On 22/04/24 22:23, occam wrote:
> On 20/04/2024 00:36, Peter Moylan wrote:
>> On 20/04/24 00:47, Bertel Lund Hansen wrote:
>>> Hibou wrote:
>>>
>>>> Perhaps self-taught AI marks the divide, the point at which
>>>> humans can no longer unravel its inner workings.
>>>
>>> Isn't that the case with the Go program that beat the world
>>> champion? I read that the decisive move was one that a human
>>> never would have thought of.
>>
>> The example that sticks in my mind was a program designed to prove
>> theorems in mathematics. It proved that the base angles in an
>> isosceles triangle are equal because ABC is congruent to ACB.
>> That's a proof that probably would not have occurred to humans.
>>
>
> I doubt however that the geometry program in this case learned the
> rules of geometry from naught. The case Bertel is referring to is
> the AI program AlphaGO (by Deepmind). In this case the program
> learned to play 'Go' by itself, by playing games against itself. By
> tallying up scores and eliminating non-optimal moves, it was very
> much an exercise in self-learning.

AlphaGO had to be told the rules of the game to begin with. Similarly,
the geometry program started with axioms (not theorems). There has to be
a starting point.

--
Peter Moylan http://www.pmoylan.org
Newcastle, NSW

On Sun, 21 Apr 2024 21:19:45 +0300, Phil Carmody <pc+usenet@asdf.org>
wrote:

>Bertel Lund Hansen <gadekryds@lundhansen.dk> writes:
>> Rich Ulrich wrote:
>>> Wouldn't it be nice if AI would fill in my credit card number
>>> for me when I order something? along with my address, etc.?
>>> Yeah, Amazon knows that because it reads my cookies and
>>> knows who I am. We don't call it AI.
>>
>> I certainly don't. There's no I involved, just a mere computer look-up
>> in a table. The same goes for your other examples.
>>
>> A pocket calculater telling me that 5*6=30 has no AI either.
>
>But only an AI can tell you that 60*60*600 = 216,000.
>
>At least that's what one told me last week.
>

I read a couple of months ago that ONE of the chat-bot providers
bragged that they had incorporated an arithmetic fact-checker
which should prevent that error.

AI -
Here in the U.S., I see reports now and then of lawyers being
chastised by courts for their AI-created briefs when they (still)
include false references. That was a problem I saw mentioned
near the start, with article author, title, and page number being
invented for some journal name, vol, etc., that was real.

I just now recognized that 'chastise' is an unusual word in that
we Americans do not use -ize for the -ise. I typed -ize and it
looked so wrong.
https://books.google.com/ngrams/graph?content=chastize%2Fchastise&year_start=1800&year_end=2019&corpus=en-US-2019&smoothing=0

for AmE, -ize hit its peak around 1980, about 7%, but fell again
to below 1%.

--
Rich Ulrich

On 23/04/2024 01:48, Peter Moylan wrote:
> On 22/04/24 22:23, occam wrote:
>> On 20/04/2024 00:36, Peter Moylan wrote:
>>> On 20/04/24 00:47, Bertel Lund Hansen wrote:
>>>> Hibou wrote:
>>>>
>>>>> Perhaps self-taught AI marks the divide, the point at which
>>>>> humans can no longer unravel its inner workings.
>>>>
>>>> Isn't that the case with the Go program that beat the world
>>>> champion? I read that the decisive move was one that a human
>>>> never would have thought of.
>>>
>>> The example that sticks in my mind was a program designed to prove
>>> theorems in mathematics. It proved that the base angles in an
>>> isosceles triangle are equal because ABC is congruent to ACB.
>>> That's a proof that probably would not have occurred to humans.
>>>
>>
>> I doubt however that the geometry program in this case learned the
>> rules of geometry from naught.  The case Bertel is referring to is
>> the AI program AlphaGO (by Deepmind). In this case the program
>> learned to play 'Go' by itself, by playing games against itself. By
>> tallying up scores and eliminating non-optimal moves, it was very
>> much an exercise in self-learning.
>
> AlphaGO had to be told the rules of the game to begin with. Similarly,
> the geometry program started with axioms (not theorems). There has to be
> a starting point.
>
True. But look at the facts. The 'rules' of the game of Go are so simple
- and so few - that they are not worthy of comparison with geometry:
1- You can put a piece anywhere on the board which is not already
occupied by another piece; 2- black and white pieces alternate in play.
The rest is just computing the score at any given instant.

