Re: Bots (was: thoughts on new modes (Ie))

[Jukka Tapani Santala <e75644@xxxxxxxx>]

On Thu, 26 Nov 1998, Bjorn Reese wrote:
> [ First, apologies for being slightly off-topic ]

*snip* Speaking off off-topic...

> and I could continue ad infinitum. With the paramounting interest in the
> field of Intelligent Agents, I predict that sooner or later we will also
> see bots that are participating in the discussions.

Sooner. I have for years used rather simplistic bots to "augment"
discussions on channels. Eliza/Racter types, a relatively simple
pattern-oriented "expert system", have been very succesful in tricking
people on MUD type systems into thinking they're real persons. More
advanced models are getting more and more convincing, even beating Turing
tests in limited subjects.

In fact, I was told that with an older version of one of my Eliza/Ractor
style IRC-bots somebody had came to the main channel, appearing very
distraught, asking people to "make him talk to me again". She thought the
bot was a real person and had got angry at her, because it was programmed
to only respond when it's name was said. (I met her later and
verified/explained the story)

Ofcourse, in general, such bots haven't got very warm reception around
channels. It's not that they're dumper than your average AOL:er, but as
some people put it to me, "We sit on front of our computers and get
online, not to talk to some program, but to talk to other people".
Strange, this world we live in, isn't it? But then, how can they tell
there really is a real person out there, on the other side.

Oh yeah, and another extreme of the scale was met, I guess, when a real
user got kicked off the aforementioned channel, when they thought I had
brought one of my bots back in. I get people occasionally mix me for one
of my bots; but then, that's part of the game.

I guess what I'm trying to say here is that we've long since passed past
that point where anybody could, with sufficient certainty, tell what is a
bot and what is a human on IRC. Most are something between, clients with
heavy scripting, anyway. So it's kinda silly to try to limit the other
kind.

Some more thoughts:

 - bots have reached the same kind of "status value", perhaps even more
   so, that for example cars have in real life. The analogue is quite
   clear, and visible.
 - a more just way of deciding on "unneccessary" would be to disconnect
   clients that idle for longer than some given time. It's even been tried
   on many codebases (And withmy luck exists in this;). Unfortunately,
   this just leads to people circumveiting them by timer-actions. (Altough
   many do them anyway believing it helps avoid lag-disconnect)

So in summary, I don't think trying to stop/limit bots really has much
future. Ofcourse, this doesn't mean having servers do more things for
users isn't a worthwhile goal.

 -Donwulff