AGI ethics (was: distributive programming assistant)

[YKY (Yan King Yin, 甄景贤)]

On Sat, Jul 16, 2011 at 5:35 PM, <ivanv...@gmail.com> wrote:
> When I think of AI, I see it as in standard movie's intelligent
> robot-concept.
> Something like fully autonomous system with these props:
>
> 1. autonomous world exploration and learning machine
> This can be resolved by integrating sensors to knowledge base. These
> sensors would be represented by a few predicates that would be
> tested by AGI for acquiring new knowledge.
>
> 2. by acquired knowledge base extended with inductive and deductive
> consequences (AGI), system should decide what to do with it's own
> output (at least a console to talk trough). This decision should reflect
> each wish of *every* living being (I need a help here). This is what's
> partialy described in algorithm.
> Decision of what to do would be represented by yet nonexisting state
> of input sensors (to recognize failure or success)
>
> 3. the machine could build a theory for reaching wanted state (like
> theorem proofing or automatic algorithm construction wich is,
> I believe very well studied and could be implemented by existing
> technology. At least, deductive proof that canonnical sequence of
> some output states results with wanted state in Universe.
>
> 4. test the theory in practice (*risky part*)
>
> As the most of the repeatable process is resolved, I believe that
> we are not far away from autonomous system that should be guided
> by living beings. In algorithm in link, I propose a very friendly
> behaviour that fulfills everyone's wishes, and that is not telling NO
> to any living being.
>
> But I'm not a God, I might be making a big mistake, I might be crazy
> and I don't want to decide alone what should do a machine that may be
> as intelligent as hell. In one word, I'm not certain that no being would
> suffer from mine decision.
>
> I suppose that this is very big question which will human race face at
> the moment when the last peace of puzzle -detecting wishes- would
> come to the light. Maby some consort of very wise people could
> make that decision.
>
> I'm writing this text to point to that last peace of puzzle: detecting
> living being's wishes. I don't need you to ultimately help me. I just think
> that this would be interesting on some day which You, people are bringing
> sooner and sooner.

My stance towards AGI is very liberal (in the sense of "laissez
faire"). I don't think using AGI to govern people (or to "play God")
is practical or feasible. Of course AGI will lead to new forms of
human organization, but it would be emergent, not designed from the
top down.

I have a vision of creating an AGI project that transcends national
boundaries and is free of discrimination. But I do realize that this
ideal may be met with resistance. It led me to think about whether
there is a scientific basis of morality.

And, guess what, I couldn't find it.

Everyone has some notions of what is "fair", but when we look closer,
the standards of fairness are always relative and often very
arbitrary. We don't really know what is fair. Nevertheless, we
continue to live by vague notions of fairness.

My ideal way of running an AGI project may be called "meritocracy",
where people would be rewarded for their contributions. Meritocracy
implies that there should be no discrimination based upon notions such
as age, sex, race, nationality, etc. They seem to be "irrelevant" to
the project.

However, I sometimes have doubts whether this is actually the "right"
ideal. As I said before, there is no scientific basis for morality.

The main problem with this approach is that it is "ahistorical". In
particular, I have trouble convincing people of different races and
nationalities to collaborate together and forget their differences.

Usually, in international business, most people manage to discard
their national / racial / cultural differences to collaborate.
However, the nature of AGI tends to drive people to extremes.

Perhaps I should try to emphasize that AGI is no different from other
forms of high-tech. People tend to label Transhumanism and
Singularity as "elitist", which is to some extent true, but it is also
quite exaggerated.

Just think back on how people used to deify technology pioneers like
Thomas Edison or Charles Lindbergh. Looking back, they're just rather
ordinary human beings, with their own flaws....