In contrast, the axioms of (Euclidian) geometry assume the statements
of: 1- point; 2- line ; 3- circle (radius); 4- right angle; 5 -
parallelism ; 6- infinity; 7- intersection (of lines); 8 - proof (or
equivalence). Given all the above, there has to be more fiddly rules -
specified by humans - of what the AI can do and not do. Without knowing
the details of the geometry program, I'd say there is at least an order
of magnitude of 'coaching rules' before such a program can run.

On 2024-04-23 03:20:13 +0000, Rich Ulrich said:

> On Sun, 21 Apr 2024 21:19:45 +0300, Phil Carmody <pc+usenet@asdf.org>
> wrote:
>
>> Bertel Lund Hansen <gadekryds@lundhansen.dk> writes:
>>> Rich Ulrich wrote:
>>>> Wouldn't it be nice if AI would fill in my credit card number
>>>> for me when I order something? along with my address, etc.?
>>>> Yeah, Amazon knows that because it reads my cookies and
>>>> knows who I am. We don't call it AI.
>>>
>>> I certainly don't. There's no I involved, just a mere computer look-up
>>> in a table. The same goes for your other examples.
>>>
>>> A pocket calculater telling me that 5*6=30 has no AI either.
>>
>> But only an AI can tell you that 60*60*600 = 216,000.
>>
>> At least that's what one told me last week.
>>
>
> I read a couple of months ago that ONE of the chat-bot providers
> bragged that they had incorporated an arithmetic fact-checker
> which should prevent that error.
>
> AI -
> Here in the U.S., I see reports now and then of lawyers being
> chastised by courts for their AI-created briefs when they (still)
> include false references. That was a problem I saw mentioned
> near the start, with article author, title, and page number being
> invented for some journal name, vol, etc., that was real.
>
> I just now recognized that 'chastise' is an unusual word in that
> we Americans do not use -ize for the -ise.

What a surprize (as Jane Austen spelled it)!

> I typed -ize and it
> looked so wrong.
> https://books.google.com/ngrams/graph?content=chastize%2Fchastise&year_start=1800&year_end=2019&corpus=en-US-2019&smoothing=0
>
>
> for AmE, -ize hit its peak around 1980, about 7%, but fell again
> to below 1%.

--
Athel cb

occam hat am 23.04.2024 um 08:13 geschrieben:

> True. But look at the facts. The 'rules' of the game of Go are so simple
> - and so few - that they are not worthy of comparison with geometry:
> 1- You can put a piece anywhere on the board which is not already
> occupied by another piece; 2- black and white pieces alternate in play.
> The rest is just computing the score at any given instant.

Your rules are not enough to play Go. You must add at least:
3 - Pieces on the board never move. They can only be taken off the board
when they have no adjacent free point.
4 - You cannot put a piece where it has no adjacent free point. It would
go off the board instantly because of rule 3.
5 - You ignore 4 if your piece also takes the last free point of one or
more pieces of your opponent. Instead, that piece or those pieces of
your opponent are taken off the board.
6 - You cannot put a piece where it reproduces a former whole board
configuration.

Example on a 5 by 5 board, odd numbers for Black and even numbers for
White. Some very bad plays in order to keep my explanation short.

