From: "Phillip M. Hallam-Baker" <hallam@ai.mit.edu>
Date: Sun, 15 Dec 1996 16:00:20 -0800 (PST)
To: "'cypherpunks@toad.com>
Subject: Re: Why PICS is the wrong approach
Message-ID: <01BBEABA.4237CB30@crecy.ai.mit.edu>
MIME-Version: 1.0
Content-Type: text/plain




Dale Thorn <dthorn@gte.net> wrote in article <58q40v$k22@life.ai.mit.edu>...
> Vladimir Z. Nuri wrote:

> If Firefly is an example of what PICS is or could become, the hell with
> PICS.  Firefly encourages and rewards group behavior and suppresses
> individuality.  Firefly would reward the discussion of the latest album
> by a Columbia or Capitol artist, and discourage discussion of material
> from independent (real independent) labels.  I know because I've been
> there and spent quite a bit of time trying to get a rating.

I think there are two issues here, Firefly and PICS. Confusing one
with the other is a bad thing. PICS is simply one way of applying labels
to content. Its sole reason for being was to head off the CDA. Now
that the judicial route has been taken I see little likelyhood that PICS
can succeed since either the supreme court will uphold the CDA
and we have the Singapore scenario or the CDA gets booted out 
and the matter is over.

If the US congress had wanted to do any good instead of making
itself look good then the voluntary approach of PICS with the
multiple rating schemes was the one most likely to work. I don't
think that it would stop 16 year olds from seeing pornography
but since children of that age are in most jurdisdictions permitted
by law to engage in sex on their own account it seems a bit bizare
to prohibit them from seeing pictures of sexual acts.

The problem with Firefly is that its a good(ish) idea baddly 
implemented. The much vaunted "agent" technology uses only a
very primitive nearest neighbour type match. There is no attempt
to draw structural inferences from the material, such as abstractions.
For example if I enjoy Dire Straights and Peter Gabriel then a
shop assistant would peg me as a late 70s rock fan and point
me towards the Fleetwood Mac and such. Firefly has 50% of
the structure needed to produce interesting matches but 
lacks the ability to make inferences. At least if it is the same
technology as Ringo, the previous generation.

This shortcomming of Ringo means that it is very slow work
training it. To get a useful measure it needs hundreds of data
points. When I visited I was faced with page after page of US
90s chart bands which I've not heard of and have no interest 
in. 

	Phill