=)
KY

[ivanv...@gmail.com]

OK

[Ivan Vodišek]

If we get morality formalized, a dreamfull life can be achieved.
Imagine a world in which labour of living beings would be fully
rooted out. No work, no hunger or thirst, no lack of goodies
which divide poor from rich, and at the end, no sane motivation
for a crime. I could live there, I think :D

Please, allow me to jump to the ruff part =|
Morality maybe consequents from need to make everyone happy.
If just one living unit is not happy with some action, that action
should not be done. No compromise. Just actions that are wanted
or neutral.
No one wants anyone to suffer. Some other's suffering may be just
a point in a process for fulfilling some other personal wish. But the
way for reaching personal wish should be consisted only by
acceptable actions. I strongly believe that other's suffering is no
option for anyone sane?
There, I said it, but only cause I want You to correct me if I'm wrong :?

If You all agree on  proposed universal ethic, I'd like to try to
connect super-state-achiever-respect-anyones-wishes-machine on
following wish detector:

Anything that anyone do is result of process of achieving some
wanted state of Universe. We can draw a parallel:
   "process" = program or GP parse tree
and
   "wanted state" = specifications of program or GP theory

The process of achieving wanted state is moreover just a process
of maintaining some more general state that got ruined. So, to detect
a wish, we only have to detect state which a living being is trying to
maintain, and that state is theory of GP about everything.

About everything? Why not just living beings? Well, I don't have
a method for recognizing living beings, but that seems to be
irrelevant.
Suppose that we have a state with a stone about 2m in the air.
Some living being wants that stone on 1m in the air. Universe
wants the stone to obey it's rule of gravity and acceleration and who
knows what else. Rules of the Universe are always obeyed, so it
doesn't meter to us because intersect between living being's wish
and Universe's rules is always living being's wish!

I say: Machine needs to obey EVERY rule that is proofed to exist
in Universe, including living beings wantings and rules of the
unliving matter. In the case of conflict (just between livings,
Universe is always right), the machine does nothing until it finds
some acceptable actions, and for Universe, and for livings.

What if I'm wrong? let our children forgive me, I can't stand this
suffer anymore, please either argue me, either let me try.

i.

[Matt Mahoney]

Perhaps you know that http://singinst.org/ is also interested in AGI ethics. But I think we are a long way from having to worry about it because we are a long way from AGI.

 

-- Matt Mahoney, matma...@yahoo.com

---

From: Ivan Vodišek <ivanv...@gmail.com>
To: general-in...@googlegroups.com
Sent: Thursday, July 28, 2011 2:33 PM
Subject: Re: [GI] AGI ethics (was: distributive programming assistant)

[YKY (Yan King Yin, 甄景贤)]

On Fri, Jul 29, 2011 at 8:09 AM, Matt Mahoney <matma...@yahoo.com> wrote:

Perhaps you know that http://singinst.org/ is also interested in AGI ethics. But I think we are a long way from having to worry about it because we are a long way from AGI.

You have continually attacked my project while approving of other similar projects.  I have asked you for an explanation and you couldn't give any.  I'm afraid you're going crazy.  If you still can't give a reasonable explanation, I'll ban you from this list for 1 month.

Perhaps you're suffering from an exaggeration of the consequences of AGI promulgated by Singularitarians and Luddites alike, and also fueled by science fiction.  (I like sci-fi too, but I don't think they are to be taken literally.)

Another problem is that life extension technology implies the relatively easy modification of physical appearance, which would imply the end of white supremacy.  Perhaps your white guilt is causing you to panic.

I think you need to calm down and sort these things out in your head, lest you become like the Unabomber.

Sometimes you are contributing useful ideas, but that does not give you the right to make false accusations against me.

I have already gone from "wanting to be like the young Bill Gates" to "becoming like the old Bill Gates without having been like the young Bill Gates".  Isn't that enough?  LMAO

KY

[J. Andrew Rogers]

2011/7/28 YKY (Yan King Yin, 甄景贤) <generic.in...@gmail.com>:

> Another problem is that life extension technology implies the relatively
> easy modification of physical appearance, which would imply the end of white
> supremacy.  Perhaps your white guilt is causing you to panic.

You just painted yourself as an emotionally insecure racist for the
whole Internet to see. Was that your intent?

> I think you need to calm down and sort these things out in your head, lest
> you become like the Unabomber.