.. . . . .
.. . . . .
.. . 1 . .
.. . 3 2 .
.. 5 4 . 6

.. . . . .
.. . . . .
.. . 1 . .
.. . 3 2 .
.. 5 4 7 6
7 is legal, notwithstanding rule 4, because it takes the last free point
of 4 and takes it off the board. Now we have
.. . . . .
.. . . . .
.. . 1 . .
.. . 3 2 .
.. 5 . 7 6

.. . . . .
.. . . . .
.. . 1 . .
.. . 3 2 .
.. 5 8 7 6
8 is illegal, notwithstanding rule 5, because after taking 7 away it
reproduces
.. . . . .
.. . . . .
.. . 1 . .
.. . 3 2 .
.. 5 4 . 6

If 8 were legal, Black could play 9 at 7, then White could play 10 at 8
etc. forever.

On 23/04/24 16:13, occam wrote:
> On 23/04/2024 01:48, Peter Moylan wrote:
>> On 22/04/24 22:23, occam wrote:
>>> On 20/04/2024 00:36, Peter Moylan wrote:
>>>> On 20/04/24 00:47, Bertel Lund Hansen wrote:
>>>>> Hibou wrote:
>>>>>
>>>>>> Perhaps self-taught AI marks the divide, the point at
>>>>>> which humans can no longer unravel its inner workings.
>>>>>
>>>>> Isn't that the case with the Go program that beat the world
>>>>> champion? I read that the decisive move was one that a human
>>>>> never would have thought of.
>>>>
>>>> The example that sticks in my mind was a program designed to
>>>> prove theorems in mathematics. It proved that the base angles
>>>> in an isosceles triangle are equal because ABC is congruent to
>>>> ACB. That's a proof that probably would not have occurred to
>>>> humans.
>>>>
>>>
>>> I doubt however that the geometry program in this case learned
>>> the rules of geometry from naught. The case Bertel is referring
>>> to is the AI program AlphaGO (by Deepmind). In this case the
>>> program learned to play 'Go' by itself, by playing games against
>>> itself. By tallying up scores and eliminating non-optimal moves,
>>> it was very much an exercise in self-learning.
>>
>> AlphaGO had to be told the rules of the game to begin with.
>> Similarly, the geometry program started with axioms (not theorems).
>> There has to be a starting point.
>>
> True. But look at the facts. The 'rules' of the game of Go are so
> simple - and so few - that they are not worthy of comparison with
> geometry: 1- You can put a piece anywhere on the board which is not
> already occupied by another piece; 2- black and white pieces
> alternate in play. The rest is just computing the score at any given
> instant.

I see that there's no rule to say who wins. Is the game-playing machine
supposed to deduce the rule?

I don't know how to play Go myself, and the above rules are not
sufficient to tell me how to play. I'm guessing, though, that there are
rules saying (a) how many pieces can be placed on the board in a single
move, (b) whether those pieces can be a mixed set, or have to be all
black or all white, (c) whether the pieces come from a common pool of
unlimited size, or whether each player only gets a certain quota of
black pieces and a certain quota of white pieces, (d) whether the board
(I assume it's two-dimensional) has a rectangular or hexagonal grid or
some other pattern, and whether it's finite in both directions. If so,
those have to be added to the list of axioms.

I've now read Silvano's additions to your rules, but those are still not
sufficient to explain the game to someone who doesn't know it. I think I
can deduce the meaning of "free point" from his examples, but is the
game-playing software given enough examples for it to be able to deduce
the meaning of "free point"? Also, I still can't guess what sort of move
stops the game so that a winner is declared.

> In contrast, the axioms of (Euclidian) geometry assume the
> statements of: 1- point; 2- line ; 3- circle (radius); 4- right
> angle; 5 - parallelism ; 6- infinity; 7- intersection (of lines); 8 -
> proof (or equivalence). Given all the above, there has to be more
> fiddly rules - specified by humans - of what the AI can do and not
> do. Without knowing the details of the geometry program, I'd say
> there is at least an order of magnitude of 'coaching rules' before
> such a program can run.

