AI making us lazy?

gennan · Post by **gennan** » Sat Oct 30, 2021 1:31 pm

While I'm at it: that black shape may also work nicely in situations like these:

Click Here To Show Diagram Code: [go]$$W $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . X . b . . . X . . | $$ | . . O , . O . . . . . . . . . . . . . | $$ | . . . . X O . . O . X . 1 . . X . . . | $$ | . . . . . . . . . . a . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ------------------------- $$[/go]

invades black's area, threatening to link up with a or escape around b.
Black may be at a loss how to prevent both threats with one move.

Click Here To Show Diagram Code: [go]$$W $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . X . 2 . . . X . . | $$ | . . O , . O . . . . . . . . . . . . . | $$ | . . . . X O . . O . X . 1 . . X . . . | $$ | . . . . . . . . . . a . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ------------------------- $$[/go]

But

defends against both threats. Even linking-up underneath with a is not working anymore:

Click Here To Show Diagram Code: [go]$$W $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . X . 2 . . . X . . | $$ | . . O , . O . . . a 7 9 0 . . . . . . | $$ | . . . . X O . . O 5 X 8 1 . . X . . . | $$ | . . . . . . . . . 4 3 6 . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ------------------------- $$[/go]

White needs to defend against black a now, so the

invasion is failing.

CDavis7M · Post by **CDavis7M** » Sat Oct 30, 2021 3:06 pm

gennan wrote: invades black's area, threatening to link up with a or escape around b.
Black may be at a loss how to prevent both threats with one move. But defends against both threats. Even linking-up underneath with a is not working anymore:

$$W
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . . X . 2 . . . X . . |
$$ | . . O , . O . . . a 7 9 0 . . . . . . |
$$ | . . . . X O . . O 5 X 8 1 . . X . . . |
$$ | . . . . . . . . . 4 3 6 . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ -------------------------
$$
Click Here To Show Diagram Code
[go]$$W $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . X . 2 . . . X . . | $$ | . . O , . O . . . a 7 9 0 . . . . . . | $$ | . . . . X O . . O 5 X 8 1 . . X . . . | $$ | . . . . . . . . . 4 3 6 . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ------------------------- $$[/go]
White needs to defend against black a now, so the invasion is failing.

What about

Click Here To Show Diagram Code: [go]$$W $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . X . X 2 . . X . . | $$ | . . O , . O . . . a 7 9 1 . . . . . . | $$ | . . . . X O . . O 5 X 8 O . . X . . . | $$ | . . . . . . . . . 4 3 6 . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ------------------------- $$[/go]

Of course, other things are possible.

Next post: Gennan's tips and tricks making me lazy?

kvasir · Post by **kvasir** » Sun Oct 31, 2021 4:05 am

Click Here To Show Diagram Code: [go]$$W $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . Y 4 2 . . . X . . | $$ | . . O , . O . . . . . . 3 . . . . . . | $$ | . . . . X O . . O . X . 1 . . X . . . | $$ | . . . . . . . . . . 5 6 . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ------------------------- $$[/go]

for

looks like a really bad exchange. I find it very hard to imagine any sequence that could work better with

+

than without these two moves.

I think it is a really good advise to learn how to play normal moves rather than study the exceptional. For one thing it is because you need some baseline to compare with. One also needs something to fallback on in time trouble and when confused. Black didn't have the marked stone in the original position, in this case

is not really normal and this position can quickly become very complicated if white tries to justify the invasion.

CDavis7M · Post by **CDavis7M** » Sun Oct 31, 2021 1:12 pm

kvasir wrote:
$$W
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . O . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . X . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . . Y 4 2 . . . X . . |
$$ | . . O , . O . . . . . . 3 . . . . . . |
$$ | . . . . X O . . O . X . 1 . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ -------------------------
$$
Click Here To Show Diagram Code
[go]$$W $$ | . . . , . . . . . , . . . . . , . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . O . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . X . . . . . . . . . . . . . . . . | $$ | . . . O . . . . . . Y 4 2 . . . X . . | $$ | . . O , . O . . . . . . 3 . . . . . . | $$ | . . . . X O . . O . X . 1 . . X . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ | . . . . . . . . . . . . . . . . . . . | $$ ------------------------- $$[/go]
for looks like a really bad exchange. I find it very hard to imagine any sequence that could work better with + than without these two moves.