You realize, of course, that the above reflects more on you than on
Matt given your prior assertions. Matt may be fervent and perhaps even
a bit off-base at times but he is civil.

Credibility matters. Irrational, emotional outbursts like the above
are a fantastic way to ensure that you are ignored by people that do
not want to be associated with such uncultured behavior.

--
J. Andrew Rogers

[YKY (Yan King Yin, 甄景贤)]

On Fri, Jul 29, 2011 at 12:43 PM, J. Andrew Rogers <jar.m...@gmail.com> wrote:

 

> Another problem is that life extension technology implies the relatively
> easy modification of physical appearance, which would imply the end of white
> supremacy.  Perhaps your white guilt is causing you to panic.

You just painted yourself as an emotionally insecure racist for the
whole Internet to see. Was that your intent?

He kept attacking me, so I inferred that:

1. he is either racist;

2. or the grape has < 7 pH value

but I can't scan his brain so I don't know which is the case.  Either way, he failed to explain why I was singled out as unethical and my theory was singled out as unsound while other people and projects similar to mine are exceptional.  Can YOU explain that?

So I was just helping him to sort out his racism OR sourness (but I cannot determine which is the case).

You realize, of course, that the above reflects more on you than on

Matt given your prior assertions. Matt may be fervent and perhaps even
a bit off-base at times but he is civil.

He is attacking me under the guise of civility.  It wasted my time and blocked progress in my project -- I consider that a very serious offense -- many people want to have AGI and life extension sooner, and he has no right to impede our progress.

Credibility matters. Irrational, emotional outbursts like the above
are a fantastic way to ensure that you are ignored by people that do
not want to be associated with such uncultured behavior.

If you cannot explain his asymmetrical behavior, you will also be banned for 1 month.  Note:  I am giving both of you time to defend yourselves.

KY

[YKY (Yan King Yin, 甄景贤)]

PS:  Sorry, Andrew, I'm not going to ban you, because you're not responsible for Matt's behavior, nor were you aware of what happened in the past few YEARS.

KY

[Ivan Vodišek]

>>> Another problem is that life extension technology implies the relatively
>>> easy modification of physical appearance, which would imply the end of white
>>> supremacy.  Perhaps your white guilt is causing you to panic.
>>
>>

>>You just painted yourself as an emotionally insecure racist for the
>>whole Internet to see. Was that your intent?

>

>He kept attacking me, so I inferred that:

>1. he is either racist;

>2. or the grape has < 7 pH value

>but I can't scan his brain so I don't know which is the case.  Either way, he failed to explain why I was singled out as unethical and
>my theory was singled out as unsound while other people and projects similar to mine are exceptional.  Can YOU explain that?

Perhaps white people are guilty for having such image about them. I can't blame
anyone. But please, note that interpeople differences are greater then interrace
differences. Allthought my skin is piggy-white, I don't agree with all decisions my
race has done through history and even now. Yes, my skin is white, but that's not
important cause I'm *person* above all.

i.

[Matt Mahoney]

YKY, I was referring to my own approach to AGI being decades away from a solution. You have seen my cost estimate. But go ahead and ban me. I am leaving. Anyone who wants to continue the discussion can find me on the AGI and Singularity lists at http://www.agiri.org/email/ if you aren't already on them.

 

-- Matt Mahoney, matma...@yahoo.com

---

From: "YKY (Yan King Yin, 甄景贤)" <generic.in...@gmail.com>
To: general-in...@googlegroups.com
Sent: Thursday, July 28, 2011 11:20 PM

Subject: Re: [GI] AGI ethics (was: distributive programming assistant)

[Ivan Vodišek]

Matt, i'm sorry i've judged you in my post. i'm certain than You are not mean person.
If anyone feels that way, it must be some mistake cause there are no such a thing as
a mean person. Please take my apologies for being subjective.

ivan vodišek

2011/7/29 Matt Mahoney <matma...@yahoo.com>

[YKY (Yan King Yin, 甄景贤)]

I'm sorry that this happened...  but maybe it's better that Matt has his own project so we can pursue our goals peacefully without constant arguing...