--
Peter Moylan http://www.pmoylan.org
Newcastle, NSW

On 23/04/2024 13:24, Peter Moylan wrote:
> On 23/04/24 16:13, occam wrote:
>> On 23/04/2024 01:48, Peter Moylan wrote:
>>> On 22/04/24 22:23, occam wrote:
>>>> On 20/04/2024 00:36, Peter Moylan wrote:
>>>>> On 20/04/24 00:47, Bertel Lund Hansen wrote:
>>>>>> Hibou wrote:
>>>>>>
>>>>>>> Perhaps self-taught AI marks the divide, the point at
>>>>>>> which humans can no longer unravel its inner workings.
>>>>>>
>>>>>> Isn't that the case with the Go program that beat the world
>>>>>> champion? I read that the decisive move was one that a human
>>>>>> never would have thought of.
>>>>>
>>>>> The example that sticks in my mind was a program designed to
>>>>> prove theorems in mathematics. It proved that the base angles
>>>>> in an isosceles triangle are equal because ABC is congruent to
>>>>>  ACB. That's a proof that probably would not have occurred to
>>>>> humans.
>>>>>
>>>>
>>>> I doubt however that the geometry program in this case learned
>>>> the rules of geometry from naught.  The case Bertel is referring
>>>>  to is the AI program AlphaGO (by Deepmind). In this case the
>>>> program learned to play 'Go' by itself, by playing games against
>>>>  itself. By tallying up scores and eliminating non-optimal moves,
>>>>  it was very much an exercise in self-learning.
>>>
>>> AlphaGO had to be told the rules of the game to begin with.
>>> Similarly, the geometry program started with axioms (not theorems).
>>> There has to be a starting point.
>>>
>> True. But look at the facts. The 'rules' of the game of Go are so
>> simple - and so few -  that they are not worthy of comparison with
>> geometry: 1- You can put a piece anywhere on the board which is not
>> already occupied by another piece; 2- black and white pieces
>> alternate in play. The rest is just computing the score at any given
>>  instant.
>
> I see that there's no rule to say who wins. Is the game-playing machine
> supposed to deduce the rule?

I admit my rules were oversimplified and incomplete. (see Silvano). The
object is to surround completely the pieces of the opposite colour -
whence you remove those 'captured' pieces from the board. Whoever has
the most number of pieces in the end (simple count) wins. The metric for
'whose winning' is much simpler to compute than, say, in chess.

>
> I don't know how to play Go myself, and the above rules are not
> sufficient to tell me how to play.

The rules are not the main issue. You can learn them in 3 minutes flat.
It's the search space that's the challenge - much larger than even that
of chess.

As a result of this sub-thread, I rewatched the AlphaGo video of the
event of 2016. It is well worth the 1 hr:29 min time. It is one of the
most gripping videos I have watched in a long, long time. It could well
be dubbed as "the beginning of the end of human dominance in game play".
And that's just the start.

https://www.youtube.com/watch?v=WXuK6gekU1Y

> I'm guessing, though, that there are
> rules saying (a) how many pieces can be placed on the board in a single
> move, (b) whether those pieces can be a mixed set, or have to be all
> black or all white, (c) whether the pieces come from a common pool of
> unlimited size, or whether each player only gets a certain quota of
> black pieces and a certain quota of white pieces, (d) whether the board
> (I assume it's two-dimensional) has a rectangular or hexagonal grid or
> some other pattern, and whether it's finite in both directions. If so,
> those have to be added to the list of axioms.
>
> I've now read Silvano's additions to your rules, but those are still not
> sufficient to explain the game to someone who doesn't know it. I think I
> can deduce the meaning of "free point" from his examples, but is the
> game-playing software given enough examples for it to be able to deduce
> the meaning of "free point"? Also, I still can't guess what sort of move
> stops the game so that a winner is declared.
>

Peter Moylan hat am 23.04.2024 um 13:24 geschrieben:

> I see that there's no rule to say who wins.

