Robust Efficiency With Successive Approximations

  • 28 Replies
  • 4189 Views
*

HS

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1178
Re: Robust Efficiency With Successive Approximations
« Reply #15 on: September 28, 2019, 11:35:13 pm »
That makes sense, someone else would have to create IKW themselves, from D, then communicate what they have realized to you. Still, I would say that, at first, those would be zipped files, and some work would have to be done to integrate them into your perspective, before they could serve the same function as genuine information, knowledge, and wisdom.

I just made BEAR up. Just imagining the flow of reactions or data through a person. First things happen in the environment outside the body. The first things to encounter this data are the senses, (skin, eyes, ears...). This stream of data would have the most volume, because you can't create more data, and the senses would narrow it down to concentrate useful things. So the first stage is data being filtered by the body (B).

Then, when I feel something with my senses, I feel a corresponding vague opinion, a value judgement. Therefore (E).

This (E) motivates "me" awareness (A), to take an action. And I, (Awareness), employ reason (R) to figure out a good action.

So you get this chain of things that are dependent on each other. The way I see it, R(A(E(B))) are functions of each other. The stream of info gets filtered and used for things at each stage, so it gets narrower until reason (R) makes 1 choice. Therefore B is the bottom of the pyramid, and R is at the top. This process happens fairly fast; eg "I open the door, a cold wind passes right through my thin shirt, it feels freezing, I feel displeasure, my awareness stretches out for a solution, it settles on logic (from experience, my little brother would just stand there very aware of his displeasure, lol), logic runs through cause and effect typically one or two step, like Jacket=Shield, and that's it.

GCRERAC is not my idea. It's a tried and true linear progression (unless the person is insane, or a step is skipped because the goal is small) of the way people go about things. Used by writers to give characters verisimilitude and believability. That's basically enough for me to run with it, just thought I'd adapt it for AI. Jim Butcher is a writer who's familiar with this, and has all the details described in his blog : https://jimbutcher.livejournal.com/ (Scroll down to, Scenes December 2006, Sequels December 2006)

*

AndyGoode

  • Guest
Re: Robust Efficiency With Successive Approximations
« Reply #16 on: September 29, 2019, 07:38:44 pm »
I updated my chart with some minor additions, some of which are based on what you mentioned.

What you called B for body I'm calling P for physical for physical senses and filtering. I know of two filters in the body/brain, one firmware that is normally not changeable except under the effect of DMT or psilocybin, therefore the mushroom icon influencing that filter, and one software that the brain changes, such as predisposition to recognize certain objects or attention focusing mechanisms.
Reflexes, like knee jerk upon the knee being tapped or eye blink upon sudden bright light hitting the eye, bypass cognition in the interest of safety via speed, so there exist hardwired responses for those, analogous to interrupts in an operating system.
I made Maslow's hierarchy more explicit via an added key.
I'll keep updating that diagram as I have time, interest, and clarity in what you wanted to add.
« Last Edit: September 29, 2019, 07:59:06 pm by AndyGoode »

*

HS

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1178
Re: Robust Efficiency With Successive Approximations
« Reply #17 on: September 29, 2019, 08:19:00 pm »
"PSLEA" Another one!  ;D These are like cocaine for me... Thanks for the diagram, I'll refer to it when I try to do version 2.0 of mine.

*

AndyGoode

  • Guest
Re: Robust Efficiency With Successive Approximations
« Reply #18 on: October 01, 2019, 04:46:05 am »
Thanks, H.S. and Korrelan.

Here's an updated, more detailed version of yesterday's chart.

Adding memory makes it more complicated. Maybe I should put the two pink triangles inside of memory, too--I'll have to think about it later.
The two big pink triangles exist because two copies of the world are stored, and those copies of the world contain emotional content and physical content as well as factual/intellectual content. That's where the system will feel emotion or pain, if it has any.
H.S.'s BEAR is treated like this--B for the body filtering is handled by the various filters shown, and what is left, the E-A-R, is basically emotional and intellectual, already implied by the pink triangles, and the reason, or R, is handled by the reasoning module. I took out the mushroom effect on the firmware filter because a very large number of things are filtered by the body, so that filter is treated more generically now since things other than mushrooms can affect it. That particular mushroom effect was only conjecture, anyway, as we all discussed in an earlier thread about the supernatural. But at least the previous diagram was stoner friendly.  ;)
The two yellow triangles are the same as in my previous chart, but with their interior details suppressed. Note that one thing that happens is there is now a pink hierarchy of pink hierarchies, since each yellow triangle contains pink triangles as well as being embedded in a pink triangle.
I'm pretty sure all the functionality of the previous diagram is present in this one, too, except sometimes there exists an extra processing step to get to the correct destination.
The Maslow hierarchy should be modified to be more generic for both people and machines, but that's about another day of thought in itself.

