Learning by example?
Let me continue with the previous position, Yet one move further, after 1,Nxb7 Qxb7
 |
| White to move |
3r3k/nq4pp/p4p2/1p2p3/4P3/2P2N1P/PnB1QPP1/R5K1 w - - 0 2
I can imagine that I reach this position by logical reasoning and see it before my minds eye. How should I proceed? Somehow I have the feeling that producing more word salad from this point onwards is not very effective.
How does System 1 learn? Is it enough to describe what the position is about? Or do I need to invent the description all by myself? In other words, must I approach this from the study room or from the training room?
I'm of course addicted to the study room. But the document that I produced there (see the previous post) seems to be sufficient as guideline for the description of this position.
Robert gave a lot of examples for karate and stuff that seem to imply that there comes a moment that System 1 must just follow the instruction (while System 2 must shut up for a moment). This seems such moment.
Your distinction between the study room and the training room is very useful but not often made or kept distinct.
ReplyDeleteIMHO
The study room should be focused toward meta-learning—learning what learning methods work and what learning methods do not work. It targeted on HOW to learn, NOT WHAT to learn.
The training room reverses the emphasis: in the training room, the emphasis should be on the WHAT (the set of examples that provides the most succinct learning experience in minimal training time); the HOW should have already been determined in the study room.
The playing room is the "final exam" of the training room's efficacy.
PART I:
ReplyDeleteTemposchlucker asks several important questions:
(1) How does System 1 learn?
(2) Is it enough to describe what the position is about?
(3) Or do I need to invent the description all by myself?
(4) In other words, must I approach this from the study room or from the training room?
(1) How does System 1 learn? System 1 learns best through experience.
I could write (and have written) a veritable smorgasbord of word salads trying to describe the kind of learning experience that System 1 needs to ingrain the appropriate pattern recognition into long-term memory. All the words in the world would still fail to convey the proper idea.
Let me revisit the notion of a Zen koan and Beginner’s Mind from a different perspective. Please note the emphasis I placed on Beginner’s; that is crucial to understanding how System 1 learns. Please note that I am NOT a Zen religious ‘guru’ possessing some esoteric “secret;” I have nothing more than my own personal experience to guide this exposition. And yes, unfortunately it’s another word salad. I don’t have a teleportation mechanism to be able to demonstrate what I mean by experience.
I want you to conduct an experiment in three parts. Please don’t jump from one part to the next using System 2; if you do, you will miss the significance of “SEEing” with Beginner’s Mind.
Start with this most famous and simplest Zen koan:
What is the sound of one hand clapping?
Part 1: Surface analysis using System 2
Here is a rewrite of the koan, seen from the initial System 2 perspective:
What is the sound of ONE HAND CLAPPING?
Your System 2 will contemplate this and instantly arrive at this conclusion: ONE HAND and CLAPPING when combined is illogical, making no sense whatsoever when seen by System 2. Feel the strain this places on your mind. You have an immediate negative reaction to it: it is logically IMPOSSIBLE! As a result, you dismiss it without further consideration.
Part 2: Essential analysis, still using System 2
Here is another rewrite of the koan, still seen from the System 2 perspective:
What is the SOUND of one hand clapping?
If you persist (still using System 2) in studying with a slightly different perspective (in spite of the illogical nature of the statement as first seen), you MAY arrive at a possible logical solution to the conundrum. By shifting focus from the obviously illogical ONE HAND CLAPPING to the possible SOUND, you can understand the logical answer. You verbalize the question along with the logical answer:
The SOUND of one hand clapping is SILENCE.
The logical mind triumphs!
Or does it? The Zen adept replies that this verbalized answer is NOT the answer to the question. We are back to the original question with a ready answer that is—WRONG.
How can THAT be possible?!?
PART II:
ReplyDeletePart 3: Experiencing the answer, using System 1 only
What is the sound of one hand clapping?
EXPERIENCE the answer. This is the gestalt “AHA!” experience.
If the answer is verbalized as ”SILENCE”, the answer is NOT actual silence!! The correct answer is experiential: in response to the question, answer with the experience of silence itself, not with verbalization. The question is verbalized; the answer is EXPERIENCED by System 1. The correct response (experienced by System 1) is [ ] [without the brackets indicating some word response by System 2].
