Sorry, but I have to ask

Do you think if they assigned the computer that played the best chess, or the game of GO, could also accomplish successful DI with a robot arm?

If I understand it, it's more or less a black box sort of program given the basic parameters of the goal, then can come across relationships that humans may not see?

Quote: rxwine

Do you think if they assigned the computer that played the best chess, or the game of GO, could also accomplish successful DI with a robot arm?

If I understand it, it's more or less a black box sort of program given the basic parameters of the goal, then can come across relationships that humans may not see?
link to original post

maybe with quantum computing... maybe now who knows though... i believe so for all of your questions

Quote: rxwine

Do you think if they assigned the computer that played the best chess, or the game of GO, could also accomplish successful DI with a robot arm?

If I understand it, it's more or less a black box sort of program given the basic parameters of the goal, then can come across relationships that humans may not see?
link to original post

That software? No.

A derivative software with a high precision setting/launching appurtenance? Maybe.

I think it would get easier with heavier (albeit fairly weighted) dice. Just a hunch, no science to back my opinion, and I am fascinated to hear why I'm wrong.

So, we build a contraption to launch dice a certain distance down a standard craps table, ensure that we have a legal throw (ie, bounces off the back wall) and landing free from any other obstacle (like cheques), would we be able to demonstrate dice influence?

Maybe. But you knew that would be the answer anyway.

First, there would be a very long calibration period. First target: ensure that the exact same signals are given to the actuators in the arm. This means that the output of the A/D converters are exactly the same each time. As the computer warms up, you have to compensate for the various electrical components' changes as well. Resistors change their ohms, capacitors their capacitance, chips change output voltage. One of the hardest problems in the early days, for instance, is compensating for the expansion of the hard drive disks as they warmed up. It's not trivial, but it can be done. Assumption: the actuators of the 'arm' get the exact same signals every time.

Second, we'd have to ensure that the arm moved in the exact same way each time. The arc, from a given point, would swing back, forward, and release the load at precisely the same time, in the same manner, every time under the same signals. There would always be errors, since we're now dealing with the mechanical. Bearings never respond exactly, contaminants in the moving surfaces creates errors in the arm positioning, acceleration, release point, and such. Now, you could generate feedback to the controller, which could alter the signals sent with micro-second precision, so the final release point occurs at the same place, with the same velocity, in all three dimensions. Again, it's doable, but the sources of error get to be ridiculous. Even tremors in the foundation of the building may need to be dealt with.

Once you do all that, you now need to ensure that the arm picks up the dice the exact same way, every time. The dice will require micrometer loading precision, lest off-axis thrust cause them to drift away from their ideal trajectory.

Then, I think, you'll have a chance to speak meaningfully about 'hitting the same spot on the felt', as the first goal of your computer arm.

There's more, but that's enough 'wall of text' for me tonight.

If they can land a robot on an asteroid --AN ASTEROID -- they can build a robotic arm to precisely throw two dice legally to get the same result on each throw (barring changes to the felt or rubber over time).

Quote: BillHasRetired

…Resistors change their ohms, capacitors their capacitance, chips change output voltage...
link to original post

Had a hard time reading this sentence without thinking of one of David Addison’s go-to lines in Moonlighting. “Do bears bear? Do bees bee?” Maybe it’s just late…!

It shows within the first few seconds…

I appreciate the responses, and the OP is an interesting question. Now if we could only get the casinos to let my android throw in my place without accusing me of using a device to cheat…!

Quote: AlanMendelson

If they can land a robot on an asteroid --AN ASTEROID -- they can build a robotic arm to precisely throw two dice legally to get the same result on each throw (barring changes to the felt or rubber over time).
link to original post

And it probably still won't be able to gain an advantage.

Quote: BillHasRetired

As the computer warms up, you have to compensate for the various electrical components' changes as well. Resistors change their ohms, capacitors their capacitance, chips change output voltage. One of the hardest problems in the early days, for instance, is compensating for the expansion of the hard drive disks as they warmed up. It's not trivial, but it can be done. Assumption: the actuators of the 'arm' get the exact same signals every time.
link to original post

You "obviously" need the critical parts in an oven, like they used to do with high precision frequency crystals.

Can't rely on mechanical limits; if the throwing arm slaps the mechanical stop, it will obviously deform both components at the contact point by half the width of an atom, and there goes your repeatability out the window as the new geometry kicks in...

In precision manufacturing plants they have been able to engineer repeatability in mechanical systems (tools and apparatus that move objects) to a very high degree. Great attention has to be given to temperature of all moving parts. I have toured one such U.S. manufactring plant, which makes parts for classified federal missions, and was deeply impressed by the tolerances that they achieve and the state of the art of precision lathes, etc.

So, designing a mechanical system for throwing dice with extremely high precision is within the state of the art by several orders of magnitude, in my opinion.

DI with a robot arm?



