Well, here's my two cents.

USING AI TO AID

Some players may use LLMs to aid them in their playing of the game. Let's
look at the rules for precedent - Agora treats a player as a single entity,
so if you're playing, that's still as much of an entity as you playing with
augmentation. My personal opinion is that it's kind of cheating, because
/you/ should be playing, not you with training wheels, but Agora is first
and foremost a game about rules, so the rules say it's okay.

USING AI AS A PLAYER

Personally I have no solid evidence with which to argue against this point.
Although agentic LLMs aren't as reliable as humans, there are also humans
who screw things up all the time (c.f. my auction recently). So there's no
reason to exclude LLMs based on this, because that would be a double
standard. However, agentic ultimately are controlled by a player, with
permission to control the minutiae. So then, let's look at precedent. I
remember an message (either by email or Discord, I can't remember) from
Janet, in which she expressed her approval of treating me and my headmates
as separate players if necessary. Although we self-evidently haven't taken
that offer up, it still highlights an important point - players must
function as one singular entity. You see, I don't control my headmates. The
difference is, using agentic LLMs is like having an owner having dogs on a
leash; the dogs can walk around as they please, but they're ultimately
limited by both the length of the leash and the position of the owner. The
headmate case is more like a pack of wolves; the dogs have some sort of
intercommunication techniques, but it's still ultimately unbounded. So, if
you use, say, 20 agentic LLMs, and you make the leash as short as possible,
you're essentially being the same player 20 times.

THE RULES

I mean - let's figure this out like a judgement. R869, in its first
paragraph, states that:
>    Any entity (including a group of entities who confederate with the
>    intent of forming a single person under this Rule) that is or ever
>    was able to willingly communicate original ideas is a person.
>    Rules to the contrary notwithstanding, no other entities are
>    persons.

So, firstly, I call to attention that it says that you're allowed to be a
confederating group of entities, backing up my earlier point. But, you must
be "able to willingly communicate original ideas". Can an agentic LLM
communicate original ideas? Arguably, kind of. But is it willingly doing
so? Arguably, less so. But, let's give AI the benefit of the doubt.
But, the nail in the coffin is really
>    The basis of a person is the set of all persons that are
>    (recursively) part of em, in addition to emself. Rules to the
>    contrary notwithstanding, a person CANNOT become Registered if eir
>    basis overlaps with that of any current player.
QED. An agentic LLM should essentially be treated as the controlling
player + the machine, and so if you register with an LLM, you can't
register with another. This addresses the AI-swarm fears, because a single
player can't register, say, 400 AIs in eir name.

CONCLUSION

But, of course, we still haven't resolved what happens if one person uses
an LLM. I mean - there's nothing against it. Although some people may have
distaste for it, the rules have nothing against it, and that's what
matters, not opinion. We could just edit R869 to clarify the AI stance
further, but the precedent is already there.
--
juniper

Reply via email to