*

HS

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1178
Re: Robust Efficiency With Successive Approximations
« Reply #19 on: October 01, 2019, 05:34:52 am »
Cool. I also did some updates. Managed to integrate BEAR into the rest of the system, which freed up Awareness to travel around the loop, and got rid of some redundancies. Then I added Maslow's hierarchy, and broke D's influence over C, as on second thought it didn't make much sense. I assume Anticipation (A) will include a time component, which could then be combined with PSLEA, to make a sensible choice.

IPM3" border="0

*

AndyGoode

  • Guest
Re: Robust Efficiency With Successive Approximations
« Reply #20 on: October 02, 2019, 01:58:49 am »
which freed up Awareness to travel around the loop

I have yet to deal with Awareness. I probably won't, until I can publish some ideas I have on that topic.

Here's today's update from me...

The most important addition is the prediction mechanism with the purple arrow as related to Jeff Hawkins' book "On Intelligence." He called the brain an "organ of prediction," so that's possibly the most important function the brain does, so I felt I had to show that function.
I separated the key from the main diagram since both were getting so large.
I showed a couple places where something like the Scientific Method takes place.
I decided not to deal with GCR-ERAC since that deals with story creation, and in my diagram the system is doing analysis of what already exists, not creation. I haven't gotten to creativity yet, and I might not.

*

AndyGoode

  • Guest
Re: Robust Efficiency With Successive Approximations
« Reply #21 on: October 04, 2019, 01:14:49 am »
Here's my latest update.

Mostly I streamlined the readability of the main diagram by hiding memory as a big blue layer parallel to the big white processing layer. That way I didn't have to keep drawing arrows across the diagram every time something had to go into or come out of memory. That's the way it is with neural nets, which our brain is, anyway--the links serve as both memory and processors, so memory and processing are extremely closely coupled in neural networks.
I also added an internal clock, which is needed by both animals and computers. The exact way biology makes those is not fully known, although a lot is now known about the internal clocks for circadian rhythms, which are 24-hour timers, which are on a longer time span than is needed to time computations in the minutes range, for example.

https://en.wikipedia.org/wiki/Time_perception
https://en.wikipedia.org/wiki/Suprachiasmatic_nucleus
https://www.nigms.nih.gov/education/pages/Factsheet_CircadianRhythms.aspx

Here's an example of how the diagrammed architecture would work...
Suppose you're playing a game of chess and you're getting low on time. You can't just call a Stockfish-like subroutine to give you a list of the best moves because you can't set the time on that program and it could take a long time to produce the desired list. Therefore you need to post a new goal, on the fly, that says you want a satisfactory move within one minute, say. The reasoning module updates the goals in the intellectual layer of the virtual world state, with a conjuncted goal like 'produce satisfactory move' AND 'finish within 1 minute', then begins reasoning on that move. By monitoring memory, to which the clock state is continually sent--note the down arrow icon next to the clock, which means that information is sent to memory--the reasoning module knows when time is up, whereupon it will use the best move found so far, activate its mechanical fingers via the actuators signal, or maybe just move an icon on a computer screen via electronic output, and make that move.
« Last Edit: October 04, 2019, 09:32:05 pm by AndyGoode »

*

goaty

  • Trusty Member
  • ********
  • Replicant
  • *
  • 552
Re: Robust Efficiency With Successive Approximations
« Reply #22 on: October 04, 2019, 04:44:19 am »
That's the way it is with neural nets, which our brain is, anyway--the links serve as both memory and processors, so memory and processing are extremely closely coupled in neural networks

That's the meta-data thing,  meta-code.
But all of computers is really that tho,   because truth tables are data that is processing as well.  its a bit of a bad concept, but can bring about good ideas anyway.

My special "processing tree" system is the same,   im writing all this code which actually stands for code :),  I found it a little disorientating, because I couldn't snap out of seeming that my code was just comments as I was doing it!  But code is just comments.  Code standing for code,  data data data, driving me nuts.

Writing an actual NES game inside a NES emulator library would be similar I think. going through 2 levels of abstraction as you write it...  :) weird.


*

HS

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1178
Re: Robust Efficiency With Successive Approximations
« Reply #23 on: October 04, 2019, 10:23:38 pm »
I'm trying to complete my next update but my computer isn't handling it very well. I might have to transfer the modeling program to an SSD. Which, from a certain perspective is promising, seeing as the whole point of physical neural nets, is that they won't have issues with calculating/modeling themselves.

*

goaty

  • Trusty Member
  • ********
  • Replicant
  • *
  • 552
Re: Robust Efficiency With Successive Approximations
« Reply #24 on: October 04, 2019, 11:15:03 pm »
I'm trying to complete my next update but my computer isn't handling it very well. I might have to transfer the modeling program to an SSD. Which, from a certain perspective is promising, seeing as the whole point of physical neural nets, is that they won't have issues with calculating/modeling themselves.

you don't need to loop your cells if you have them as a physical connection,  and electronics isn't hard, its just hard to believe in.    If you get better at doing it, best keep it to yourself, because as far as I can tell, they don't like showoffs on electronics forums, they are serious Q&A only,  but showoff to me anytime... this forum seems to be friendly that way.  that's why I stick around here, cause the rest of the internet doesn't want me.