I'm sure he could

Quote: AxelWolf

Quote: AlanMendelson

If they can land a robot on an asteroid --AN ASTEROID -- they can build a robotic arm to precisely throw two dice legally to get the same result on each throw (barring changes to the felt or rubber over time).
link to original post

And it probably still won't be able to gain an advantage.
link to original post

Of course a precise robot would have an advantage. Remember there are robots that perform brain surgery now.

A robot could set and throw the dice to a precise spot with precise delivery that would result in the dice reacting the same way each time.

By the way, I thought the "DI is impossible" argument ended after the discussion about what those European researchers found?

The question isnt about the possibility-- the question is who has the ability?

Quote: AlanMendelson

Quote: AxelWolf

Quote: AlanMendelson

If they can land a robot on an asteroid --AN ASTEROID -- they can build a robotic arm to precisely throw two dice legally to get the same result on each throw (barring changes to the felt or rubber over time).
link to original post

And it probably still won't be able to gain an advantage.
link to original post

Of course a precise robot would have an advantage. Remember there are robots that perform brain surgery now.

A robot could set and throw the dice to a precise spot with precise delivery that would result in the dice reacting the same way each time.

By the way, I thought the "DI is impossible" argument ended after the discussion about what those European researchers found?

The question isnt about the possibility-- the question is who has the ability?
link to original post

Nope, because they didn't do it in real casino conditions and all the things that come along with that.

Thought of a name. We had Deep Blue, AlphaGo.

How about CrapArm?

Quote: AxelWolf

Quote: AlanMendelson

Quote: AxelWolf

Quote: AlanMendelson

If they can land a robot on an asteroid --AN ASTEROID -- they can build a robotic arm to precisely throw two dice legally to get the same result on each throw (barring changes to the felt or rubber over time).
link to original post

And it probably still won't be able to gain an advantage.
link to original post

Of course a precise robot would have an advantage. Remember there are robots that perform brain surgery now.

A robot could set and throw the dice to a precise spot with precise delivery that would result in the dice reacting the same way each time.

By the way, I thought the "DI is impossible" argument ended after the discussion about what those European researchers found?

The question isnt about the possibility-- the question is who has the ability?
link to original post

Nope, because they didn't do it in real casino conditions and all the things that come along with that.
link to original post

Tell me something. Card counters who practice at home and perfect their counting at home... do they simply fall apart when faced with real casino conditions?

Do the rules of physics change under real casino conditions?

Quote: AxelWolf

Nope, because they didn't do it in real casino conditions and all the things that come along with that.
link to original post

Be sure to make the robot female so it doesn't get distracted by the cocktail waitresses!

Quote: 100xOdds

DI with a robot arm?



I'm sure he could
link to original post

What a sight it would have been if Brent Spiner had shown up at a craps table exhibiting any of Commander Data’s behavior and started dice setting! Now, not so much, due to aging… (Sorry, Brent, but we all must age!)

Quote: AlanMendelson

Of course a precise robot would have an advantage. Remember there are robots that perform brain surgery now.

My back surgery was robotically assisted.

Quote: camapl

Quote: 100xOdds

DI with a robot arm?



I'm sure he could
link to original post

What a sight it would have been if Brent Spiner had shown up at a craps table exhibiting any of Commander Data’s behavior and started dice setting! Now, not so much, due to aging… (Sorry, Brent, but we all must age!)
link to original post

Didnt you know Data was an expert DI?

https://youtu.be/0EHMBfmayUA

Quote: 100xOdds

DI with a robot arm?



I'm sure he could
link to original post

And he did, in the episode "The Royale", although his "influence" consisted of actually changing the shape of the dice by squeezing them.

As for whether or not a calibrated robot arm could repeat a throw consistently, you also have to take into account any changes in the air that might affect each throw. Yes, we did land on an asteroid, but presumably it was smart enough to make subtle course corrections en route - something dice can't do once they leave the shooter's hand.

Quote: ThatDonGuy

Quote: 100xOdds

DI with a robot arm?



I'm sure he could
link to original post

And he did, in the episode "The Royale", although his "influence" consisted of actually changing the shape of the dice by squeezing them.

As for whether or not a calibrated robot arm could repeat a throw consistently, you also have to take into account any changes in the air that might affect each throw. Yes, we did land on an asteroid, but presumably it was smart enough to make subtle course corrections en route - something dice can't do once they leave the shooter's hand.
link to original post

The doctorate students in Norway said air currents had no impact on their study.

There are some important physical differences between the Norwegian scientific study and real-life craps tables.
- the back wall
- with pyramidal surface texture

There are some important physical differences between the Norwegian scientific study and real-life craps tables.
- the back wall
- with pyramidal surface texture

Duplicate post

And theres an issue about lengths of the table and the surface of the table but none of those issues matter because the point of the study was that DICE ROLLS ARE NOT TRULY RANDOM. And they are INFLUENCED by how they're set.
link to original post

Quote: AlanMendelson

Quote: AxelWolf

Quote: AlanMendelson

Quote: AxelWolf

Quote: AlanMendelson

If they can land a robot on an asteroid --AN ASTEROID -- they can build a robotic arm to precisely throw two dice legally to get the same result on each throw (barring changes to the felt or rubber over time).
link to original post

And it probably still won't be able to gain an advantage.
link to original post

Of course a precise robot would have an advantage. Remember there are robots that perform brain surgery now.