It was cheeky, hence the

above. I did look with Leela already (making me less motivated?) and playing

is not "good" but it's not really "bad" either (by my standards). Also, Leela did suggest that black extend right rather than connect at

.

kvasir wrote:I think it is a really good advise to learn how to play normal moves rather than study the exceptional.

I agree is the goal is improvement. But if the goal is entertainment, then it is really good advise to review the exceptional. I didn't actually study any of this. I reviewed it for entertainment. And this game's position was just the first thing that I recalled when seeing that shape. Next time, my first recall might be what Gennan said.

This sort of goes back to the original question in the first post: "Ryan Li’s recommendation is to generally use AI less, because knowing what the correct move is in a situation does not help furthering your understanding of the game." For me, I'll use Ghost Intelligence less, but only after Halloween.

jlt · Post by **jlt** » Sun Oct 31, 2021 1:26 pm

Ryan Li’s recommendation is to generally use AI less, because knowing what the correct move is in a situation does not help furthering your understanding of the game.

The same could be said about josekis. Maybe we would benefit by deviating more often from josekis.

Knotwilg · Post by **Knotwilg** » Tue Nov 02, 2021 3:05 am

Click Here To Show Diagram Code: [go]$$W Invasion, interception $$ ------------------ $$ | . . . . . . . . . $$ | . . 7 5 6 a . . . $$ | . . 1 2 8 O . B . $$ | . . 3 X . . . . . $$ | . . . 4 . . . . . $$ | . . 9 . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . .[/go]

Before AI we played this joseki, up to pro level.

Click Here To Show Diagram Code: [go]$$W Invasion, interception $$ ------------------ $$ | . . . . . . . . . $$ | . . 7 5 6 8 . . . $$ | . . 1 2 a O . B . $$ | . . 3 X . . . . . $$ | . . d 4 . . . . . $$ | . . 9 b . . . . . $$ | . . . . c . . . . $$ | . . . . . . . . . $$ | . . . . . . . . .[/go]

AI played this move and so we have investigated why this would be better. The general understanding is that the aji of the cut at A here is less cumbersome than the aji of White dropping down at A in the previous diagram.

AI has showed us the right move and through studying variations we have improved our understanding of the game.

We can of course agree on the tautology "using AI in a lazy way will make us lazy".

John Fairbairn · Post by **John Fairbairn** » Tue Nov 02, 2021 3:32 am

AI played this move and so we have investigated why this would be better. The general understanding is that the aji of the cut at A here is less cumbersome than the aji of White dropping down at A in the previous diagram.

I beg to differ, not with you but with the "general understanders".

I'm totally prepared to accept the crawl as better if for no other reason that, when a beginner, I thought the crawl was the obvious move and couldn't ever completely understand why the connection was favoured.

The bit I have difficulty in accepting is the alleged modern reasoning. "Less cumbersome" is rather abstract and meaningless. It is insufficient. It is a "lazy" explanation. Which means AI has so far taught us zilch about this position (except to play A not B on the monkey see, monkey do principle). I think it is reasonable for us to expect a more concrete explanation.

FWIW I think a passable concrete explanation can be that the crawl effectively creates a Go Seigen group on the side if (here) White cuts. I came to this conclusion after seeing someone (I think it was Sumire) willingly accept the cut and give up the three cut-off stones. We already know from Go Seigen that offering to give up three stones is very "cumbersome" for the opponent. It seems to me that bots and the best modern players seem increasingly willing to give up even bigger sacrifices. If I'm right about that, three stones must seem even more trivial now than in Kamakura days. And going on from there, AI is therefore not teaching us anything about the tactics of one joseki at all really - it is teaching us about whole-game strategy.

