The religion of AI

160 views
Skip to first unread message

Cosmin Visan

unread,
May 7, 2019, 2:15:11 PM5/7/19
to Everything List
"However, I think the notion of robots with IQs of 10,000 rather gives the game away. A human being can solve an IQ puzzle that he or she has NEVER seen before. A computer has to be told how to do it by its programmers – either in the form of a ready-made algorithm or by feeding it large amounts of LABELLED data. If presented with a different puzzle, the computer is completely clueless again." - This seems so obvious to me that I wonder why AI religious believers don't see it.

https://unherd.com/2018/05/influential-religion-youve-never-heard/?=refinnar

Jason Resch

unread,
May 7, 2019, 2:40:46 PM5/7/19
to Everything List

One AI teaches itself how to play Go, Chess, and Shogi better than any other human.
https://spectrum.ieee.org/tech-talk/robotics/artificial-intelligence/mb

One AI teaches itself how to master 49 different video games:

Jason

On Tue, May 7, 2019 at 1:15 PM 'Cosmin Visan' via Everything List <everyth...@googlegroups.com> wrote:
"However, I think the notion of robots with IQs of 10,000 rather gives the game away. A human being can solve an IQ puzzle that he or she has NEVER seen before. A computer has to be told how to do it by its programmers – either in the form of a ready-made algorithm or by feeding it large amounts of LABELLED data. If presented with a different puzzle, the computer is completely clueless again." - This seems so obvious to me that I wonder why AI religious believers don't see it.

https://unherd.com/2018/05/influential-religion-youve-never-heard/?=refinnar

--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To post to this group, send email to everyth...@googlegroups.com.
Visit this group at https://groups.google.com/group/everything-list.
To view this discussion on the web visit https://groups.google.com/d/msgid/everything-list/3ce0c2b9-0189-44c8-8a3f-e0e76648b219%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Brent Meeker

unread,
May 7, 2019, 3:28:23 PM5/7/19
to everyth...@googlegroups.com


On 5/7/2019 11:15 AM, 'Cosmin Visan' via Everything List wrote:
"However, I think the notion of robots with IQs of 10,000 rather gives the game away. A human being can solve an IQ puzzle that he or she has NEVER seen before. A computer has to be told how to do it by its programmers – either in the form of a ready-made algorithm or by feeding it large amounts of LABELLED data.

You are at least two years out of date.  AlphaGo was not given labelled data, it was only given the rules of the game. It had never "seen" a game of Go.  Adversarial training and unsupervised learning don't need labelling.  Leslie Valiant has made a critical observation in this area. He showed that ensembles of artificial neurons can act as a symbol, and that logical operations can be performed on such ensembles, even though the distribution of the nodes and wiring is initially completely random.

Brent

If presented with a different puzzle, the computer is completely clueless again." - This seems so obvious to me that I wonder why AI religious believers don't see it.

https://unherd.com/2018/05/influential-religion-youve-never-heard/?=refinnar

Brent Meeker

unread,
May 7, 2019, 3:34:01 PM5/7/19
to everyth...@googlegroups.com
But can AI write nonsense theories of everything as well as human arm-chair philosophers?  When they write a book explaining why humans can't really be conscious because they can't even balance their bank account, THEN we'll know they've master human reasoning, including artificial stupidity.

Brent

John Clark

unread,
May 7, 2019, 3:34:58 PM5/7/19
to everyth...@googlegroups.com
On Tue, May 7, 2019 at 2:15 PM 'Cosmin Visan' via Everything List <everyth...@googlegroups.com> 
wrote:

. > A human being can solve an IQ puzzle that he or she has NEVER seen before. A computer has to be told how to do it by its programmers

Modern programs can get good enough to beat any human at Chess or GO in just a few hours (by playing games against itself) starting from knowing NOTHING about the games except for the basic rules that tell it what moves are legal. You couldn't do that, no human could, but a computer can.   

 John K Clark




Cosmin Visan

unread,
May 7, 2019, 4:53:51 PM5/7/19
to Everything List
By saying such things you just prove that you have no understanding of computer science whatsoever. You are completely oblivious that those AIs are first programmed by someone in certain specific ways. You probably think of them as some sort of magical spirits that appear out of nowhere by themselves.

Cosmin Visan

unread,
May 7, 2019, 4:54:43 PM5/7/19
to Everything List
And how are "rules of the game" not labelled data ?

John Clark

unread,
May 7, 2019, 5:05:50 PM5/7/19
to everyth...@googlegroups.com
On Tue, May 7, 2019 at 4:53 PM 'Cosmin Visan'  <everyth...@googlegroups.com> wrote:

> By saying such things you just prove that you have no understanding of computer science whatsoever

We do know 2 facts about computer science that you do not, computers exist and so do computations.

