks like, and creating a fake reality in your brain in which you've already found the screw, then when reality matches the fake closely, you know you've found it. I noticed something interesting in a similar experience I've had, I was recently playing with my younger brothers building lego. They of course have a huge bin full of random lego pieces, and very often I was looking for two specific pieces at a time. However when I focused on finding both at the same time, my ability to find them decreased significantly compared to when I focused on finding just one at a time. This shows how the human brain has limited multitasking abilities, if you want to do a job well, you have to focus and give that one task your undivided attention.

*

LOCKSUIT

  • Emerged from nothing
  • Trusty Member
  • *************
  • Transformer
  • *
  • 2165
  • First it wiggles, then it is rewarded.
Re: The last invention.
« Reply #268 on: February 01, 2018, 06:18:50 am »
Hey korrelan I noticed there is many cortical columns in the human brain, are each a feature detector? I thought you have ex. 50 features detectors in layer 1, then ex. 40 higher-level representations/features in layer 2 ex. eye/nose...and layer 6 faces/scenes/complex groups ex. dining table made up of many things. Again, SIX layers....now internet tells me thhere is thousands of these "6 layer" nets......??.....or, is it just 1 six-layer? simply the perpendicular facing cortical columns are all part of the 6 layer sheet? And each is a feature detector mini-processor?

Reason I'm confussed is because if the whole neocortex was 1 "6 layer" sheet, containg allll things ex. faces horses lines tables motion etc, then each neuron is ....................oh wait, ex. 8 features means 8 clone networks right? Hence many cortical columns.....acting each as a "neuron" making up the whole neo-sheet hierarchy correct? Ex. nose column+eye column=face column in "layer3"?

« Last Edit: February 01, 2018, 08:32:45 am by LOCKSUIT »
Emergent

*

Bob

  • Roomba
  • *
  • 3
Re: The last invention.
« Reply #269 on: February 02, 2018, 12:48:28 am »
Wow, korrelan, you really have something going on with this project. :o

LOCKSUIT, for as far as I understand it, the six layers in a cortical column are just the way how the column processes information. Different parts of the neocortex are responsible for lower/higher level processes but a single column doesn't do everything from line detection to face recognition. It does just a single thing. Columns are connected through the thalamus which acts as a "relay station" between them. Columns are also interconnected in other ways. So some low-level information goes through the thalamus to one part of the neocortex, gets processed (e.g. line detection), something is sent back to the thalamus, then sent to a higher level part of the neocortex (e.g. to do more complex detection). Then rinse and repeat? I haven't really looked at how the brain handles visual information but I would guess it works in similar ways for different sensory input.

I like the following image. I think it illustrates this quite well.
https://goo.gl/images/LPHnmJ
I'm not sure what the paper this image comes from says about it because I don't have access.

Thank you for the video. I learned something from it. It's a bit weird to see a cat that way though.

 


Users Online

79 Guests, 1 User
Users active in past 15 minutes:
DemonRaven
[Trusty Member]

Most Online Today: 109. Most Online Ever: 208 (August 27, 2008, 09:36:30 am)

Articles