So,

We can of course agree on the tautology "using AI in a lazy way will make us lazy".

Yes. Absolutely.

Knotwilg · Post by **Knotwilg** » Tue Nov 02, 2021 4:05 am

John Fairbairn wrote: FWIW I think a passable concrete explanation can be that the crawl effectively creates a Go Seigen group on the side if (here) White cuts.

That's at least a conceptually richer explanation than merely describing the aji as more or less cumbersome. Let's say that was just me describing the advance in understanding somewhat lazily.

John Fairbairn wrote: And going on from there, AI is therefore not teaching us anything about the tactics of one joseki at all really - it is teaching us about whole-game strategy.

I agree. In the Suzuki thread, if I was able to learn from that position at all by using KataGo, it's that again and again, AI is looking for sente because with every local move, temperature may suddenly drop, or otherwise stated, chances for temperature to drop are higher than to rise, so you should always be on the lookout for sente. That matters more than the actual local result. My words/understanding.

We can of course agree on the tautology "using AI in a lazy way will make us lazy".
Yes. Absolutely.

And the same applies to human (pro) advice. The main difference with human advice is the reasoning behind the move. I'm not convinced that getting the best move + reasoning would make us less lazy than merely getting the best move.

John Fairbairn · Post by **John Fairbairn** » Tue Nov 02, 2021 5:04 am

In the Suzuki thread, if I was able to learn from that position at all by using KataGo, it's that again and again, AI is looking for sente because with every local move, temperature may suddenly drop, or otherwise stated, chances for temperature to drop are higher than to rise, so you should always be on the lookout for sente. That matters more than the actual local result.

I don't use AI or look at AI-AI games, though I do look at lots of games by pros who are following AI. That probably gives me a skewed version of what is going on in the AI 'mind', though I do think it's easier for us amateurs to learn from such pros than from AI direct.

However, reality is that my go environment is also heavily affected by western trains of thought, notably what I read here. Yet I often feel a strong disconnect between that and what Japanese pros say. Sente is one of the main areas of polarisation.

I imagine western go players include many chess or ex-chess players and so we are influenced by chess thinking and chess terms. As such, we take very much to heart Bronstein's famous dictum "The most powerful weapon in chess is to have the next move.” Add to that the fact that westerners adore the lilt of the word sente, we end up using it far, far more than you ever see it in Japanese (or Chinese) texts. In fact, we use it as a kind of portmanteau word, blending two or three meanings that are usually differentiated in the oriental languages. That, I think, makes it difficult for us to have conversations in terms such as your quote above. (That's leaving aside losing people like me once you start talking about temperature

)

My instinct ("trained" by seeing other relevant and very common words such as ichidanraku in Japanese) is to make a firm distinction between having the next move, having a forcing move and having the initiative. Having the next move means at most controlling the next move. Having the initiative means controlling the game. Controlling the game means you often don't need to worry about having next move.

The concept is not at all strange to us - we have just chosen, perhaps for historical reasons as well as chess influence, to ignore it in go.

To give an example from western sport, NFL football shows it most starkly. One side is literally given the ball to make the next move. After they have their turn, the ball is given to the other side. They then have the next move. It is such a stark turnaround that different squads of players are brought on (defense and offense) in each phase. In the early part of the game, each side is trying to score or to prevent the other side from scoring. But as the game proceeds, one side may pull ahead with touchdowns and so have a lead. They then have the initiative in the game.

Imagine a situation where the Patriots are leading 45-3 with two minutes left on the clock. The Jets have the ball. They have the next move. The Patriots have the initiative. They can just let the clock run down.

But that matters a lot more than being a more description. Imagine a similar situation but with the score at 21-14 in favour of the Patriots. The Jets have the ball and so can still tie the game with a converted touchdown, but they have no timeouts. The Patriots have the initiative (the lead), but more than that they even control the next move - they don't have it but they control it. That's because the Jets, with no timeouts to stop the clock, have to choose plays that end up running out of bounds (that stops the clock), and they can't make do with a field goal. They are effectively playing under a handicap. They control the ball but not the game.

