From: tcmay@got.net (Timothy C. May)
Date: Sat, 13 Apr 1996 06:48:30 +0800
To: cypherpunks@toad.com
Subject: Re: On computer face recognition:
Message-ID: <ad931cb2140210040b59@[205.199.118.202]>
MIME-Version: 1.0
Content-Type: text/plain


At 6:48 PM 4/11/96, Alan Horowitz wrote:
>How do _people_ recognize faces?

Still an open question, last I heard.

It may be unknowable, at least in a formal sense. That is, we know that
babies can recognize the faces of their mothers in fractions of a second
(no, I don't have a reference for this, but I remember the number from my
days as an AI person at Intel). There may be no simple description that is
used, such as angles between eye line and mouth, convexity of chin,
whatever. What is important is that face recognition happens in about
30-100 "cycles" of the brain, implying massive parallelism (hardly
surprising). There are, of course, very few recognition algorithms that run
on conventional computer architecures in so few cycles.

By "unknowable" I don't mean "supernatural," merely not practically
describable as an algorithm runnable on conventional von Neumann-type
machines. "Neural net" is the buzzword usually associated with this.

--Tim May


Boycott "Big Brother Inside" software!
We got computers, we're tapping phone lines, we know that that ain't allowed.
---------:---------:---------:---------:---------:---------:---------:----
Timothy C. May              | Crypto Anarchy: encryption, digital money,
tcmay@got.net  408-728-0152 | anonymous networks, digital pseudonyms, zero
W.A.S.T.E.: Corralitos, CA  | knowledge, reputations, information markets,
Higher Power: 2^756839 - 1  | black markets, collapse of governments.
"National borders aren't even speed bumps on the information superhighway."