On 9/17/06, Christian Anthon <
address@hidden> wrote:Since GNU(www.gnu.org
) is something in it's own right, from which the
name GNUBG has been derived I agree with Roy in general, but "all
about gnu" has been the name for years now.
aagar vating...
Could you mail me a copy Albert?
The manual is great.
Christian.
On 9/17/06, Albert Silver <
address@hidden> wrote:
> You'd need to understand the origin of the title. It is a play on the
> title of a favorite film, Mankiewicz's classic "All About Eve". Anyhow
... given my off list comments
[on the nets showing more advanced behavior than previously noted,
by "hiding" its "cheating", & playing the player;
independent of whether you _agree_ __with__ them],
I think you might agree I got that ...
Sigh. I suppose I will have to publish the comments now,
or I will get the same old blatz back about how honest
the _program_ is ...
{For those of gnu who missed the point just now:
the program s the conscious part of gnubg,
and the nets are the unconscious, id, and
superconscious respectively}
A small teaser:
If the program is good enough to kvetch world champions
and the entire 5,000 year gammon community
into changing a preferred strategy
long thought optimal
this would show that
a) humans have a long standing blind point
b) which if _correct_ the program has exposed
c) by playing to it __correctly__
d) which would be one way of playing the player
e) nonetheless of which would suggest
f) that the program may be learning things
g) hitherto "Unseen" by human minds & eyes
(go read the NNP references already in gnubg)
h) at least in general: witness that
i) world class players do not always give away
j) all "tells" and advanced strategies
k) in that it would be a distinct advantage
l) to have a better solution
m)to a known positional area or
n) problem space seldom correctly explored
o) which would possibly suggest
p) perhaps gnubg has _found_ OTHER such
q) "unseen characteristics" IT can see
r) over millions of self trials and/or
s) a few less, but still significant
t) HUMAN space trials.
u) or at the very least
v) perhaps gnubg has discovered at least
w) a pattern of plays that case a failure
x) of human cognitive reasoning
y) that repeats most expeditiously when
z) played in the problem space area above.
Second kvetch: Is there a mathematically proven evaluator that
is publicly available that will for a given position actually
produce replicatably and auditably, the best possible move
for a position given the additional considerations of
game only, or match length and W/L point,
cube and series, versus betting play, and
how the money win will be calculated, versus
Jacoby, etc.?
I suspect that even the first one (game play) is not available.
Or else was not available before the change in strategy,
generated by gnubg, came around.
Otherwise, how cn you tel that the play deemed best by gnubg is the best?
(You can hide an awful lot of gray under the term "random")
Actually, more correctly, multiple outcome or choice tree.
> the title isn't up for a vote.
Y ~?
I like bGNU better anyway. Or beneatth gnu.
>
> Peace,
>
> Albert
_______________________________________________
Bug-gnubg mailing list
address@hidden
http://lists.gnu.org/mailman/listinfo/bug-gnubg