Let me answer between the lines. There are slightly different rules, but
I do not want to write a detailed article and the result is in most
cases exactly the same. At the end of the game, i.e. when both players
agree that there are no more sensible moves available:
- First, determine the territories, i. e. the parts of the board
surrounded only by pieces of the same colour.
- Second, put the opponent's pieces taken off the board into the
opponent's territory. The same applies to opponent pieces inside a
territory which cannot form a territory of their own.
- Third, count who has the most free points surrounded only by their own
pieces. That's the winner.

B for Black and W for White:
.. . B W .
.. . B W .
.. . B W .
.. B B W .
.. B W W .
8 points for Black and 5 for White. The normal Go board is 19 x 19.

> Is the game-playing machine
> supposed to deduce the rule?

Of course not. Its programmers will settle for one set of rules.

> I don't know how to play Go myself, and the above rules are not
> sufficient to tell me how to play. I'm guessing, though, that there are
> rules saying (a) how many pieces can be placed on the board in a single
> move,

ONE!

> (b) whether those pieces can be a mixed set, or have to be all
> black or all white,

Like in chess, it's Black vs. White. Have you ever seen a chess player
move pieces of both colours in the same game???

> (c) whether the pieces come from a common pool of
> unlimited size, or whether each player only gets a certain quota of
> black pieces and a certain quota of white pieces,

Theoretically unlimited. In practice, usually 100-150 pieces each before
both players agree that there are no more sensible moves available.

> (d) whether the board
> (I assume it's two-dimensional) has a rectangular or hexagonal grid

Neither. It's a 19x19 square.

> or
> some other pattern, and whether it's finite in both directions. If so,
> those have to be added to the list of axioms.

It normally is. There are some odd variants, with different board shapes
or the possibility to play more than one piece at a time, but they're
nonstandard, just like in chess.

Anyway, you just taught me that some aspects are obvious to any player,
even an absolute beginner, but not to an alien or a machine.

> I've now read Silvano's additions to your rules, but those are still not
> sufficient to explain the game to someone who doesn't know it. I think I
> can deduce the meaning of "free point" from his examples, but is the
> game-playing software given enough examples for it to be able to deduce
> the meaning of "free point"? Also, I still can't guess what sort of move
> stops the game so that a winner is declared.

The 5x5 board again, this time with the number of adjacent free points
when the board is completely empty. Diagonals do not count. I think it
is very trivial to tell that to a programme.
2 3 3 3 2
3 4 4 4 3
3 4 4 4 3
3 4 4 4 3
2 3 3 3 2

Back to my previous game, let's imagine Black is completely stupid and
keeps playing.
.. . B W .
.. . B W .
.. . B W .
.. B B W .
.. B W W .
Inside his own territory reduces it by one (in other rule sets, it gives
one point for the new piece and subtracts one for the now occupied
point) and inside White's territory it also reduces his territory by
one, because that stone will be taken away at the end and put inside
Black's territory.

It might be a good idea only if White had fallen asleep and keeps
passing, i.e. not playing anything.
.. . B W 7
.. . B W 3
.. . B W 1
.. B B W 5
.. B W W 9
Black plays 1 and loses one point, then 3 and loses another point, then
5 and 7 (the same). Only after 9 the White stones would have no free
points left and be taken off the board. But White can play 8 after 7 and
then it's the black pieces which are taken off the board.

Now we have 4 Black pieces off the board reducing Black's territory at
the end of the game, so Black has only 4 points left.
.. . B W y
.. . B W x
.. . B W x
.. B B W y
.. B W W W

If Black plays at one of the x points, White plays at the other and
Black can never play at y, because he has no free points left and White
still has one.

Anyway, if you really want to know more,
<https://en.wikipedia.org/wiki/Go_(game)> should be a good starting point.

