[agi] Re: Big machine learning scientist embraces consciousness in machines

Conscious entity embraces consciousness. Hmmm wonder why. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tb2b962ade3189dc2-M6b899ef72f81e5c1fdf08600 Delivery options:

[agi] Re: The KC of Permutations

On Saturday, December 28, 2019, at 5:43 PM, James Bowery wrote: > Is there a higher level language -- the language of permutations -- in which > one can describe them? In this case estimating KC is minimizing representational complexity across many paths in a spectrum of reversible transforms

[agi] Re: Real AGI Brain

It could cost me up to a 1,000$ so although I do have some hundreds it may put me in a bit of a hole and why do that... So small donations would help ex. 30 here 100 there. -- Artificial General Intelligence List: AGI Permalink:

[agi] Re: Real AGI Brain

On Saturday, January 04, 2020, at 1:37 AM, rouncer81 wrote: > I dont think a cell would ever self ignite,  wouldn't it always be a pattern > off the sensor. > How is something supposed to self ignite, it goes against determinism it > seems, so it could be out-ruled! > Do you know of another

[agi] Re: Real AGI Brain

I can make text generation as well remember, so actually yes, yes and yes.  But I dont know how gpt-2 worked until you told me tho,  its got a certain style to it,  mine is different. -- Artificial General Intelligence List: AGI Permalink:

[agi] Re: Real AGI Brain

Let's take a deep layer architecture. Every node is connected to every node in every layer. Nodes are representations that can be activated by multiple very different inputs, after training. Bottom nodes are a-z alphabet inputs+outputs. So what happens if you hear 'hi there'? Well these 7 nodes

[agi] Re: Real AGI Brain

If no one says otherwise, I'm gonna do it. I'm goin for it starting today. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T409fc28ec41e6e3a-Mcc8a5152cf1fd06a1c8e617a Delivery options:

[agi] Re: Real AGI Brain

if it could chat it would impress me most,   a good thing to think of is what is the purpose of the robot speaking? is it trying to get you to buy something? is it trying to say funny jokes? is it getting you to understand something? (thats a hard one.) It seems to be basic manipulative

[agi] Re: Real AGI Brain

It loves/seeks: More big diverse data by collecting/generating it. More compressed hierarchy (node count Cost (error, Hutter Prize shows lossless compression leads to less data but can re-generate all back still). More related data to goal nodes. It wants to come to equilibrium, there would

[agi] Re: Real AGI Brain

It wants to compress, it wants to expand, it wants to come to equilibrium to get goals. To get goals it must find the maze path using data, which lead it to no problems (dead ends). -- Artificial General Intelligence List: AGI Permalink:

[agi] Re: Real AGI Brain

I'm reading https://lena-voita.github.io/posts/emnlp19_evolution.html -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T409fc28ec41e6e3a-M25bcc7c6c5fb6a41102feda0 Delivery options:

[agi] Re: Real AGI Brain

"We also view every model as learning a function from input X to output Y." All we need do is compress the net by making sure it is a small world net where each word explain each other enough, same activations. It's not mapping just 1 input to output, all possible mutations are mapped.

[agi] Re: Real AGI Brain

"Information about input is getting lost, information about output is accumulated" Yes some nodes are missing/combined and by the end of the leaking it gains more context weight from all context so [w w w w w w w w w w] describe w w w [w] w w w w w w. Mine looks at both sides all context and

[agi] Re: Real AGI Brain

So all similar nodes are mapped while preserve the correct nodes/paths. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T409fc28ec41e6e3a-M8f1d8a157ee454344784187d Delivery options:

[agi] Re: Real AGI Brain

They use words I use, evolution, extract, compress -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T409fc28ec41e6e3a-M69d8e3f22e11169a39a4ce0b Delivery options: https://agi.topicbox.com/groups/agi/subscription

[agi] Re: Real AGI Brain

"representations of individual tokens, not the entire input" Each part of a node is mapped/pathed. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T409fc28ec41e6e3a-M58a967e0a20031eaa3f72276 Delivery options:

[agi] Re: Real AGI Brain

But if look at the upper activated nodes or write them, yeah it looks like each is adjusted based on all. See, the Authors say frequent words change others less and are changed by others more, btw frequent Next Tokens are chosen and also chosen based on rare story tokens. The rare tokens

[agi] Soundbites

https://www.youtube.com/watch?v=UfezfehUBEQ https://www.youtube.com/watch?v=TuKDBeVSBoo -- Artificial General Intelligence List: AGI Permalink:

[agi] Re: Real AGI Brain

I feel like these authors are missing simple things, I mean its nonesense. LM and MT and MLM are all doing translation, the entailment recovers the distant knowledge of frequency and relation to story words as I said. -- Artificial General Intelligence

[agi] Re: Real AGI Brain

My acquaintance is refactoring my previous text generator: https://github.com/AlphaGit/bladesupernovas-text-generator/blob/master/TextGenerator.py He likes it. I'm unsure what he is doing. He said he is fairly busy though. -- Artificial General Intelligence

[agi] Re: Real AGI Brain

He put up a demo, i think he'll set it up right soon: https://colab.research.google.com/drive/1wzswci20wekoIXyI_eEw-_7JiYk17A5d My text generator has been implemented a year ago and it's not even perfectly done and works, so I know what it is really doing inside.