THIS feeling is what EXPERIENCE is all about; it is NOT a word salad describing the experience, which will ALWAYS fall short in capturing the essence of the experience. And that kind of experience is how System 1 learns. This is also the “it” that operates from System 1 when doing a martial art, drawing, playing a musical instrument, playing masterful chess, in short, just about everything that demonstrates expertise.
(2) Is it enough to describe what the position is about?
NO, it is NOT enough to use words to describe what the position is about. The position (and its corresponding salient cues and patterns) must be EXPERIENCED in the same way that SILENCE is experienced.
(3) Or do I need to invent the description all by myself?
NO, you do NOT need to invent a description using words. The words are an attempt to communicate using System 2. Get rid of the words!
Remember: a picture EXPERIENCED is worth a thousand, nay, ten thousand words!
(4) In other words, must I approach this from the study room or from the training room?
This is progress. The study room is useful for investigating HOW System 1 learns best; it focuses on the learning method, not the content to be learned. The training room applies the learning method, and is focused on WHAT is presented for System 1 to absorb. As much as possible, ELIMINATE THE WORDS! Lather, rinse, repeat.
The difficulty has always been separating the HOW from the WHAT because of our (almost) total reliance on System 2 for learning both method and content.
You can't skip part 1 and 2. System 2 is a friendly guy who is more than willing to cooperate. He is willing to shut up, but he needs his pacifier: understanding the why. Without an explanation he gets cranky and will sabotage the outcome.
ReplyDeleteAs this post shows, System 2 needs to acknowledge its shortcomings before he is willing to shut up.
ReplyDeleteSystem 1 and System 2 can work in perfect harmony. They both look over the shoulder of attention and are educated. That is the how logical thinking and patterns become connected. Just like a story becomes connected to the sight words.
ReplyDeleteThe art is to guide the attention.
In SEP 2017 I made some comments on the blog post Promotion example [August 23, 2017] based on Dr. Betty Edwards’ book Drawing on the Right Side of the Brain: A course in enhancing creativity and artistic confidence. Those comments address the general roles of System 1 [right side of the brain] and System 2 [left side of the brain] and how they interact in a chess context.
ReplyDeleteOne of the most important insights I gained from Dr. Edwards is that our educational system is almost totally oriented toward teaching/learning using System 2. Why? Because it is verbal based and therefore language-oriented, which facilitates transfer from teacher to student in the formal classroom. System 1 is non-verbal and thus does not communicate using written language. As a result, it becomes very difficult (if not impossible) to “listen” when System 1 tries to tell us something.
Her method for getting System 2 to “shut up” is by using an upside-down picture as the first drawing exercise. That confuses System 2 sufficiently for it to recognize that it’s ILLOGICAL. [Note the similarity to the “One Hand Clapping koan.] It is only by causing System 2 to “drop out” (STOP WITH THE WORD SALADS!) that System 1 can perform its “magic” of SEEing things as they are with nothing added, nothing removed AND communicating that “sight” through patterns to the consciousness.
I don't belief in System 2 as the enemy of System 1. It just needs a pacifier. A lot of time is wasted with fighting against System 2.
ReplyDeleteIf I gave the impression that I (or Dr. Edwards) consider System 2 to be an "enemy" of System 1, I apologize. Dr. Edwards emphasized that the two system are complementary, not antagonistic. The idea behind getting System 2 to be quiet (at least some of the time) is that we so often rely totally on it, rather than using a more balanced approach of both systems.
ReplyDeleteThe previous post gives a series of scenarios. The scenarios are executed by executing subtasks. Scenarios and subtasks have a 1:n relationship.
ReplyDeleteSome subtasks are general. They can be used in multiple scenarios. Like the task removal of the defender or add an attacker.
I have a set of 1430 combinations. How should we work with such set in the training room? First of all, the tasks of which the combination exists must be identified. Feel free to use Stockfish for identification to avoid wasting time.
Once the tasks are clear, it is time to study the threats that make the combination work.
CMCT puts us on the wrong foot. It is all about threats used to accomplish a task. A check is a threat, a mate in 1 threat is a threat.
A capture is a way to accomplish a subtask. It is a postponement move. A 1:1 move. One capture, answered by one take back. The combination is postponed for one move. In order to make a capture more forcing, you can combine it with a threat.
Threats make the combination tick. Understanding how that works is the task for the training room.