John K Clark

 

Brent Meeker

unread,
May 7, 2019, 5:06:39 PM5/7/19
to everyth...@googlegroups.com


On 5/7/2019 1:53 PM, 'Cosmin Visan' via Everything List wrote:
By saying such things you just prove that you have no understanding of computer science whatsoever. You are completely oblivious that those AIs are first programmed by someone in certain specific ways.

They are programmed in specific ways for learning...just like you were programmed, by evolution, to learn a language.

Brent

You probably think of them as some sort of magical spirits that appear out of nowhere by themselves.

On Tuesday, 7 May 2019 21:40:46 UTC+3, Jason wrote:

One AI teaches itself how to play Go, Chess, and Shogi better than any other human.
https://spectrum.ieee.org/tech-talk/robotics/artificial-intelligence/mb

One AI teaches itself how to master 49 different video games:

Jason

--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To post to this group, send email to everyth...@googlegroups.com.
Visit this group at https://groups.google.com/group/everything-list.

Brent Meeker

unread,
May 7, 2019, 5:10:57 PM5/7/19
to everyth...@googlegroups.com
They are not labelled 'true' or 'false' or 'cat' or 'dog'.   But whatever you want to call them, they are very small; contrary to your ignorant assertion that "A computer has to be told how to do it by its programmers – either in the form of a ready-made algorithm or by feeding it large amounts of LABELLED data."

Brent
--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To post to this group, send email to everyth...@googlegroups.com.
Visit this group at https://groups.google.com/group/everything-list.

Jason Resch

unread,
May 7, 2019, 5:10:58 PM5/7/19
to everyth...@googlegroups.com
Thanks for reinforcing my belief that I shouldn't interact with you.

Jason
--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-list+unsubscribe@googlegroups.com.
To post to this group, send email to everything-list@googlegroups.com.

John Clark

unread,
May 7, 2019, 5:12:48 PM5/7/19
to everyth...@googlegroups.com
On Tue, May 7, 2019 at 4:54 PM 'Cosmin Visan'  <everyth...@googlegroups.com> wrote:

>And how are "rules of the game" not labelled data ?

That's just stating the problem! Nothing, not a person and not a machine, can give an answer if they don't know the question.  Do you actually think the above is a good rebuttal??  

John K Clark




Philip Thrift

unread,
May 7, 2019, 5:28:13 PM5/7/19
to Everything List

The bottom line of the DeepMind test in 2018:

Ultimately, the team’s AI IQ test shows that even some of today’s most advanced AIs can’t figure out problems [in domains] we haven’t trained them to solve.


smitra

unread,
May 7, 2019, 7:06:04 PM5/7/19
to everyth...@googlegroups.com
There exists a limited set of problems the AI can tackle successfully
due to the way the AI was trained. But the same is true for human beings
tackling problems. While we can tackle a far wider range of problems,
this is ultimately also limited due to the way the brain came into being
and the way it ended up being trained to tackle problems. Also, one can
consider living organisms as algorithms in the sense of universal
constructors and they have a limited range of survivability also due to
the effective training that forged their biochemical machinery.

Saibal
> --
> You received this message because you are subscribed to the Google
> Groups "Everything List" group.
> To unsubscribe from this group and stop receiving emails from it,
> send an email to everything-li...@googlegroups.com.
> To post to this group, send email to
> everyth...@googlegroups.com.
> Visit this group at https://groups.google.com/group/everything-list
> [1].
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/everything-list/3ce0c2b9-0189-44c8-8a3f-e0e76648b219%40googlegroups.com
> [2].
> For more options, visit https://groups.google.com/d/optout [3].
>
>
> Links:
> ------
> [1] https://groups.google.com/group/everything-list
> [2]
> https://groups.google.com/d/msgid/everything-list/3ce0c2b9-0189-44c8-8a3f-e0e76648b219%40googlegroups.com?utm_medium=email&utm_source=footer
> [3] https://groups.google.com/d/optout

Cosmin Visan

unread,
May 8, 2019, 3:55:25 PM5/8/19
to Everything List
Since evolution is the evolution of consciousness, I wasn't programmed, but consciousness by itself solved new and never before seen problems without anyone telling her what to do.

Brent Meeker

unread,
May 8, 2019, 4:06:34 PM5/8/19
to everyth...@googlegroups.com


On 5/8/2019 12:55 PM, 'Cosmin Visan' via Everything List wrote:
Since evolution is the evolution of consciousness, I wasn't programmed, but consciousness by itself solved new and never before seen problems without anyone telling her what to do.

By random variation with selection.

Brent


On Wednesday, 8 May 2019 00:06:39 UTC+3, Brent wrote:

They are programmed in specific ways for learning...just like you were programmed, by evolution, to learn a language.

