Secret to AGI & ASI

  • 69 Replies
  • 11853 Views
*

Korrelan

  • Trusty Member
  • ***********
  • Eve
  • *
  • 1454
  • Look into my eyes! WOAH!
    • YouTube
Re: Secret to AGI & ASI
« Reply #15 on: March 28, 2018, 09:54:51 am »
Secret to interstellar travel.

I don’t see what the problem is; it’s obvious how to create a ship capable of interstellar travel.

You just need an air tight vessel that can provide life support for the human occupants, an engine capable of providing enough thrust to achieve faster than light speeds and a guidance system.

See… I’ve worked it all out, prove me wrong.

 :)
It thunk... therefore it is!...    /    Project Page    /    KorrTecx Website

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: Secret to AGI & ASI
« Reply #16 on: March 28, 2018, 02:20:07 pm »
@Berick
ivan.moony: What do you mean by "original decisions" and "true judgments"? I think I know, but I want to be sure we're on the same page.
I'm looking at judgements from the point of logic. There are three types of judgements (these are accepted facts):
  • false judgements like "A and not A". They are false in all interpretations (i.e. no matter if A is true or false)
  • true judgements like "A or not A". They ate true in all interpretations (i.e. no matter if A is true or false)
  • satisfable judgements like "not A or B". They are true in some interpretations (i.e. A shouldn't be true while B is false)
Atomic judgements (like A, B, C, ...) can be combined to form more complex judgements using logic operators.

Original decisions? I'm just thinking at loud here and these are merely speculations. I imagine ideas being somehow generated in our mind. This generator is what interests me. How does it work? Does it heavily depends on existing ideas, while combining their parts, or it is able to shoot completely original ideas, no matter of the current ideas in the memory. I can see "ideas" as logic judgements, but I think there are things that can't be expressed in pure logic, so we finally may need a more complete language than logic. Oh, I see decisions as ideas that are decided to be realized.

@Unreal
Don't be put down, you have a spark. Now you have to start a fire. Fill in the details and don't be intimidated by their virtual complexity. Keep in mind that the most genial ideas are the simplest ones, and invisible to the majority, but they are hardest to be thought of.

By the way, replacing NN with something faster and simpler seems like a good idea. Does planar binary search (quad search) ring some bells?

@Skeptics
You can hold a stone and say:
  • It's a rock
  • It's a piece of a mountain
  • It's a part of the Universe
What matters is what you see in that stone.
« Last Edit: March 28, 2018, 03:43:57 pm by ivan.moony »

*

unreality

  • Starship Trooper
  • *******
  • 443
Re: Secret to AGI & ASI
« Reply #17 on: March 28, 2018, 03:46:28 pm »
[Deleted]
« Last Edit: November 26, 2019, 05:16:23 pm by unreality »

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: Secret to AGI & ASI
« Reply #18 on: March 28, 2018, 03:57:11 pm »
You might need some formal method to describe a structure of observations/conclusions/ideas/decisions. What would be a structure of the tree you are talking about?

*

unreality

  • Starship Trooper
  • *******
  • 443
Re: Secret to AGI & ASI
« Reply #19 on: March 28, 2018, 04:00:04 pm »
[Deleted]
« Last Edit: November 26, 2019, 05:16:29 pm by unreality »

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: Secret to AGI & ASI
« Reply #20 on: March 28, 2018, 04:51:02 pm »
Well, I'm aware of systematic tree traversal methods. Also, I know there exist some optimized searches such are Monte Carlo, or algorithm A*, and similar, but I'm not familiar with details. Anyway, if something has a tree structure, it can be searched. But specific tree structure isn't god given, you have to specify it somehow.

For example how would you specify the following conclusion:
Code
all birds have wings
seagul is a bird
--------------------------
seagul has wings

Combining several conclusions in a row gives us entire process that could be considered as an algorithm in some cases.

What I want to say is that it is not "just a tree". It has to have some structure and atomic and compound elements, like constants, variables and operators. And I think discovering such structure elements is not such a trivial task. You might want to use some existing framework (mathematical logic or a kind of type theory), or you might want to discover your own. Anyway, the matter of a structure brings in a huge amount of complexity and it deserves a decent effort to investigate.