On Tue, 23 Apr 2024 14:16:18 +0200, occam <occam@nowhere.nix> wrote:

>On 23/04/2024 13:24, Peter Moylan wrote:
>> On 23/04/24 16:13, occam wrote:
>>> On 23/04/2024 01:48, Peter Moylan wrote:
>>>> On 22/04/24 22:23, occam wrote:
>>>>> On 20/04/2024 00:36, Peter Moylan wrote:
>>>>>> On 20/04/24 00:47, Bertel Lund Hansen wrote:
>>>>>>> Hibou wrote:
>>>>>>>
>>>>>>>> Perhaps self-taught AI marks the divide, the point at
>>>>>>>> which humans can no longer unravel its inner workings.
>>>>>>>
>>>>>>> Isn't that the case with the Go program that beat the world
>>>>>>> champion? I read that the decisive move was one that a human
>>>>>>> never would have thought of.
>>>>>>
>>>>>> The example that sticks in my mind was a program designed to
>>>>>> prove theorems in mathematics. It proved that the base angles
>>>>>> in an isosceles triangle are equal because ABC is congruent to
>>>>>>  ACB. That's a proof that probably would not have occurred to
>>>>>> humans.
>>>>>>
>>>>>
>>>>> I doubt however that the geometry program in this case learned
>>>>> the rules of geometry from naught.  The case Bertel is referring
>>>>>  to is the AI program AlphaGO (by Deepmind). In this case the
>>>>> program learned to play 'Go' by itself, by playing games against
>>>>>  itself. By tallying up scores and eliminating non-optimal moves,
>>>>>  it was very much an exercise in self-learning.
>>>>
>>>> AlphaGO had to be told the rules of the game to begin with.
>>>> Similarly, the geometry program started with axioms (not theorems).
>>>> There has to be a starting point.
>>>>
>>> True. But look at the facts. The 'rules' of the game of Go are so
>>> simple - and so few -  that they are not worthy of comparison with
>>> geometry: 1- You can put a piece anywhere on the board which is not
>>> already occupied by another piece;

Two exceptions to that:

1. You can not place a stone where it is surrounded by stones of the
opposite color, unless, by so-doing it surrounds and therefore
captures one or more of the stones that would have captured it.

2. You can not repeat the position

>>>2- black and white pieces
>>> alternate in play. The rest is just computing the score at any given
>>>  instant.
>>
>> I see that there's no rule to say who wins. Is the game-playing machine
>> supposed to deduce the rule?
>
>I admit my rules were oversimplified and incomplete. (see Silvano). The
>object is to surround completely the pieces of the opposite colour -
>whence you remove those 'captured' pieces from the board.

No, the primary object is to surround unoccupied intersections or
groups of unoccupied intersections. Each such unoccupied intersection
is worth one point. Each captured stone is also worth one point, but
unless one or both players is a beginner, there are usually few
captured stones.

>Whoever has
>the most number of pieces in the end (simple count) wins.

No, whoever has the most points (see above) wins.

> The metric for
>'whose winning' is much simpler to compute than, say, in chess.
>
>>
>> I don't know how to play Go myself, and the above rules are not
>> sufficient to tell me how to play.
>
>The rules are not the main issue. You can learn them in 3 minutes flat.

Yes. But you also need to play at least one game to really understand
the rules, and see how the surrounded intersections are counted.

>It's the search space that's the challenge - much larger than even that
>of chess.

Yes. Because it's easiest to surround space in the corners, then the
sides, and finally the center, unlike chess, there are four openings
going on at once. It's often easy to find a good move in one of the
corners, but that's not good enough--you need to find the best move in
the most important corner at the moment. If a player wins one corner
and loses the other three, he will very likely lose the game. A good
move isn't good enough. You need to find the best move.

The board is 19x19. It's been said that if the board were enlarged to
21x21, the influence of one corner on the adjoining corners would be
beyond the grasp of the human mind.