The list with scenarios can be enhanced in the future with new scenarios. Like endgame scenarios, or positional scenarios.
ReplyDeleteThe task is more important than the capture. Unless the task is to gain wood. Which is usually at the end of the combination.
ReplyDeleteSo you must focus on "what task does this capture fulfill?" and not on the capture itself.
ReplyDeleteFrom the study room to the training room: an example
ReplyDeleteWhat are the components of a back rank mate pattern?
There is more to the answer than just the final location of the king at the end of the mate on the back rank. DUH!
A template (a “chunk” plus slots to store related variable information from position to position) would be useful. This provides a pattern framework which has the capability to encompass more than just 1 specific concrete position.
Use the following artificial position as the basis for the template.
FEN: 6k1/ppr2ppp/8/8/8/8/PP3PPP/3R2K1 w - - 0 1
Fixed “chunk” characteristics:
Name of this template (It may have no name)
Goal: Player to move can checkmate
King confined behind an unbroken pawn shield (back rank)
Vulnerability of the back rank to an enemy heavy piece (queen or rook) attack
Template slots (for variable information):
Player to move
Attacking piece type (queen or rook)
Attacking piece location
Line of attack of the attacking piece
Focal point (square on the back rank on which checkmate can be delivered)
I’m not trying to be exhaustive, just illustrative.
Fixed “chunk” characteristics:
Mnemonic name of this template = “back rank mate”
White to move can checkmate
King confined behind an unbroken pawn shield (back rank)
Because there is no Black heavy piece on the back rank, it is vulnerable.
From the example position, we can fill the template slots with specific values.
Template slots (for variable information):
White is on the move
Attacking piece = White rook on d1
Attacking piece location = d1
Line of attack of the attacking piece = d1→d8
Focal point = d8
We now have a viable pattern to commit to long-term memory in the form of a “back rank mate” template. This is the basis for “fiddling around” to see what changes can be made (and stored in the slots of the template) without changing its essence. The number of “fiddles” can be as many as there are possible alternative values placed into the slots without changing the essence of the pattern—with as few as required to “cement” the template into long-term memory. The order in which the “fiddling around” changes occur will NOT be the same for everyone; that is unimportant.
1st “fiddling” alteration:
Attacking piece = White queen.
Check: The basic pattern still holds true.
2nd “fiddling” alteration:
Attacking piece location = e1 [also d2,d3,d4,d5,d6,d7,e2,e3,e4,e5,e6,e7 if a rook; if a queen, additionally a4,b5,c6,d7,h4,g5,f6,e7]
Check: The basic pattern still holds true.
3rd “fiddling” alteration:
Line of attack: a4→e8 [also d1→d8, e1→e8, h4→d8]
4th “fiddling” alteration:
Change the focal point = e8
Without changing anything other than the attacking piece type (queen or rook) and attacking piece location, we have 36 different “patterns” for concrete positions, stored within one template in long-term memory. Adding 4 different lines of attack to the previous two slots yields 144 “patterns.” Adding two different focal points yields 288 “patterns.” [Maybe this arithmetical process is how someone estimated that Grandmasters “recognize” 300,000 “patterns.”]
This is just one basic template. It does not begin to scratch the surface for possible variations.
Here’s an interesting question: Do we have to actually see every single concrete alteration in order to cement the generic template into long-term memory? I certainly hope NOT!!
Here’s an interesting question: Do we have to actually see every single concrete alteration in order to cement the generic template into long-term memory? I certainly hope NOT!!
DeleteNo, you only need 1 logical conceptual pattern. If you strongly insist on a personal approach, you might want to adopt the 7 circles of stupidity.
PART I:
ReplyDeleteNow let’s make a more significant alteration—adding a queen (for both sides) and changing the line of attack. In short, we now have a preparatory move that must prepare the attacker(s) and the lines of attack in order to set up for the actual checkmate. This will require a modification to our template to “capture” this new precondition.
FEN: r5k1/p1Q2ppp/1p6/q7/8/8/PPP3PP/5RK1 w - - 0 1
The position still meets two of the fixed criteria for a pattern match:
King confined behind an unbroken pawn shield (back rank)
Vulnerability of the back rank to an enemy heavy piece (queen or rook) attack
The problem is that the most “obvious” focal point (according to our original definition) is that a one-mover is no longer available: 1. Qd8+ Rxd8 removes the possibility of checkmate.
