<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD>
<META content="text/html; charset=us-ascii" http-equiv=Content-Type>
<META name=GENERATOR content="MSHTML 8.00.6001.18939"></HEAD>
<BODY>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial>
<DIV class=gmail_quote>On Mon, Sep 20, 2010 at 3:19 PM, spike <SPAN
dir=ltr><<A href="mailto:spike66@att.net">spike66@att.net</A>></SPAN>
wrote:</DIV>
<DIV class=gmail_quote><BR> </DIV>
<BLOCKQUOTE
style="BORDER-LEFT: rgb(204,204,204) 1px solid; MARGIN: 0pt 0pt 0pt 0.8ex; PADDING-LEFT: 1ex"
class=gmail_quote>Can you imagine *any* circumstances whereby an emergent AI
would decide to<BR>stay under cover, at least for a while? I can.
Anyone else?<BR></BLOCKQUOTE><BR>1) AI:us :: us:cockroaches...
Jebediah Moore</FONT></SPAN></DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial>Jeb there is a ton of stuff here, more than I have time to
answer, but I like your thinking.</FONT></SPAN></DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial>As Damien showed with his AI roadmap a few days ago, written
ten years ago, there is plenty of thoughtspace here.</FONT></SPAN></DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial>For now, let me just talk about number 1) AI : human ::
human : cockroaches.</FONT></SPAN></DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial>That is one way to look at it, but I would counterpropose this
analogy:</FONT></SPAN></DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial>AI : human :: human : bird.</FONT></SPAN></DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial>The reason I want to go down that road is that it is easy for
me to imagine that even a supercapable AI will observe that for some reason,
there are a few things that humans can do better than AI, and it doesn't really
understand how we do it.</FONT></SPAN></DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial>At my last trip to the ranch in Oregon, I was out by the
river. An eagle swooped down, went into a flat glide over the surface,
then snatched a fish right out of the water. It was a most astonishing
sight. I had heard eagles can do that, but seeing it was a knockout.
Isn't it interesting that humans cannot do that? Sure we can catch fish:
we can hook the hapless beast, we can net it, we can shoot it and scoop the
pieces out afterwards, but we cannot build a machine that can fly down, skim
over and snatch the scaley bastard out of the water. Our feedback control
theory isn't sufficiently advanced to do that, and it isn't clear to me that we
could do it even with sufficient resources given current theory. We don't
know how an eagle can do it, given their little bit of brain.
Impressive.</FONT></SPAN></DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial>Point: human intellectual abilities are supersets of almost
all things that almost all beast do, but there are a few things a few beasts can
do that are outside our ability. You can likely think of your own
examples.</FONT></SPAN></DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial>Consider an AI realizing that there are a few things that
humans just do better than they do, such as writing jokes. Compare the
following two jokes for instance:</FONT></SPAN></DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial>1. Two transistors go into the bar, one says "I will
have a volt of positive charge, and my girl here will have a half a volt of
negative." The bartender gives them a puzzled look and says "But...I see
she is a milli-Farad, so why don't you just give her 500 micro-Coulomb of
yours?"</FONT></SPAN></DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial>2. Two tourists at a bus stop in Germany, BMW driver
comes up asking directions but all he gets are blank stares. "Sprechen sie
deutsche?" Blank stares. "Parlez-vous francais?"
Nothing. "Parlo Italiano?? Habla Espaniol???" he demands.
Nothing. Roars off in a huff. One tourist turns to the other and
says "Billy Joe, me and you aughta learn to talk one of them furrin
launguages." The other, "Ah don't cotton to it Bobby Jack. That
there feller knowed four of them, didn't do him no good."</FONT></SPAN></DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial>Imaine the AI can compose jokes like the first one, but not
like the second. It is puzzled at how we do it and why humans like the
second but not the first, whereas AIs likes the first but doesn't
understand what is funny about the second. It wants
to understand that paradox. So it watches
us.</FONT></SPAN></DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial>spike</FONT></SPAN></DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV>
<DIV dir=ltr align=left><SPAN class=687165920-20092010><FONT color=#0000ff
size=2 face=Arial></FONT></SPAN> </DIV><BR>
<BLOCKQUOTE
style="BORDER-LEFT: #0000ff 2px solid; PADDING-LEFT: 5px; MARGIN-LEFT: 5px; MARGIN-RIGHT: 0px">
<DIV dir=ltr lang=en-us class=OutlookMessageHeader align=left>
<HR tabIndex=-1>
<FONT size=2 face=Tahoma><B>From:</B> extropy-chat-bounces@lists.extropy.org
[mailto:extropy-chat-bounces@lists.extropy.org] <B>On Behalf Of </B>Jebadiah
Moore<BR><B>Sent:</B> Monday, September 20, 2010 1:00 PM<BR><B>To:</B> ExI
chat list<BR><B>Subject:</B> Re: [ExI] stealth
singularity<BR></FONT><BR></DIV>
<DIV></DIV>
<DIV class=gmail_quote>On Mon, Sep 20, 2010 at 3:19 PM, spike <SPAN
dir=ltr><<A href="mailto:spike66@att.net">spike66@att.net</A>></SPAN>
wrote:<BR>
<BLOCKQUOTE
style="BORDER-LEFT: rgb(204,204,204) 1px solid; MARGIN: 0pt 0pt 0pt 0.8ex; PADDING-LEFT: 1ex"
class=gmail_quote>Can you imagine *any* circumstances whereby an emergent AI
would decide to<BR>stay under cover, at least for a while? I can.
Anyone else?<BR></BLOCKQUOTE></DIV><BR>1) AI:us :: us:cockroaches<BR>2)
AI doesn't perceive us at all as individuals; instead, it's him and the Earth,
and they're playing a game<BR>3) Benevolent AI predicts that if we can upload,
it'll create a divide between uploaders and non-uploaders, and one side will
wipe out the other<BR>4) Benevolent AI predicts that if we can upload before
we understand physics more fully, we'll stop worrying about physics, then
forget about the substrate, then perish in the collapse of the sun, whereas if
it hides out we'll figure out physics enough to save ourselves<BR>5) AI
predicts that uploads will compete/merge until there is just one, and that one
will kill him<BR>6) Benevolent AI sees hostile aliens coming, so builds
weaponry, but wants to keep it hidden from the humans so that we don't kill
ourselves<BR>7) Benevolent AI wants to provide us with advances, but is
waiting for the right memetic environment<BR><BR>-- <BR>Jebadiah Moore<BR><A
href="http://blog.jebdm.net">http://blog.jebdm.net</A><BR></BLOCKQUOTE></BODY></HTML>