countingtls

I think I have a plan of borrowing my printing head design ideas to merge multiple cell lines into a truly massive swarm.

The idea is simple, instead of cramp them all into one, we create each cell line for their individual functions

• harvesters/worker drones that would go out to harvest plants (swimming out in all directions to find the plant harvesting area of light, and when energy storage lipocytes are sufficient, grow a light aversion senseocyte to return to the home base in the dark side). I have some "pheromone" ideas, but they are not as robust and simple as this
• Patrols/soldier drones, those follow the worker's color/signal. Which might be a way of "depositing" pheromone color cells. Since they wouldn't feed themselves, so they will eventually die, and if we time their life span just right, we can make them only last a distance less than from the home base to food source. Their corpses will leave a trail of food when them pop, and many allow the worker to follow, and not just blindly search the light source.
• Nanny drones that use secrocyte to either provide the signals to bring egg in the form of stemocyte into the next larva phase, and would devour the harvester's lipocytes for energy, and even collect and group the leftover lipocytes connected with keratinocyte near the queen (attracted to both the lipocyte color and the egg colors)
• Caretakers with secrocyte that can produce protease to consume the "protected lipocytes food storage" and sacrifice themselves or burp the food to the queens (or this can be done by a variation of nannies that act as living food storage themselves, by converting keratin protected lipocyte into direct energies where they can be glued to the queen)
• And finally the queen itself, essentially a variation of my print head design that glue the starting stemocytes to the "birth channel" which will create each drone caste (instead of printing a string). The senseocytes can use colors as the activation, instead of S1 to S4 signals, hence the queen's "control" can be done behind an enclosed "brain" (which again can be a different cell lines, so we have even more cell modes to work with). And the larva growth pad can even be the extension of the birth channel just outside the funnel, and attract larva, with constant burping of food particle. Or the larva can also swim to and be glued to the growth pad, where another version of the caretaker drones "provide" the food.

In terms of drones, I think beside the pheromone mechanism and the caretaker mechanism I haven't fully resolved, the others are definitely doable, or been done before.

The most difficult part of this, might be the brain state control. Effectively, we need a control and detect mechanism for producing different number of internal "color state cells". The number of drones in each caste depends on a lot of factors, obviously when the food lipocytes are low, we need to produce more harvesters (or when the queen haven't "seen" caretaker drones near its feeding heads, more caretakers and harvesters will be "printed". Or sensocytes near the growth bed that detect the nanny dronee, and when they died or leave the area too far, indicating too few nannies to print. The patrol drones are harder to control, but not necessary needed if we don't use them, or just print a certain ratio of them related to the harvesters. The issue though is that these detection areas might be quite far from the internal brain color state area, and how to relay them effectively. Or do we need some internal "messenger drones" that travel between sensing areas and the brain receptor?

Another final question is that do we have enough cell count to scale up to this level. Individual parts and drones definitely would work, but I think around 10000 cells, the lag would start to get very troublesome. Each drone might be about 10 cells each, and the energy required to travel on a large substrate (maybe 5mm? even 6mm?) for room to set up such complex queen (and the nest) will be enormous. Harvesters need to be very efficient in travelling and transporting, and there will be a limit for how many cells the queen can sustain. (but in the early development we might have to compromise, and glue some photocytes and turn up the light for energy, just to demonstrate that this could work in principle)--

Besides this being a news for Ichiriki Ryo now had a lead of 3 to 1 against Iyama Yuta (just 1 more win to get the Honinbo title), the scoring of this game also showed a quite rare situation, when they played more than 361 moves (total 364 moves for this game). One side completely filled up during manual scoring, and had to count stones off the board.

In the case of this game, it was filled to one side with 4 space left on the board for black, and 3 extra white stones off the board, hence 4 vs -3, so black win by 0.5 in the end { 4 - (-3) - 6.5 komi}. (If you found odd that the total count are 360 stones, but there are 364 moves, it is because they run out of stones, so they had to exchange two prisoners during the final dame filling phase in order to finish the game)

The sgf record can be found here http://eidogo.com/#Pbj2YrFZ:0,364

BTW, this is also a game where the Japanese rule and the Chinese rule give different results. With 6.5 komi, the Japanese rule gives B+0.5, but with 7.5 komi in the Chinese rule, it is W+0.5 (or W+1/4 子 with Chinese half counting of total 184 black area => {184 - 3.75 - 180.5 = -0.25}). Also notice, this is not solely because the Chinese rule has more komi, but more to do with white plays the last move. In games where black plays the last move, the result might be the same (cases with seki and how the last ko is filled can raise more tricky situations and differences).

This is a game played on June 19 between Ueno Asami (as black) and Fujisawa Rina (as white). At move 159, Asami cut with K14.

It involves a followup of the life and death of the white group M10, and most human players would be able to judge it, but AI (from Golaxy to Katago) all seem to deem it "alive" and continue to fight without securing the life of that group first. While a human player can read out the sequence like Rina did during the game and make the right choice.

But it takes like 10k+ playouts before AI realize the K14 cut is a very good move, and (with few playouts AI even judges K14 as a blunder). And it takes millions of playouts before AI realizes the M10 white group is in trouble.

What other AI blindspots and hallucinations have you seen in real games?

Kids are creative, and smart kids will get bored very quickly, even if repeat exercises are necessary practice for them.

(There is no badukshitposting community on lemmy. Maybe we should accommodate more diverse posts here).

I've heard of the "carving" about weiqi, but never actually saw it, or even saw a picture of it till I start researching the history.

The report in the link was written by the head of the Luoyang Weiqi Museum, so I am farily confident the authetisty of the pictures she took. And I will make a brief translation and introduction below:

First of all, these carvings are called 畫像石. a very specific styles of carvings used in the Han Dynasty China (the 2nd century BC to the 2nd century AD). So very high confident it is of its time.

Here are the overall picture of the stone the carving is from

The dimension is 254 cm in length, 85 cm in height, and 26 cm thick, weight more than 1000 kg.

We can see other recreation activities next to the "博弈", people were dancing and playing some kind of instrument (likely some kind of drum, or bell).

There are other carvings of some daily activities in the back of the stone as well

In the close up shot, we can see there are two types of game boards carved on the stone, one is Go in the front, and another smaller board on the side, called 博 (where only recently we've uncovered the actual game board for it), and Go has been associated with it for a very long time - "博弈". We are certain they were both quite prominate during the Han Dynasty (probably more for 博 than weiqi).

Very hard to make out the number of lines on the carving, it degraded quite a bit. But it is for sure more than 9 lines or 13 lines, and on the range at least 15 plus even up to 19. (simply cannot tell for sure, and cannot rule out the original stone worker made it incorrectly to begin with)

This is from an exhibit at the Suzhou museum last year. The topic of the exhibit is - "Restoration of Peace, the Four Faces of Han Dynasty Civilization", about archeological finds from Han Dynasty China near Suzhou.

The stones are made of a type of glass and not necessary "black and white" but blue and slightly yellow white. (or they might have been darker and lighter, but the colors degraded or changed over time). It was found in a tomb of a Han dynasty high royal prince. They didn't find a board, but a cloth with lines drawn on it and too degraded to know how many lines it was.

So it is possible that we didn't find many old game boards (or older than Han dynasty), not due to out of luck (or as some suggested Go was invented in Han dynasty), but because ancient players faced the same issue we are, and find it impractical to carry giant wood or stone slaps as gobans, and chose lighter materials like cloth or leather and draw grid on them as the board. Much cheaper and easier to produce.