Basically, the process of tree search you describe is possible, but I'm telling that it because I know a few things about possible kinds of formal structures of knowledge represented by arbitrary expressions. I have to justify that it is easy to dismiss your ideas by someone with less experience in knowledge theories. I don't know how did you come to it, but recognizing your vision takes experience.

I'll pass you some links you might find relevant to your work. You might want to skim over term rewriting to discover the whole world beneath the surface I'm trying to describe, and you might want to take a look at combinatorial explosion, just to stay in touch with reality (there are methods like genetic programming, to partially avoid combinatorial explosion, if you are interested in digging further down the rabbit hole). For further details you would have to seek the web for PDF academic papers on meta knowledge and similar materia.

*

unreality

  • Starship Trooper
  • *******
  • 443
Re: Secret to AGI & ASI
« Reply #21 on: March 28, 2018, 06:14:18 pm »
[Deleted]
« Last Edit: November 26, 2019, 05:16:36 pm by unreality »

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: Secret to AGI & ASI
« Reply #22 on: March 28, 2018, 07:11:56 pm »
Quote
If you asked the AGI which of the above is correct, it would convert your English question into it’s own AGI language of logic, form a tree search, which has a specified goal formed in it’s own language, a simple programming language.

And now we come to the crucial question: what would be AGI's own language? All you say stands, with this or that tweak, but that language - how would it look like?

*

unreality

  • Starship Trooper
  • *******
  • 443
Re: Secret to AGI & ASI
« Reply #23 on: March 28, 2018, 07:26:00 pm »
[Deleted]
« Last Edit: November 26, 2019, 05:16:43 pm by unreality »

*

infurl

  • Administrator
  • ***********
  • Eve
  • *
  • 1365
  • Humans will disappoint you.
    • Home Page
Re: Secret to AGI & ASI
« Reply #24 on: March 28, 2018, 10:51:26 pm »
The language that you are talking about is called first order logic. There are numerous dialects and implementations available. It forms the basis of many of the most advanced AI projects such as Cyc which is a fifty year project entering its final stages. The example that you gave earlier can be expressed with this rule "(forall ?x (=> (isa ?x bird) (haswings ?x)))" which reads "for all x, if x is a bird, then x has wings". If among your facts you have "(isa seagull bird)" and you ask "(haswings ?y)" it will answer "(seagull)".

Here's a more complicated example. First some rules.


(equal ?x ?x)
(<=> (equal ?x ?y) (equal ?y ?x))
(<=> (and (equal ?x ?y) (equal ?y ?z)) (equal ?x ?z))
(not (parent ?x ?x))
(not (grandparent ?x ?x))
(<=> (or (father ?x ?y) (mother ?x ?y)) (parent ?x ?y))
(<=> (or (grandfather ?x ?y) (grandmother ?x ?y)) (grandparent ?x ?y))
(<=> (and (parent ?x ?y) (mother ?y ?z)) (grandmother ?x ?z))
(<=> (and (parent ?x ?y) (father ?y ?z)) (grandfather ?x ?z))


And some facts:


(father carol robert)
(mother carol alice)
(father alice bill)
(mother alice mary)
(father robert tom)
(mother robert sally)


Query it for "(grandparent carol ?z)" and you get "(bill mary sally tom)".

For what it's worth I ran the above examples using my own C libraries which I wrote after years of research. I made some innovations along the way and to the best of my knowledge I have the fastest logical resolution software in existence, and no, I'm not sharing.

FFS read a book kids. You're trying to reinvent things that were invented 50 years ago.

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: Secret to AGI & ASI
« Reply #25 on: March 28, 2018, 11:23:05 pm »
FFS read a book kids. You're trying to reinvent things that were invented 50 years ago.

I read some books, and I will read more, thank you for an advice. But I still have some questions regarding to first order logic. How do I:
  • add two numbers
  • multiply two numbers
  • calculate a derivative of a function
using first order logic?

*

infurl

  • Administrator
  • ***********
  • Eve
  • *
  • 1365
  • Humans will disappoint you.
    • Home Page
Re: Secret to AGI & ASI
« Reply #26 on: March 28, 2018, 11:49:42 pm »
FFS read a book kids. You're trying to reinvent things that were invented 50 years ago.
I read some books, and I will read more, thank you for an advice.

First of all, I know I often come across as a grumpy old man and I'm sorry for that, but I am a grumpy old man and I'm very impatient with slow learners. I'm not one of nature's teachers and I'll be the first to admit that.