It is my belief that AI bots (or at least pros who follow bots) play to have the initiative, and don't really care who has next move. They want sente = control, not sente = next move. Control in go is probably definable not as who has next move, but where the next move will be played, by either side. Until we unravel that distinction in English, I don't think we will make much progress in fathoming what AI is up to, but I'm even more certain that we won't be fully in synch with what oriental pros are saying about the topic.

kvasir · Post by **kvasir** » Tue Nov 02, 2021 5:14 am

Click Here To Show Diagram Code: [go]$$W $$ ------------------ $$ | . . . . . . . . . $$ | . . O O X a . . . $$ | . . O X X O . X . $$ | . . O X . . . . . $$ | . . . X . b . . . $$ | . . O . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . .[/go]

My understanding is that the computer dislikes this positions because white can play a or b as forcing moves and there is little downside to this even when white doesn't follow up with anything. Both of these moves were explored by human players for years but at least a was considered (or taught) to be a negative and b seems to have been rarer.

Click Here To Show Diagram Code: [go]$$W $$ ------------------ $$ | . . . . . . . . . $$ | . . O O X X . . . $$ | . . O X . O . X . $$ | . . O X . . . . . $$ | . . . X . 1 . 2 . $$ | . . O . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . . $$ | . . . . . . . . .[/go]

For comparison in this case if

then

because it is not so easy to cut.

The difference between the connection and the connection under is rather small when you check with the computer and I have seen the computer recommend the connection in situations when the forcing moves don't make sense. What is more surprising is that there have been people playing the connect under move for years but not so much on the top pro level. Even more surprising is that the decision often doesn't seem to affect the next many moves much. Is this ultimately how AI makes us lazy? We seem to chance one move and magically have better percentage without really changing how we continue from that point.

mumps · Post by **mumps** » Tue Nov 02, 2021 5:34 am

John Fairbairn wrote: It is my belief that AI bots (or at least pros who follow bots) play to have the initiative, and don't really care who has next move. They want sente = control, not sente = next move. Control in go is probably definable not as who has next move, but where the next move will be played, by either side. Until we unravel that distinction in English, I don't think we will make much progress in fathoming what AI is up to, but I'm even more certain that we won't be fully in synch with what oriental pros are saying about the topic.

I think this is a common misinterpretation of AI bots. They have no concept of initiative or any other human concept of Go. They are merely playing to get one more point than their opponent and are evaluating the position solely in these numeric terms.

Sadly, this also means that they cannot explain why they make a move in a way that we can understand and use.

Perhaps all we can do is the same as Mark: play through (AI) games without attempting to 'understand' what's going on and just trying to feel the flow?

John Fairbairn · Post by **John Fairbairn** » Tue Nov 02, 2021 6:55 am

I think this is a common misinterpretation of AI bots. They have no concept of initiative or any other human concept of Go. They are merely playing to get one more point than their opponent and are evaluating the position solely in these numeric terms.

Thanks for the salutary reminder, Jon. I think we do know this* but the urge to anthropomorphise overtakes us and we do end up forgetting. Does my dog really love me or has he just worked out how to get food and shelter? We prefer to believe the former.

But does it not follow from your, dare I say it, cynical view that even trying to follow the AI flow is also bound to fail? What I am getting at is that there may be no flow to follow. The bots' evaluations are always in the moment. There is no detectable (or consistent) strategy behind them. That leaves us humans like fish out of water. Getting back into the pool with humans pros may be the best option, no? Just use AI for ideas and motivation, not as gospel.

*With the caveat that there may a version of katago that tries to optimise the points difference????

PS I've just met the second person in my life I know who plays croquet!

Knotwilg · Post by **Knotwilg** » Tue Nov 02, 2021 9:30 am

mumps wrote: I think this is a common misinterpretation of AI bots. They have no concept of initiative or any other human concept of Go. They are merely playing to get one more point than their opponent and are evaluating the position solely in these numeric terms.