Also I apologize for saying some inappropriate things when I was upset...

KY

[YKY (Yan King Yin, 甄景贤)]

I'm not sure if I've done the right thing...

Matt has some "reasons" that make him think Genifer won't work:  eg, that logic-based AI is old fashioned AI;  people have tried it and it failed.  etc etc

I have explained many times to him:  Genifer is an improvement over GOFAI, it's different.  But he doesn't get it.

Next month, same thing happens.  Ad infinitum.

Personally I think Matt's reasons are "fucking stupid".  But I also want to run the project "democratically".

Every time this project is gaining some momentum, he interrupts and ruins it.  It's driving me crazy.

What should I do?

KY

[William Taysom]

> What should I do? ... explained many times ... next month, same thing happens

Don't feed the trolls. If someone brings up something that you don't want to in talk about, then ignore it.

[YKY (Yan King Yin, 甄景贤)]

On Sat, Jul 30, 2011 at 5:29 PM, William Taysom <wta...@gmail.com> wrote:

> What should I do? ... explained many times ... next month, same thing happens

Don't feed the trolls.  If someone brings up something that you don't want to in talk about, then ignore it.

Well, Matt may or may not be a troll -- we can't scan his brain so we don't know =)

We have 2 options:

1.  Do it the small way, ie, everyone has to agree with the theory of "YKY & friends", or go to another project.  So it would not be democratic.

2.  Merge with Opencog and set up some democratic rules.  There will be a lot of management overhead.  The advantage is that we'll have large-scale collaboration.

It's hard to choose...  I'm thinking...

KY

[Ivan Vodišek]

2011/7/30 YKY (Yan King Yin, 甄景贤) <generic.in...@gmail.com>

To make a decision a classic approach can be taken:

- list differencies between Genifer and Opencog
- list advanteges and disadvanteges of each choice

My personal opinion is: if diferences exist it's better to breath them life,
just to make an experiment that in the end may happend to be the
right one for community.

I'm for a new thing, if anyone asks. Maby i did'nt mentioned, i'd be happy
to program any part of Genifer in any language that's required from me. I'll
give up with my super-duper language cause Genifer may be more important.
I have 24/7 time on my own. Where should I start?

[YKY (Yan King Yin, 甄景贤)]

On Sat, Jul 30, 2011 at 6:52 PM, Ivan Vodišek <ivanv...@gmail.com> wrote:

To make a decision a classic approach can be taken:

- list differencies between Genifer and Opencog
- list advanteges and disadvanteges of each choice

We're both based on some kind of probabilistic logic, and Ben has shown some interest in adopting my fuzzy-probabilistic logic idea.  He's currently on vacation in Mongolia, but I'll meet him in December in Hong Kong =)

The main reason I keep wanting to collaborate with Opencog is that I fear AGI would be a big project that requires a community effort.  It is easy to write a prototype Genifer, but when the prototype grows it may become resource-intensive.

My personal opinion is: if diferences exist it's better to breath them life,
just to make an experiment that in the end may happend to be the
right one for community.

I'm for a new thing, if anyone asks. Maby i did'nt mentioned, i'd be happy
to program any part of Genifer in any language that's required from me. I'll
give up with my super-duper language cause Genifer may be more important.
I have 24/7 time on my own. Where should I start?

I'm really glad you have the time...  maybe you want to look at the inductive learning problem?  It's a very interesting and rich topic and I'll provide you with all the info I have...

I'd try to finish the inference engine, ie, the Bayesian stuff <-- programming that is dreadfully tedious IMO.

KY

[YKY (Yan King Yin, 甄景贤)]

On Sat, Jul 30, 2011 at 2:26 AM, Matt Mahoney <matma...@yahoo.com> wrote:

YKY, I was referring to my own approach to AGI being decades away from a solution. You have seen my cost estimate. But go ahead and ban me. I am leaving. Anyone who wants to continue the discussion can find me on the AGI and Singularity lists at http://www.agiri.org/email/ if you aren't already on them.