What does this new position tells us? There is now a defender of the back rank. The preparatory solution is to remove the defender. Apparently, that is not possible because the defender is tucked away safely on a8.
We should look for a way to get to the back rank that does not allow the BRa8 to remain safe while defending the back rank. Concurrently, we need to find a new line of attack that will maintain the back rank weakness and allow a heavy piece to enter the back rank with checkmate. Note that the Black queen cannot provide any assistance; it is out of play for defensive purposes.
Is there a weak square that is insufficiently guarded? Yes, the f7-square has two attackers to one defender [2:1]. The f7-square is adjacent to the focal point. Regardless of which White piece captures on f7, the Black king cannot capture because the attacking piece is defended. If it is attacked from f7, it must abandon its defense of f7 AND f8 (which has magically become the focal point for a back rank mate). Only the White queen can attack f8 AND BKg8 simultaneously from f7, forcing the Black king to move to h8 (which maintains the criteria of a pawn shield blocking its escape from the back rank). So, it is logical to capture on f7 with the queen, not the rook.
After 1. Qf7+ Kh8, the focal point f8 is now available (via exchange on f8) to put the WRf1 on f8 with checkmate. 2. Qf8+ Rxf8 3. Rxf8#.
I’m intentionally sliding over a lot of details in the interest of reducing the amount of word salad to be digested. THE WORD SALAD IS NOT PART OF THE TEMPLATE!
I miswrote the first move: 1.Qxf7+ illustrates that the capture on f7 is not intended as gain of material (that is incidental); it is the creation of a threat against the Black king which must respond by moving away.
DeleteToo much word salad gives indigestion.
PART II:
ReplyDeleteFor the preparatory stage, we have to threaten the Black king (to force it away from the defense of the focal point; initially, that focal point was B.A.D. [2:2]). The weaknesses of f7 and f8 attract attacking force, and allow this sequence to create checkmate.
Again, it is possible to “fiddle around” with the position, moving the Queen around, moving the defenders around, etc. I’m not even going to attempt to calculate the number of different positions that can lead to the same essence. Hopefully, one concrete example is sufficient to illustrate the possible modification(s) to the template that could occur.
It is vitally important to learn to “SEE” potential alterations predicated on an elementary essence. That greatly multiplies the number of different concrete positions that can be recognized by the underlying pattern.
I estimate that this new pattern template can be “burned” into long-term memory with fewer than 10 different examples.
It is through the same process that more and more complications can be added to an elementary template, expanding its scope, so that more and more positions will trigger pattern recognition of the essence and the corresponding preparatory moves.
The interesting AHA! occurs when two or more patterns have been ingrained into long-term memory and we recognize a more abstract general pattern that incorporates some or all of these multiple patterns. There is really no practical limit to how far this process can be taken. The ideal is to inculcate a sufficient number of different patterns so as to enable the recognition of the underlying essences of known patterns embedded in a completely “never-seen” position arises. The ingrained pattern templates should have sufficient elasticity to “stretch” to cover the new position, with a certain amount of mental effort to bridge the gaps. This is the epitome of SKILL.
At a much more abstract level, there is the following CHESS TEMPO categorization of a "family" of checkmates:
ReplyDeleteEdge Mate
◦ Back Rank Mate
◦ Edge Pin Mate
◦ Queen Cutoff Mate
◦ Side File Mate
◦ Smothered Mate
◦ Suffocation Mate
Edge Mate – General Category
Edge mates refer to back rank and side file mates and their variants such as the queen cutoff mate. Note that this should not be applied to every mate that happens on an edge of the board, it is specific to the conditions of the back rank mate etc such as being hemmed in by the king's own pawns and pieces.
Perhaps there might be some disagreement over including a variant like the "Queen Cutoff Mate" in the general category with the back rank mate. Go back to the elementary criteria (king behind a pawn shield with no luft) and you will find the commonality.
SLICING AND DICING THE CATEGORIES INTO WORD SALADS IS N-O-T THE POINT!! THE POINT IS TO "SEE" THE PATTERNS AND ASSOCIATED ACTIONS!
You are almost there. The scenarios are familiar enough. What we need to learn is to build the combination with threats. To couple the logic of System 2 to the threat-patterns of System 1.
Delete