The universe doesn't know it follows Newton's, pardon, Einstein's laws of mechanics, or Maxwell's equations in the electromagnetic field, or Schrodinger's or Dirac's equations. The universe doesn't know the speed of light, Planck's constant etc etc. The universe simply IS.

All of these are human concepts which help us understand the physical and cosmological reality. The fact the universe is unaware of the concepts according to which it works should not discourage us investigating and conceptualizing.

The same happens when we try to understand Go. AI acts as a magnifying glass, or an idiot savant, or whichever analogy you like, for us to understand Go better. We do so collectively and need language, hence terms, to discuss and articulate.

So when we say "AI favor sente" we don't really think AI consciously tries grabbing sente, althought I wouldn't be so sure about what's really in that neural network when left to its own resources and what isn't.

Now back to John, who says AI wants "control" rather than the next move and finds Western go players think of having the next move more than Eastern (pro) players (or AI) and this may be induced by our chess intuition.

I can't speak for the rest of you but I doubt my brief and youthly exposure to Chess still dominates those 25+ years of intensive Go playing and studying. When I talk about sente (maybe wrongly so, but I try to stick to the more common terms, even if they are Japanese) I really mean "the right (and duty) to play elsewhere because the local situation is not so big/urgent anymore (i.e. temperature has dropped)". That is a sharper notion than mere "control". Of course you want control. It probably means you are leading. But the obnoxious striving for the right to play elsewhere is what has been reinforced and encouraged by studying AI and what I find a useful sharpening of my approach.

gowan · Post by **gowan** » Tue Nov 02, 2021 2:33 pm

John Fairbairn wrote:
I think this is a common misinterpretation of AI bots. They have no concept of initiative or any other human concept of Go. They are merely playing to get one more point than their opponent and are evaluating the position solely in these numeric terms.
Thanks for the salutary reminder, Jon. I think we do know this* but the urge to anthropomorphise overtakes us and we do end up forgetting. Does my dog really love me or has he just worked out how to get food and shelter? We prefer to believe the former.

But does it not follow from your, dare I say it, cynical view that even trying to follow the AI flow is also bound to fail? What I am getting at is that there may be no flow to follow. The bots' evaluations are always in the moment. There is no detectable (or consistent) strategy behind them. That leaves us humans like fish out of water. Getting back into the pool with humans pros may be the best option, no? Just use AI for ideas and motivation, not as gospel.

*With the caveat that there may a version of katago that tries to optimise the points difference????

PS I've just met the second person in my life I know who plays croquet!

My emphasis above. I agree with the approach to using AI just as a pointer to where we might have made a mistake and, in addition, where a better move might be. There is no "flow" in AI play because Ai has no grasp of strategy on any large scale. I don't know much (anything?) about how AI programs choose moves but if there is any strategy it must lie in the heuristics (algorithm?) the AI uses to choose what moves to try for playouts and how many playouts to use. We use strategy as a compass in determining where to play and our feeble ability to playout (read) is how we decide which particular move to make. Our pathetic ability to read requires that we have higher level concepts and strategy to compensate.

Kirby · Post by **Kirby** » Tue Nov 02, 2021 5:01 pm

If you're in a room with a human pro go player, and you don't share a common language to communicate, is it all that different than reviewing with computer AI?

In both cases, the only explanation you can get are the sequences they recommend. No explanation can be given since there is no shared language but the moves themselves.

Still useful, in my opinion.

Life In 19x19

AI making us lazy?

Re: AI making us lazy?

Re: AI making us lazy?

Re: AI making us lazy?

Re: AI making us lazy?

Re: AI making us lazy?

Re: AI making us lazy?

Re: AI making us lazy?

Re: AI making us lazy?

Re: AI making us lazy?

Re: AI making us lazy?

Re: AI making us lazy?

Re: AI making us lazy?

Re: AI making us lazy?

Re: AI making us lazy?

Re: AI making us lazy?