Yes.  that's the "mental bound" of AF members, lol.  Your talents are wasted here.  In fact........maybe you should go someplace where the value of your colorful text and argumentative brilliance won't go completely unacknowledged....and buddy, they're going completely unacknowledged here.

(November 7, 2016 at 9:38 am)Mathilda Wrote:

(November 7, 2016 at 8:33 am)chimp3 Wrote: I recently read a book on AI and it said that we have yet to produce anything as intelligent as the insect. I am not very knowledable on this subject but this does not merit god status to me. Nature does a much better job with random mutation and natural selection.

Eric Kandel spent a large part of his career mapping out each of the 18,000 neurons of sea slugs in order to understand memory formation. He won the Nobel prize for it. A fruit fly has about 250,000 neurons. Compare this with the human brain which has about 86 billion neurons. Each neuron is more than a simple integrating unit that most artificial neural networks use. Each neuron has significant computational power compared to an entire artificial neural network. For example, dendritic trees can encode binary logic.

What ProgrammingGodJordan hasn't taken into account is though that any strong AI, any artificial general intelligence needs to be self organising.

To explain this, think of what computers are and what they can do.

They're super fast idiots.

They need explicit instructions in heavily constrained environments without any noise. Absolutely everything has to be made explicit. Animals and humans on the other hand are autonomous and do not need to be micro-managed. The field of AI is about taking this ability of natural intelligence and giving it to computers to make them more autonomous. You could create a program for a robot to build a car. Everything would have to be in the exact place because if something is even slightly off then the robot will fail and won't be able to correct itself. AI would allow it to adapt, like how a human worker does not have to care where the screwdriver is, they see it, pick it up, rotate it and then use it.

But here's the difficulty. How do you encode explicit instructions for noisy environments that you yourself do not know about in advance? The more explicit your instruction, the less robust it is to the noisy real world. Whereas if a system is completely self organising then it is shaped solely by its environment. A wholly self organising system can then adapt to other environments without depending on any instructions explicitly coupled to one specific environment. If all a neural network adapts to is a continuous signal that changes over time for example, then it does not matter what's causing that signal to change.

Take humans and animals for example. No one opens up our heads and directly injects an electrical current into our brains in order to teach us. We learn from our environment. We sense it, we interpret those senses, choose an action, act within an environment and then sense that the changes in that environment. We are part of a sensory / action loop. We are a part of our own environment.

But this opens up another can of worms, how do you engineer self organising systems? To put this in context, a good analogy for electricity and voltage is to think of water and pressure. So using nothing but water pumps, pipes, cisterns and valves, how would you arrange them in a network to naturally adapt to the water that you feed into it? How would you set it up to remember previous water pressures? Or to regulate its water pressure over time to remove spikes and cope with reduced input? Or to select the best source of water? All without any explicit instructions to do so. After all, a simple model of a neuron builds up voltage over time from input signals until it reaches a threshold and fires. This is a glimpse of the challenge that lays ahead in AGI.

Now we're moving away from computational and informational systems but physical systems. In fact most research in the field of self organisation is still performed by physicists. But that's what the brain is. A wholly self organising biophysical system. And if you want an artificial general intelligence that robustly adapts to new environments, you need to remove explicit encoding and instructions.

[*A*]
See Google deepmind's alpha go/Atari q player.

Neural models SELF ORGANIZE on the horizon of GRADIENT DESCENT bound calculations, as some cost sequence minimizes error on input signals.

Such are non trivial models that like humans, LEARN ENVIRONMENTALLY. 

Bi dimensional games via atari q, are a microcosm of real life. (2d games, though of low resolution, provide non trivial variation, in the task sequence analysed)

These models take:
(0) Pixels (as humans do)
(1) Controller access absent explicit reward mapping (as humans fair)
(2) Reward notation /score (as humans fair)



[*B*]

Mathilda, likely lacks accurate comprehension of said field, (as observed amidst her ignorant responses).

Rather, it is perhaps pertinent that you observe Geoffrey Hinton's stipulations regarding phenomena abound 'AlphaGo'. Hinton is crucial (amongst Few) in the generation of modern neural bound machine learning.


[*C*]

By extension, I have constructed complex neural models, that have derived as high as 76/500+ amidst the international kaggle scale.
 
In addition, on comprehension of quantum computing, deep reinforcement learning, and causal learning (uetorch tower blocks) I had come to compose non trivial general approximations. (See 'THOUGHT CURVATURE').