>As a result of this sub-thread, I rewatched the AlphaGo video of the
>event of 2016. It is well worth the 1 hr:29 min time. It is one of the
>most gripping videos I have watched in a long, long time. It could well
>be dubbed as "the beginning of the end of human dominance in game play".
>And that's just the start.
>
>https://www.youtube.com/watch?v=WXuK6gekU1Y
>
>
>> I'm guessing, though, that there are
>> rules saying (a) how many pieces can be placed on the board in a single
>> move,

One.

>> (b) whether those pieces can be a mixed set, or have to be all
>> black or all white,

One black or one white. Just as in chess, one player plays the white
stones, the other black.

>(c) whether the pieces come from a common pool of
>> unlimited size, or whether each player only gets a certain quota of
>> black pieces and a certain quota of white pieces,

One player has 181 white stones, and the other 181 black stones.
That's normally way more than needed.

A minor point. The pieces aren't called "pieces." They're called
"stones" (or in Japanese, "ishi" or "go ishi." They are traditionally
made from slate (black) and shell (white), but such a set of stones
are very expensive. Mine are made from plastic.

>>(d) whether the board
>> (I assume it's two-dimensional) has a rectangular or hexagonal grid or

Square (actually nearly square). 19x19. The game is played on the
intersections, not on the boxes the lines form, as in chess.

>> some other pattern, and whether it's finite in both directions.

Yes.

>> If so,
>> those have to be added to the list of axioms.
>>
>> I've now read Silvano's additions to your rules, but those are still not
>> sufficient to explain the game to someone who doesn't know it. I think I
>> can deduce the meaning of "free point" from his examples, but is the
>> game-playing software given enough examples for it to be able to deduce
>> the meaning of "free point"? Also, I still can't guess what sort of move
>> stops the game so that a winner is declared.
>>
>

On Tue, 23 Apr 2024 14:16:18 +0200, occam <occam@nowhere.nix> wrote:

>As a result of this sub-thread, I rewatched the AlphaGo video of the
>event of 2016. It is well worth the 1 hr:29 min time. It is one of the
>most gripping videos I have watched in a long, long time. It could well
>be dubbed as "the beginning of the end of human dominance in game play".
>And that's just the start.
>
>https://www.youtube.com/watch?v=WXuK6gekU1Y

I never saw this before, but I'm watching now. Very good.

On 23/04/2024 18:00, Ken Blake wrote:
> On Tue, 23 Apr 2024 14:16:18 +0200, occam <occam@nowhere.nix> wrote:
>
>
>> As a result of this sub-thread, I rewatched the AlphaGo video of the
>> event of 2016. It is well worth the 1 hr:29 min time. It is one of the
>> most gripping videos I have watched in a long, long time. It could well
>> be dubbed as "the beginning of the end of human dominance in game play".
>> And that's just the start.
>>
>> https://www.youtube.com/watch?v=WXuK6gekU1Y
>
>
> I never saw this before, but I'm watching now. Very good.

As an ex-tournament chess player yourself, I'd be very interested in
your take on 'intuition', as a result of watching this video.

Lee Sedol is no flake. As the world champion of Go (for the last 10
years), he clearly puts a lot of faith in his intuition and its role in
his playing of the game.

Then comes move #37 (game #2, around 49mins into the video). Poof!

It was dispiriting to witness the realisation of 8 million Koreans (as
well as many other human around the world) that the machine was better
at Go than the best human.

On Wed, 24 Apr 2024 12:36:49 +0200, occam <occam@nowhere.nix> wrote:

>On 23/04/2024 18:00, Ken Blake wrote:
>> On Tue, 23 Apr 2024 14:16:18 +0200, occam <occam@nowhere.nix> wrote:
>>
>>
>>> As a result of this sub-thread, I rewatched the AlphaGo video of the
>>> event of 2016. It is well worth the 1 hr:29 min time. It is one of the
>>> most gripping videos I have watched in a long, long time. It could well
>>> be dubbed as "the beginning of the end of human dominance in game play".
>>> And that's just the start.
>>>
>>> https://www.youtube.com/watch?v=WXuK6gekU1Y
>>
>>
>> I never saw this before, but I'm watching now. Very good.
>
>As an ex-tournament chess player yourself, I'd be very interested in
>your take on 'intuition', as a result of watching this video.