Brent
--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To post to this group, send email to everyth...@googlegroups.com.
Visit this group at https://groups.google.com/group/everything-list.

Cosmin Visan

unread,
May 8, 2019, 4:26:13 PM5/8/19
to Everything List
No. By rationality.

Brent Meeker

unread,
May 8, 2019, 5:01:42 PM5/8/19
to 'Cosmin Visan' via Everything List
Rationality is a property of decision processes. 

Brent
--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To post to this group, send email to everyth...@googlegroups.com.
Visit this group at https://groups.google.com/group/everything-list.

Cosmin Visan

unread,
May 8, 2019, 5:06:26 PM5/8/19
to Everything List
Of course. Decisions which are taken using free will.

Brent Meeker

unread,
May 8, 2019, 5:17:06 PM5/8/19
to everyth...@googlegroups.com
So you can will yourself to experience new qualia?

Brent
--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To post to this group, send email to everyth...@googlegroups.com.
Visit this group at https://groups.google.com/group/everything-list.

John Clark

unread,
May 9, 2019, 8:12:48 AM5/9/19
to everyth...@googlegroups.com
On Wed, May 8, 2019 'Cosmin Visan' v <everyth...@googlegroups.com> wrote:

> No. By rationality.

If something is rational then there is a reason, a cause, for it doing what it did. If it was irrational then there was no reason, no cause, for it doing  what it did. There has long been a word for events without causes, "random".  

> Decisions which are taken using free will.

I have no idea what you mean by "free will" and, because you obviously haven't given the matter any serious thought, you don't know what you mean by "free will" either. However I do know there are only 2 posabilities, events happen because of cause and effect or they don't.

  John K Clark


Cosmin Visan

unread,
May 9, 2019, 4:06:12 PM5/9/19
to Everything List
Only because you don't have absolute will it doesn't mean you don't have limited will. I see a lot of strawman fallacy around here.

Brent Meeker

unread,
May 9, 2019, 4:19:31 PM5/9/19
to everyth...@googlegroups.com
OK, what limits it?  physics?

Brent
--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To post to this group, send email to everyth...@googlegroups.com.
Visit this group at https://groups.google.com/group/everything-list.

John Clark

unread,
May 9, 2019, 6:26:26 PM5/9/19
to everyth...@googlegroups.com
On Thu, May 9, 2019 at 4:06 PM 'Cosmin Visan'  <everyth...@googlegroups.com> wrote:

>>I have no idea what you mean by "free will" and, because you obviously haven't given the matter any serious thought, you don't know what you mean by "free will" either. However I do know there are only 2 posabilities, events happen because of cause and effect or they don't.
> Only because you don't have absolute will it doesn't mean you don't have limited will. I see a lot of strawman fallacy around here.

Did you choose to respond to my post for a reason? If you did then your very response demonstrates the falsehood of the idea you're trying to convey. On the other hand if you responded for no reason then you are being irrational and irrational arguments are obviously not to be taken seriously. And if there is anything in the world we can be absolutely positively 100% certain of its that you either did or you didn't.  

I have said many times before there is no law of logic that demands every event have a cause and I stand by that, however every argument that claims to be logical must have a cause. 

 John K Clark

 

Cosmin Visan

unread,
May 10, 2019, 2:40:56 AM5/10/19
to Everything List
Evolution.

Bruno Marchal

unread,
May 10, 2019, 7:15:36 AM5/10/19
to everyth...@googlegroups.com
On 7 May 2019, at 21:33, 'Brent Meeker' via Everything List <everyth...@googlegroups.com> wrote:

But can AI write nonsense theories of everything as well as human arm-chair philosophers?  When they write a book explaining why humans can't really be conscious because they can't even balance their bank account, THEN we'll know they've master human reasoning, including artificial stupidity.

The universal machine(s), unprogrammed, are born maximally intelligent and conscious. 

It will take some time before they become as stupid as the human, and feel superior, or inferior (the main symptom of stupidity).

But they can wake up, from times to times, with or without “brain perturbation technic”.

Bruno 



Bruno Marchal

unread,
May 10, 2019, 7:19:18 AM5/10/19
to everyth...@googlegroups.com

On 7 May 2019, at 23:10, 'Brent Meeker' via Everything List <everyth...@googlegroups.com> wrote:

They are not labelled 'true' or 'false' or 'cat' or 'dog'.   But whatever you want to call them, they are very small; contrary to your ignorant assertion that "A computer has to be told how to do it by its programmers – either in the form of a ready-made algorithm or by feeding it large amounts of LABELLED data.”


We can program a computer with universal goal, like “help yourself”.

We cannot tell the computer “that the Heaven will help him”. That is true, but become false when asserted.

Bruno




Brent

On 5/7/2019 1:54 PM, 'Cosmin Visan' via Everything List wrote:
And how are "rules of the game" not labelled data ?

