Am 25.03.2013 um 14:07 schrieb "Onno Meyer" <[email protected]>:

> Hello Thomas,
> 
> I found that we had brainstormed on GURPSnet about skill programs before, in 
> January 2012. Then it was about my large robot cruiser.
I will look that up when I am back home.
> 
> Thomas wrote:
>>> Say you have an AI trained to an impressive skill level and you want it to 
>>> self-learn to become better. Like any other character in that situation, it 
>>> would be learning without a teacher, without books, on the job. The GM 
>>> should ask the players just what that job is …
>> 
>> Not one AI, dozens, hundreds. Each time a device returned from the field it 
>> exchanges its new experiences with all the others. The knowledgebase of 
>> Version 1 is used to create that for Version II, so that your Battlesuit Mk. 
>> VII can look back on the experience of 200 years of infantery combat even on 
>> the day it leaves the factory.
> 
> Agreed, with several caveats:
> You have to integrate the lessons from the different AIs into one. How? 
> Having different AIs fight and learn against each other puts you at the mercy 
> of the artificial scoring rules of the wargame.
> Will the special conditions of the most recent wars hide universal truths? If 
> an AI had to learn from recent US Army experience, would it be any good 
> against a 'peer competitor'?
> Once you reach a reasonably high skill, will real mission data improve the 
> program, or just confirm what it already knew?
As for a chess program or a program that drives a car, I would expect that the 
foundation is laid by some manual programming, based on human experience, 
theory of games, statistics and alike, the rest is exceeding the 
database/knowledge base. I am not an expert for Artificial Learning, but what I 
read about the topic, it looks for me that the basic concepts already exists. 
Extrapolated this into the future should allow for "indefinite learning 
capability" in a robot brain from the Traveller universe ...

I put that in quotes because for each learning process there is point where the 
effort to gain another level gets indefinite. GURPS reflects that by increasing 
the point costs for each level. Learning a Martial Art is another sample for 
that.

For our battle suits, after some iteration the merging of experience will be 
not more than fostering already known "facts" until something changes on the 
battle field.

> 
> Another problem is that what's good for the goose is good for the gander -- 
> if I want sensor ops skill programs to see through stealth, but not tactics 
> skill programms to out-think human player characters, what can I do? And if 
> the program makes it possible to see a stealthy battlesuit with a 
> helmet-mounted sensor of another battlesuit, what does that do to the 
> stealth-vs-sensors race in grav tanks?

SensorOp is in my eyes a very problematic skill, especially in the spotlight of 
the discussion of "artificial skills".

First, if your sensor is not more than a regular torchlight, you may have any 
skill level you want, you will not detect a starship in space under stealth 10 
km away. On the other side, with enough data just basic knowlegde of the craft 
allow you to find everything.

What I want to say is that success (in the real world) depends more from volume 
and quality of data than from skill (as far as the skill exceeds some basic 
level, of course). The data quality in turn depends from the sensor equipment 
(and that, of course, from maintenance, and that is dependent from skill …).

So there are skills that have a "highest useful level", any access is wasted.

> 
> The effect of sensor skill is capped since no detection is possible if total 
> non-skill modifiers add up to -10 or worse (VE170).

What I said above ...

> [...]

> Regards,
> Onno
> _______________________________________________
> GurpsNet-L mailing list <[email protected]>
> http://mail.sjgames.com/mailman/listinfo/gurpsnet-l

--
Thomas Thrien
Allemagne

Geo 51° 28' 12" N 7° 32' 17" E

Es heißt, der Klügere gibt nach. Doch wenn die Klügeren immer nachgeben, dann 
passiert nur noch, was die Dummen wollen …


Of course it has a meaning when a black cat crosses your way from left to right 
…
It means, that the cat, coming from your left, wants to go somewhere right from 
you ...
_______________________________________________
GurpsNet-L mailing list <[email protected]>
http://mail.sjgames.com/mailman/listinfo/gurpsnet-l

Reply via email to