Re: Friendliness successfully IMPLEMENTED!

From: Mark Waser (
Date: Wed Mar 19 2008 - 18:56:29 MDT

> Humans and AIs have radically different cognitive structures. Trying
> to implement Friendliness on both at the same time isn't going to
> work.

OK. I agree with the first sentence and strongly disagree with the second.
What basis do you have for believing the second?

> My initial thought was that you had actually written out a technical
> specification and tested it on working software.

There is no software sophisticated enough to test Friendliness at this time.
I am sorry for confusing you. It was *not* intentional.

> I'm afraid of this list being infected with impossible Friendliness
> architectures.

If it's impossible then you shouldn't be afraid.

> Since none of us have any idea who you are or where you live, this
> would be rather difficult.

Not at all correct. There are a decent number of people on this list who
have met me in person at either the AGIRI Workshop in Maryland in May of
2006 or the AGI08 Conference held earlier this month in Memphis. If you
Google my name, the top entry tells you where I work. If you add my state
(obvious from my workplace) to my name, the top nine entries are mine and
can tell you what my last job was, where I go to church, two of the last
races that I ran in, etc., etc. Are all your statements this bad? The
Bayesian probability of your statement about implementing Friendliness on
both not going to work just took a drop.

> One of Christianity's main principles is that everyone acts
> unethically all the time. This isn't quite true, but it's plausible
> enough, given most people's understanding of human cognitive
> architecture. What makes you different from everyone else?

No, that principle says that everyone is imperfect and makes mistakes. I am
no different from everyone else.

> If you're intelligent enough to rewrite human architecture to that
> extent, without the benefit of ultratechnology, you should be
> intelligent enough to figure out how to provide evidence for it.

I see. You can't figure out a decent proof either. Why not be honest and
admit it rather than launching another attack when all I did was ask for a

>> What cooperative Friendliness? I'm not seeing any cooperation here at
>> all. I asked for assistance. I said "If you believe that this is
>> incorrect, *PLEASE* free me by showing me where this meme is harmful and
>> rid me of my infection." You came back and attacked me for being
>> misleading and then demanded proof that I was a good person.
> It only becomes cooperative when it spreads to a large group of
> people, which (if I understand correctly) is what you're assuming.

I'm not trying to promote cooperative Friendliness. This is a side-track.

>> Prove your statement by freeing me of the meme or making me willingly
>> perform an unethical act.
> I don't have to. If I'm powerful enough, I can simply *shoot you*.
> Poof, there goes Friendliness.

Yup. You can kill my Friendliness by killing me. Is that your solution to
the Friendly AI problem as well? I don't believe that it's going to work.

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:02 MDT