On Tuesday, 7 May 2019 22:28:23 UTC+3, Brent wrote:

You are at least two years out of date.  AlphaGo was not given labelled data, it was only given the rules of the game. I
--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To post to this group, send email to everyth...@googlegroups.com.
Visit this group at https://groups.google.com/group/everything-list.
To view this discussion on the web visit https://groups.google.com/d/msgid/everything-list/b0101d94-ccbc-4c49-b4bb-fef7d323048c%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.


--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To post to this group, send email to everyth...@googlegroups.com.
Visit this group at https://groups.google.com/group/everything-list.

Bruno Marchal

unread,
May 10, 2019, 7:26:33 AM5/10/19
to everyth...@googlegroups.com
On 8 May 2019, at 21:55, 'Cosmin Visan' via Everything List <everyth...@googlegroups.com> wrote:

Since evolution is the evolution of consciousness, I wasn't programmed, but consciousness by itself solved new and never before seen problems without anyone telling her what to do.

But machines can do that too. There is an entire field of theoretical computational Learning theory. Most result result are negative, in the sense that we cannot even compute the gain in learning abilities when we weaken the identification criteria (to solve the inference inductive aspect of learning).

A good book, which requires a bit familiarity in Recursion theory, is 

OSHERSON D.N., STOB M.and WEINSTEIN S., 1986, Systems that Learn, MIT press.

(There has been a new edition, with some more authors).

My favorite papers in this field:

CASE J., 1989, The Power of Vacillation, COLT '88, Proceedings of the 1988 Workshop on Computational Learning Theory, pp. 196-205.

CASE J. & NGO-MANGUELLE S., 1979, Refinements of inductive inference by Popperian machines. Tech. Rep., Dept. of Computer Science, State Univ. of New-York, Buffalo.

CASE J. & SMITH C., 1983, Comparison of Identification Criteria for Machine Inductive Inference. In Theoretical Computer Science 25,.pp 193-220.

CASE J., CHEN K., JAIN S., 1992, Strong Separation of Learning Classes, in Jantke (Ed.) 1992.

Bruno




On Wednesday, 8 May 2019 00:06:39 UTC+3, Brent wrote:

They are programmed in specific ways for learning...just like you were programmed, by evolution, to learn a language.

Brent

--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To post to this group, send email to everyth...@googlegroups.com.
Visit this group at https://groups.google.com/group/everything-list.

Brent Meeker

unread,
May 10, 2019, 2:03:45 PM5/10/19
to everyth...@googlegroups.com
So the mind evolves by rational acts of free will (maybe JKC will explain what that means) but that free will is limited by "evolution".

Brent
--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To post to this group, send email to everyth...@googlegroups.com.
Visit this group at https://groups.google.com/group/everything-list.

Cosmin Visan

unread,
May 11, 2019, 6:14:14 AM5/11/19
to Everything List
There are lots of ways through which mind evolves.

John Clark

unread,
May 11, 2019, 6:45:27 AM5/11/19
to everyth...@googlegroups.com
On Sat, May 11, 2019 at 6:14 AM 'Cosmin Visan'  <everyth...@googlegroups.com> wrote:

> There are lots of ways through which mind evolves.

There is only one fundamental reason anything evolves, because of random mutation and natural selection; one factor has a cause and one doesn't. 

John K Clark




 

On Friday, 10 May 2019 21:03:45 UTC+3, Brent wrote:
So the mind evolves by rational acts of free will (maybe JKC will explain what that means) but that free will is limited by "evolution".

Brent

--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To post to this group, send email to everyth...@googlegroups.com.
Visit this group at https://groups.google.com/group/everything-list.

Bruno Marchal

unread,
May 13, 2019, 8:59:39 AM5/13/19
to everyth...@googlegroups.com
On 9 May 2019, at 22:06, 'Cosmin Visan' via Everything List <everyth...@googlegroups.com> wrote:

Only because you don't have absolute will it doesn't mean you don't have limited will. I see a lot of strawman fallacy around here.

I agree. Some people have no scientific attitude when some notion are tackle, like God, or Free-will. In science, when we do find that a notion makes no sense, we propose new definition, and we proceed. But some people want to mock a notion, and usually we reject the new definition by saying that we have not the right to change the definition, but then even Earth does not exist, as it was by definition a flat surface at the beginning. In scie”ice, we change the definition all the time, and to do this properly we can use the axiomatic method, which minimise such changes.

Bruno 




On Thursday, 9 May 2019 00:17:06 UTC+3, Brent wrote:
So you can will yourself to experience new qualia?

--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To post to this group, send email to everyth...@googlegroups.com.
Visit this group at https://groups.google.com/group/everything-list.
Reply all
Reply to author
Forward
0 new messages