thoughts from FOSDEM, status

[Kévin Ottens]

Le Dimanche 20 Mars 2005 19:25, Aaron J. Seigo a écrit :
> for input or results discovery?

For result discovery... but as I said I've been asked how it could look like 
on an application/agent mapping. I never said it was the best idea...

> for traversal/discovery of the mesh... i don't see the benefits. feel free
> to enlighten me =)

I see the benefit (surely not with an application/agent mapping) because you 
could make the system (the system, not necessarily the agents) to learn some 
things thanks to the user behavior. I know at least one information retrieval 
system working like this.

But, please don't push me at presenting a more complete solution... I'm not 
pushing the use of multi-agent systems in Klink, I just want to clarify some 
points and in particular to avoid giving up on the idea on some wrong 
assumptions or rethoric arguments like "it's not so marketed anymore, so it's 
not useful". ;-)

> there will certainly be multiple simultaneous searches, and perhaps even
> for the same query. what concerns me is the concept of agent
> specialization, which is to say "this agent looks at X, this agent knows
> how to look at Y". it's a very seductive concept, especially if completely
> divorced from the requesting applications (e.g. i do not see the need for a
> "kmail query agent" and a "juk query agent"). despite its seductive nature,
> i really wonder at its added complexity.

Once again, I never said that I see a need for such a mapping.

> it's not marketted quite so much anymore ;) it's gone the way of hard AI it
> seems: reality set in and people are using it where it makes sense instead
> of the utopian concepts of world domination previously spewed by the
> marketing arms of those doing the research.

I admit that I don't know to which extend it has been marketed in the 
non-academic world. =)
But I know it's very fashioned currently in the academic world... And we 
regularly see things that put the label "multi-agent systems" on good old 
expert systems... tssss...

> in particular, search agents for public use are non-existent. at least, i
> don't see them in popular use.

And this particular example is most of the time not about multi-agent 
systems... A non trivial amount of researchs done around search agents are 
just about distributed databases without any real multi-agents concern.

> certainly; you're right this was a bit of a heavy handed, sweeping
> statement. i'm aware of MA systems used by, for instance, military
> organizations to study battlefield situations.

It's generally easy to apply MAS when you have a real life metaphor, 
battlefield being a good example. Network bandwidth use some results of the 
ants algorithms too coming from the multi-agents field.

> they seem to map well onto 
> systems which are themselves multi-agent in some way.

That's a weird statement... multi-agent systems map well on multi-agent 
systems. Wow! We're lucky. ;-)

Ok, I stop kidding... I guess I understood.

It's clearly a problem of metaphor and on how you approach the problem you 
want to solve. I admit that sometimes it's not that easy to find the right 
angle to look at a problem... it's true for other kinds of engineering 
though.

> i'm not sure this is the case here.

I'm not sure either... But I think that generally it needs more efforts at the 
design stage to find a way to use agents, or to decide (using a set of 
criteria) that agents are not the way to go.

> what would be the benefits in this case?

> how would the agents be defined?

It'll require more work than answering a mail to decide this. =)

> how would it simplify or empower (either is good =) the system?

The system would gain more flexibility, be able to adapt to the user needs, 
and answer questions you're not expecting the user to ask.

In my team we try to use some properties of the multi-agent systems giving 
them the ability to adapt to their environment, and in particular the system 
design is changing (the design of the system can be seen as how the agents 
are linked, the social relations, hence why it's able to change and evolve 
during the system lifetime).

> by treating different "types" of information as distinct.

It's a flaw of the application/agent mapping... not of multi-agent systems 
themselves.

> but the idea of application agents is not useful imo.

I guess that's the problem in this whole discussion. We started on the wrong 
basis, assuming only application/agent mapping could be used. An infon/agent 
(I hope I got the infon name correct =)) mapping could be more interesting 
for example.

Regards.
-- 
Kévin 'ervin' Ottens, http://ervin.ipsquad.net
"Ni le maître sans disciple, Ni le disciple sans maître,
Ne font reculer l'ignorance."