My stipulations are thereafter of thorough nature, whilst Mathilda's are narrow in content.


[*D*]

OBSERVE a PRIOR response of mine. Mathilda's longueur refers to class '...A', in the facing RECOLLECTION:

The paths via machine learning faces:

'...A' Hardware replication. [From  C. elegans worm neuronal replication, to Henry Markram's mammalian brain constructs]

'...B' Theory-bound compactions of neuronal cognition [Deep Neural Networks-aligned models, Convolutional Neural Networks, Recurrent Neural Networks, Long Short Term Memory, Deep Reinforcement Learning that utilizes Deep Neural Nets...]


Therein:

[0] - '...A' is but entirely non-human surpassing. [approximates worms, partial mammalian brain as observed in '...A']

and

[1] '...B' has already produced models that exceed human performance, in cognitive, non trivial tasks.[Although the human brain oscillates abound higher efficiencies entirely, brain based models excel in individual tasks/task groups as observed in '...B'.]

---

(November 7, 2016 at 5:10 pm)Rhythm Wrote: Yes.  that's the "mental bound" of AF members, lol.  Your talents are wasted here.  In fact........maybe you should go someplace where the value of your colorful text and argumentative brilliance won't go completely unacknowledged....and buddy, they're going completely unacknowledged here.

Your failure:

(0) Said statistics are but not of my invention.

Your non failure:

(1) It appears you have come to purge a degree of your ignorance, amidst the reduction (of mine) of your priorly sillily scribed sequences...

As such, (1) is, as I intend.

(November 7, 2016 at 3:54 pm)ProgrammingGodJordan Wrote:

(November 7, 2016 at 10:19 am)Mathilda Wrote: It's interesting that PGJ mentions Jeremy England for some unexplained reason yet England's research is very much related to what I was talking about with regards to self organisation.

Dissipative adaptation applies amidst matter (and likely adequate SIMULATION par such) whence said matter attributes life like properties as time diverges.. (AS I EXPRESSED PRIOR)

Albeit, in the attribution of like like properties, self-organization is quintessentially observed...

Time diverges from what? You're sort of saying the right words but putting them all in the wrong order. For example, life like properties emerge from self organisation, not the other way round.

Dissipative structures are thought to be responsible for the increase in complexity over time. This is because self organisation works by minimising free work, that is energy that can do work, and one way that the energy can be minimised is by being dissipated. My own personal findings which I really need to publish sometime are that dissipative structures per se are not enough for both complex and dynamic systems to be created.

(November 7, 2016 at 6:06 pm)Mathilda Wrote:

(November 7, 2016 at 3:54 pm)ProgrammingGodJordan Wrote: Dissipative adaptation applies amidst matter (and likely adequate SIMULATION par such) whence said matter attributes life like properties as time diverges.. (AS I EXPRESSED PRIOR)

Albeit, in the attribution of like like properties, self-organization is quintessentially observed...

Time diverges from what? You're sort of saying the right words but putting them all in the wrong order. For example, life like properties emerge from self organisation, not the other way round.

Dissipative structures are thought to be responsible for the increase in complexity over time. This is because self organisation works by minimising free work, that is energy that can do work, and one way that the energy can be minimised is by being dissipated. My own personal findings which I really need to publish sometime are that dissipative structures per se are not enough for both complex and dynamic systems to be created.

[*A*]
Your comprehension of English, is rather sub par.

I shall bi-stipulate said sentence [en variation], such that you may trivially observe of the non difference in meaning:

(0) Initial sentence:
Albeit, in the attribution of like like properties, self-organization is quintessentially OBSERVED...

(1) Alternate sentence (of precise non difference):
Albeit, self-organization is quintessentially OBSERVED in the attribution of like like properties.

Thereafter, order is irrelevant.

[*B*]

NOTE:
Neural models SELF ORGANIZE on the horizon of GRADIENT DESCENT bound calculations, as some cost sequence minimizes error on input signals.

Neural models' neuronal/synaptic values are NATURALLY updated (self computed), absent human intervention at said particular juncture; p(x|y).

[*C*]
Diverge definition:

(of a series) increase indefinitely as more of its terms are added.

Big fish/little pond synd?

(November 7, 2016 at 5:40 pm)ProgrammingGodJordan Wrote: Neural models SELF ORGANIZE on the horizon of GRADIENT DESCENT bound calculations, as some cost sequence minimizes error on input signals.