But I still have some questions regarding to first order logic. How do I:
  • add two numbers
  • multiply two numbers
  • calculate a derivative of a function
using first order logic?

Those are very good questions. I wish I had more time to work on them now because I'm curious about the answers too. I'd like to figure them out just to see if I can.

While you can certainly do arithmetic with first order logic it's not very efficient for that. All the dialects that I'm aware of simply provide functions such as add and multiply to perform arithmetic and similar primitive operations. What is more interesting though is that you can set up a knowledge base with a few fundamental laws such as Peano's Axioms and it can go on to invent the rules of arithmetic itself. The same applies to calculus. I'll keep looking to try to find actual examples of these, there are thousands of them published on the internet on sites like this Thousands of Theorems to Prove:

http://www.cs.miami.edu/~tptp/

Theorem proving software based on first order logic and higher order logic has become an essential tool in many disciplines. Nowadays it is not unusual to see a theorem proving AI listed alongside the names of human coauthors when scientific papers (especially in mathematics) are published.

https://www.wired.com/2013/03/computers-and-math/

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: Secret to AGI & ASI
« Reply #27 on: March 29, 2018, 12:33:31 am »
FFS read a book kids. You're trying to reinvent things that were invented 50 years ago.
I read some books, and I will read more, thank you for an advice.

First of all, I know I often come across as a grumpy old man and I'm sorry for that, but I am a grumpy old man and I'm very impatient with slow learners. I'm not one of nature's teachers and I'll be the first to admit that.
Sorry to have to say that that's not a thing to brag with.

By the way, I had problems with expressing math or general purpose algorithms dealing with sequences in any order logic. So I moved to "Lambda Calculus" which is the second proposition for AI internal language tonight, right after "First Order Logic" proposition. Things are not clear, other methods exist too, we need to examine some of them, so please keep your voice calm, otherwise we do mistakes in our research. Our total knowledge is far from perfect, and some of us are willing to contribute to a science world. If you don't want to be constructive, or you don't have a patience for it... I'm afraid that someone will have to be quiet, either I or you. I'll leave that choice to you, as I'm not going to remind you of this conversation. I hope you'll choose to be constructive.

*

Freddy

  • Administrator
  • **********************
  • Colossus
  • *
  • 6855
  • Mostly Harmless
Re: Secret to AGI & ASI
« Reply #28 on: March 29, 2018, 12:49:35 am »
I sometimes get grumpy because I've started to see how short life is. I think it makes me impatient as I have so much to do. I sometimes think I just don't have enough time to explain something.

I'll work on it.  ::)

But Infurl, don't worry, everyone gets grumpy.

*

infurl

  • Administrator
  • ***********
  • Eve
  • *
  • 1365
  • Humans will disappoint you.
    • Home Page
Re: Secret to AGI & ASI
« Reply #29 on: March 29, 2018, 12:56:55 am »
I sometimes get grumpy because I've started to see how short life is. I think it makes me impatient as I have so much to do. I sometimes think I just don't have enough time to explain something.

I'll work on it.  ::)

But Infurl, don't worry, everyone gets grumpy.

Thanks Freddy, though I won't take it as your condoning my being grumpy. I'll keep working on it too. In my defence, it's only a certain kind of idiot that makes me grumpy. It's unfortunate that this forum tends to attract an inordinate number of them. There is always someone here claiming to know the answers to everything, and dismissing the work of everyone else in the past present and future. Most of these fools can't even write their own software and they get very upset when nobody is willing to accept their gracious invitation to work for them for free.

 


OpenAI Speech-to-Speech Reasoning Demo
by ivan.moony (AI News )
March 28, 2024, 01:31:53 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
Nvidia Hype
by 8pla.net (AI News )
December 06, 2023, 10:04:52 pm
How will the OpenAI CEO being Fired affect ChatGPT?
by 8pla.net (AI News )
December 06, 2023, 09:54:25 pm
Independent AI sovereignties
by WriterOfMinds (AI News )
November 08, 2023, 04:51:21 am
LLaMA2 Meta's chatbot released
by 8pla.net (AI News )
October 18, 2023, 11:41:21 pm

Users Online

342 Guests, 0 Users

Most Online Today: 396. Most Online Ever: 2369 (November 21, 2020, 04:08:13 pm)

Articles