I'm not good enough at Go to have a feeling for intuition in it. My
guess is that it's significant for most good players, but that's just
a guess.

>
>Lee Sedol is no flake. As the world champion of Go (for the last 10
>years), he clearly puts a lot of faith in his intuition and its role in
>his playing of the game.
>
>Then comes move #37 (game #2, around 49mins into the video). Poof!
>
>It was dispiriting to witness the realisation of 8 million Koreans (as
>well as many other human around the world) that the machine was better
>at Go than the best human.

Rich Ulrich <rich.ulrich@comcast.net> writes:
> I just now recognized that 'chastise' is an unusual word in that
> we Americans do not use -ize for the -ise. I typed -ize and it
> looked so wrong.
> https://books.google.com/ngrams/graph?content=chastize%2Fchastise&year_start=1800&year_end=2019&corpus=en-US-2019&smoothing=0
>
> for AmE, -ize hit its peak around 1980, about 7%, but fell again
> to below 1%.

You were unwize to expect it to rize any higher.

Phil
--
We are no longer hunters and nomads. No longer awed and frightened, as we have
gained some understanding of the world in which we live. As such, we can cast
aside childish remnants from the dawn of our civilization.
-- NotSanguine on SoylentNews, after Eugen Weber in /The Western Tradition/

Silvano <Silvano@noncisonopernessuno.it> writes:
> occam hat am 23.04.2024 um 08:13 geschrieben:
>
>> True. But look at the facts. The 'rules' of the game of Go are so simple
>> - and so few - that they are not worthy of comparison with geometry:
>> 1- You can put a piece anywhere on the board which is not already
>> occupied by another piece; 2- black and white pieces alternate in play.
>> The rest is just computing the score at any given instant.
>
> Your rules are not enough to play Go. You must add at least:
> 3 - Pieces on the board never move. They can only be taken off the board
> when they have no adjacent free point.

Unnecessary. Nobody said moving a piece is a move, so moving a piece
isn't a move.

The rule you're actually missing is no air => removal.

> 4 - You cannot put a piece where it has no adjacent free point. It would
> go off the board instantly because of rule 3.

Unnecessary. Play and instant removal is 100% equivalent to a pass,
which is a legal move.

> 5 - You ignore 4 if your piece also takes the last free point of one or
> more pieces of your opponent. Instead, that piece or those pieces of
> your opponent are taken off the board.

Yup, but the removal should bave been the first thing you listed.

> 6 - You cannot put a piece where it reproduces a former whole board
> configuration.

Yup, which covers ko-fights as the most trivial case.

Phil
--
We are no longer hunters and nomads. No longer awed and frightened, as we have
gained some understanding of the world in which we live. As such, we can cast
aside childish remnants from the dawn of our civilization.
-- NotSanguine on SoylentNews, after Eugen Weber in /The Western Tradition/

Ken Blake <Ken@invalid.news.com> wrote:

> On Tue, 23 Apr 2024 14:16:18 +0200, occam <occam@nowhere.nix> wrote:
>
>
> >As a result of this sub-thread, I rewatched the AlphaGo video of the
> >event of 2016. It is well worth the 1 hr:29 min time. It is one of the
> >most gripping videos I have watched in a long, long time. It could well
> >be dubbed as "the beginning of the end of human dominance in game play".
> >And that's just the start.
> >
> > <https://www.youtube.com/watch?v=WXuK6gekU1Y>
> >
>
> I never saw this before, but I'm watching now. Very good.
>

I've just viewed it for the first time and found it spellbinding. I was
particularly impressed by the way that the vanquished human player
rose above his defeat - inspirational.

--
^Ï^. Sn!pe, PA, FIBS - Professional Crastinator

My pet rock Gordon just is.