*

HS

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1178
Re: Robust Efficiency With Successive Approximations
« Reply #25 on: October 04, 2019, 11:58:51 pm »
Their loss. Though I do have to check out their serious Q&A at some point, it might be funny.  :)

*

AndyGoode

  • Guest
Re: Robust Efficiency With Successive Approximations
« Reply #26 on: October 05, 2019, 12:25:07 am »
I'm using only OpenOffice Draw, which is free but it started aborting when my diagrams got more complicated as in the more recent diagrams above. Oh well, it's better than Paint and other free drawing programs I've tried.

I had to look up SSD. For other folks unfamiliar with that term...
https://en.wikipedia.org/wiki/System_sequence_diagram

I suspect I'm near the end of the updates of these diagrams that show my understanding of the brain and intelligent systems, since I don't believe they will go usefully into more detail. I believe the top levels of the red triangles need to be placed into memory, and Maslow's hierarchy's contents need to be upgraded to include machines, but I need to think about those issues a while first.

P.S.--Some robot diagrams that are similar to the one I created are these...
https://image.slidesharecdn.com/roboticsppt-1-161113171637/95/robotics-ppt-1-17-638.jpg
http://www.personal.reading.ac.uk/~shshawin/LN/images/pictures1-16.jpg
http://3.bp.blogspot.com/_ZGzaqHb40vU/TE_y5tMPkQI/AAAAAAAAAJY/YD2IwszZlG4/s1600/Perception.jpg
https://image.slidesharecdn.com/facerecognitionusinglaplacianfaces-140823024208-phpapp01/95/face-recognition-using-laplacianfaces-18-638.jpg
http://docsdrive.com/images/ansinet/itj/2007/fig1-2k7-1043-1049.jpg
« Last Edit: October 05, 2019, 01:17:06 am by AndyGoode »

*

HS

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1178
Re: Robust Efficiency With Successive Approximations
« Reply #27 on: October 05, 2019, 12:47:00 am »
Yeah, I didn't think I could take it any further in paint, so I began modeling a brain made from my neuron cubes, using inventor (like cad). But the parts got into the thousands and the program would stop responding and shut down. I only meant Solid State Drive by SSD.  ;D.

*

goaty

  • Trusty Member
  • ********
  • Replicant
  • *
  • 552
Re: Robust Efficiency With Successive Approximations
« Reply #28 on: October 05, 2019, 01:31:13 am »
Yeah, I didn't think I could take it any further in paint, so I began modeling a brain made from my neuron cubes, using inventor (like cad). But the parts got into the thousands and the program would stop responding and shut down. I only meant Solid State Drive by SSD.  ;D.

Yeh that's what I thought u meant!  =)   haha  SSD.

Just imagine how weird the gizmos that run an animal are,  then realize how restricted and uncreative the tech is on the electronics forums,  I say throw it in the bin, do something new.  A simple logic gate can be literally anything in the whole scheme of the universe.

 


Requirements for functional equivalence to conscious processing?
by DaltonG (General AI Discussion)
November 19, 2024, 11:56:05 am
Will LLMs ever learn what is ... is?
by HS (Future of AI)
November 10, 2024, 06:28:10 pm
Who's the AI?
by frankinstien (Future of AI)
November 04, 2024, 05:45:05 am
Project Acuitas
by WriterOfMinds (General Project Discussion)
October 27, 2024, 09:17:10 pm
Ai improving AI
by infurl (AI Programming)
October 19, 2024, 03:43:29 am
Atronach's Eye
by WriterOfMinds (Home Made Robots)
October 13, 2024, 09:52:42 pm
Running local AI models
by spydaz (AI Programming)
October 07, 2024, 09:00:53 am
Hi IM BAA---AAACK!!
by MagnusWootton (Home Made Robots)
September 16, 2024, 09:49:10 pm
LLaMA2 Meta's chatbot released
by spydaz (AI News )
August 24, 2024, 02:58:36 pm
ollama and llama3
by spydaz (AI News )
August 24, 2024, 02:55:13 pm
AI controlled F-16, for real!
by frankinstien (AI News )
June 15, 2024, 05:40:28 am
Open AI GPT-4o - audio, vision, text combined reasoning
by MikeB (AI News )
May 14, 2024, 05:46:48 am
OpenAI Speech-to-Speech Reasoning Demo
by MikeB (AI News )
March 31, 2024, 01:00:53 pm
Say good-bye to GPUs...
by MikeB (AI News )
March 23, 2024, 09:23:52 am
Google Bard report
by ivan.moony (AI News )
February 14, 2024, 04:42:23 pm
Elon Musk's xAI Grok Chatbot
by MikeB (AI News )
December 11, 2023, 06:26:33 am

Users Online

248 Guests, 0 Users

Most Online Today: 280. Most Online Ever: 2369 (November 21, 2020, 04:08:13 pm)

Articles