I'm still dissatisfied with your answer.  How did you know how many decades Genifer needs to build a common sense KB?  And why are you so sure?  And why does that not apply to Opencog or NARS?

Secondly, you said that Genifer will never succeed and then the very next post you supported Opencog to use CMR.

KY

[YKY (Yan King Yin, 甄景贤)]

2011/7/30 YKY (Yan King Yin, 甄景贤) <generic.in...@gmail.com>

Personally I think Matt's reasons are "fucking stupid".  But I also want to run the project "democratically".

OK, I also apologize for the use of foul language.  I want to eliminate all reasons why Matt is against me / our project.

KY

[Mike Dougherty]

2011/7/30 YKY (Yan King Yin, 甄景贤) <generic.in...@gmail.com>:

Let it go. You have started replying to your own messages. :)

I didn't think Matt was out to cause you such grief. He has pet
theories (involving a lot of x times ten to the yth power, etc.) You
have pet theories. (I'm not as sure how to humorously paraphrase
though) You're both working the same field. Sometimes theories
overlap enough that there is agreement. Sometimes they're in
opposition. Since nobody has produced the definitive working AI I
think everyone should keep an open mind.

The time you have spent worrying about what Matt thinks is time you
have missed thinking on your own project. Refocus. Get back to doing
what makes you happy. Life's too short to do anything else.

[Sandeep Pai]

Best reply ever!

2011/7/31 Mike Dougherty <msd...@gmail.com>

[Ivan Vodišek]

Does anyone have idea about:
How can one detect intension of a being who undertakes sequence of
actions (i.e. writes something in console input)? Assume that we have
parse rules to construct a predicate tree from input.

i.

[Mike Dougherty]

2011/8/3 Ivan Vodišek <ivanv...@gmail.com>:

As in how much evil does one have to be doing before kicking them out
of your system and fixing their damage?

Most of what you could do for an IDS would be best discussed beyond
the opening question in private email. I'd love to brainstorm, so
include me if you get any other takers. :)

[Matt Mahoney]

>From: Mike Dougherty <msd...@gmail.com>

I did my dissertation on network intrusion detection after switching from data compression because that's what my adviser had funding for. It turns out that malicious traffic does not compress as much as normal traffic. Unfortunately this method gives too many false alarms to be useful for real time detection, but it might be good for forensics after an attack.
 

-- Matt Mahoney, matma...@yahoo.com

[Ivan Vodišek]

Hi all!

I'm introducing "Sky" personality for autonomous AI. Can be
applied for avoiding morality conflicts in AGI if equipped with
social data.

https://sites.google.com/site/ivanvodisek/

♬

[YKY (Yan King Yin, 甄景贤)]

I think conflicts cannot always be solved by constraint satisfaction alone, though when it can, it'd be very nice.  But sometimes someone's wish cannot be fulfilled, such as in many competitive situations.  For example when Pete Sampras and Andre Agassi competed in tennis.  Or Garry Kasparov vs Anatoly Karpov in Chess.

In any competition the loser will feel unhappy.  In sports, we're taught to respect the rules, ie, sportsmanship.  In real life situations, the rules may not be entirely clear and therefore we often see a lot of people in the business world doing unethical things to get ahead.

KY

[Ivan Vodišek]

2011/8/22 YKY (Yan King Yin, 甄景贤) <generic.in...@gmail.com>

in sport there is super-goal to winning: playing the game the best you
can to pick the winner. this one seems solved, but maybe i'm still
missing something :|

[YKY (Yan King Yin, 甄景贤)]

On Mon, Aug 22, 2011 at 8:22 AM, Ivan Vodišek <ivanv...@gmail.com> wrote:

in sport there is super-goal to winning: playing the game the best you
can to pick the winner. this one seems solved, but maybe i'm still
missing something :|

In real life situations, the "rules" are not set clearly, and it may seem that there are no rules at all (in a sense that's true).  So, if 2 parties both think they can win, the competition can devolve into violence, killing, and wars.

We're still seeing this kind of problems...

KY