Such are non trivial models that like humans, LEARN ENVIRONMENTALLY. 

Bi dimensional games via atari q, are a microcosm of real life. (2d games, though of low resolution, provide non trivial variation, in the task sequence analysed)

These models take:
(0) Pixels (as humans do)
(1) Controller access absent explicit reward mapping (as humans fair)
(2) Reward notation /score (as humans fair)



[*B*]

Mathilda, likely lacks accurate comprehension of said field, (as observed amidst her ignorant responses).

(November 7, 2016 at 5:40 pm)ProgrammingGodJordan Wrote: My stipulations are thereafter of thorough nature, whilst Mathilda's are narrow in content.

Yeah that confirms what I suspected. You really should have paid attention to my recommendation in the shoutbox to read Christoph Koch's Biophysics of Computation. The book makes it quite clear that the kind of artificial neural networks that you are referring to are computational models inspired by the brain. Real neurons don't actually work like that. Because of how frequently you make comments about people not having written a neural network, I'm guessing that you wrote some kind of backprop ANN once (or called one of the many routines written for you in some package or library), trained it on some data and now think that you are some kind of expert. What you think you are talking about is most certainly not computational neuroscience.

Here's a tip. Leave the ego out of this.

(November 7, 2016 at 6:32 pm)mh.brewer Wrote: Big fish/little pond synd?

THEIST likely accurately describes your religious stance, on observation of your irrelevant, nonsensical anecdotes...

---

(November 7, 2016 at 6:36 pm)Mathilda Wrote:

(November 7, 2016 at 5:40 pm)ProgrammingGodJordan Wrote: Neural models SELF ORGANIZE on the horizon of GRADIENT DESCENT bound calculations, as some cost sequence minimizes error on input signals.

Such are non trivial models that like humans, LEARN ENVIRONMENTALLY. 

Bi dimensional games via atari q, are a microcosm of real life. (2d games, though of low resolution, provide non trivial variation, in the task sequence analysed)

These models take:
(0) Pixels (as humans do)
(1) Controller access absent explicit reward mapping (as humans fair)
(2) Reward notation /score (as humans fair)



[*B*]

Mathilda, likely lacks accurate comprehension of said field, (as observed amidst her ignorant responses).

(November 7, 2016 at 5:40 pm)ProgrammingGodJordan Wrote: My stipulations are thereafter of thorough nature, whilst Mathilda's are narrow in content.

Yeah that confirms what I suspected. You really should have paid attention to my recommendation in the shoutbox to read Christoph Koch's Biophysics of Computation. The book makes it quite clear that the kind of artificial neural networks that you are referring to are computational models inspired by the brain. Real neurons don't actually work like that. Because of how frequently you make comments about people not having written a neural network, I'm guessing that you wrote some kind of backprop ANN once (or called one of the many routines written for you in some package or library), trained it on some data and now think that you are some kind of expert. What you think you are talking about is most certainly not computational neuroscience.

Here's a tip. Leave the ego out of this.

[*A*]

Irrelevant.
I simply stipulate statistics.

I had long stipulated, of the brain based nature of deep neural networks.

Albeit, deep neural models have equaled/exceeded human performance on non trivial tasks, cognitive tasks.

[*B*]

(0) I have compacted scratch written (via basic java) neural models for letter recognition.

(1) I have composed non trivial models, SAMPLE: (utilizing residual neural network) for heart irregularity detection. (Deriving 76/500+ via international kaggle scale)
....

(November 7, 2016 at 6:32 pm)mh.brewer Wrote: Big fish/little pond synd?

Not even a big fish. His fragile ego means that he needs to believe that he's good at the subject but he doesn't actually understand much about it. I saw a lot of this on my Comp Sci degree as an undergrad. Young men would be boasting about how they got drunk and then went back and programmed something in C++.

No. I didn't understand it either.

(November 7, 2016 at 6:37 pm)ProgrammingGodJordan Wrote:

(November 7, 2016 at 6:32 pm)mh.brewer Wrote: Big fish/little pond synd?

THEIST likely accurately describes your religious stance, on observation of your irrelevant, nonsensical anecdotes...

Oh, ouch. And actually, based on what I and others can see in your behavior, (user name, flamboyance, attention seeking, dismissing what you can't defend) you appear to be the prototypical ego maniac with an inferiority complex. 

I could have been wrong. More like little fish/little pond synd.