A robot could set and throw the dice to a precise spot with precise delivery that would result in the dice reacting the same way each time.

By the way, I thought the "DI is impossible" argument ended after the discussion about what those European researchers found?

The question isnt about the possibility-- the question is who has the ability?
link to original post

Nope, because they didn't do it in real casino conditions and all the things that come along with that.
link to original post

Tell me something. Card counters who practice at home and perfect their counting at home... do they simply fall apart when faced with real casino conditions?

Do the rules of physics change under real casino conditions?
link to original post

You are comparing apples to oranges and you know it.

Quote: AlanMendelson

Quote: AxelWolf

Quote: AlanMendelson

Quote: AxelWolf

Quote: AlanMendelson

If they can land a robot on an asteroid --AN ASTEROID -- they can build a robotic arm to precisely throw two dice legally to get the same result on each throw (barring changes to the felt or rubber over time).
link to original post

And it probably still won't be able to gain an advantage.
link to original post

Of course a precise robot would have an advantage. Remember there are robots that perform brain surgery now.

A robot could set and throw the dice to a precise spot with precise delivery that would result in the dice reacting the same way each time.

By the way, I thought the "DI is impossible" argument ended after the discussion about what those European researchers found?

The question isnt about the possibility-- the question is who has the ability?
link to original post

Nope, because they didn't do it in real casino conditions and all the things that come along with that.
link to original post

Tell me something. Card counters who practice at home and perfect their counting at home... do they simply fall apart when faced with real casino conditions?

Do the rules of physics change under real casino conditions?
link to original post

You are comparing apples to oranges and you know it.

Quote: AxelWolf

You are comparing apples to oranges and you know it.

Actually, he may NOT know it: that could explain a lot.

I'll say it again. The Norwegian study found dice rolls were not random and the setting of the dice affected the end result.

I'll take that conclusion, thank you.

Quote: AlanMendelson

I'll say it again. The Norwegian study found dice rolls were not random and the setting of the dice affected the end result.

I'll take that conclusion, thank you.
link to original post

Did the study find that it was enough to gain an advantage in a casino?

Quote: rxwine

Do you think if they assigned the computer that played the best chess, or the game of GO, could also accomplish successful DI with a robot arm?

If I understand it, it's more or less a black box sort of program given the basic parameters of the goal, then can come across relationships that humans may not see?
link to original post

The short answer: Yes.
The slightly longer answer: Yes but...
The much longer answer: [TLDR]

Since my wife went to visit her parents for a week I've got plenty of free time so here goes....

The technique you want to use in this situation is called reinforcement learning [REF-1]. It's been used to train AIs to play both board games like Go as well as video games like DOTA. It's also been used to train robot hands to manipulate blocks [REF-2] and robot dogs to walk [REF-3].

The main difference between training an AI to play a board or video game and training a robot to walk or juggle cubes is for the later situation you need to use model-based reinforcement learning. The model is a program that simulates some aspect of the physical world. The good news is using a model speeds up the training by several orders of magnitude. The bad news is you need to build a high-fidelity sim model. Luckily a lot of software tools are now available to make the modelling and simulation work easier (e.g., NVIDIA's OMNIVERSE).

So to create a "CRAP-BOT" would require a 3 phase project:

1. create the sim model and use it to train the CRAP-BOT software
2. install the s/w in the actual CRAP-BOT hardware and then test and validate in the lab
3. take CRAP-BOT to an actual casino and see if it can shift the odds

Phase-3 is the only way to deal with the perfectly valid objections from Axel, Dietr, and others noting that the real-world is not as simple or as forgiving as researchers often assume.

Digging a level deeper, to do Phase 1 you need a craps table, some high-speed cameras, and a robot arm that can throw the dice like a human. These would be used to collect training data on the physics of how dice bounce off the surface and back wall of a craps table under varying conditions. No attempt at this point to control how the dice land; the goal is to get data covering all types of throws and bounces.The data is then used to tune the sim to match. The resulting sim is a digital twin of that one specific craps table. This gets combined with a digital twin of the CRAP-BOT. Now you can use reinforcement learning to quickly run a gazillion simulations with the result that the s/w can, at least in the simulated world, exhibit the maximum possible degree of DI.

Phase 2 would be demo of CRAP-BOT in the lab using the same table that was used to collect the training data. At this stage you would be addressing the points that BillHasRetired, Dieter, and Gordonm888 were discussing earlier [post]. You're also dealing with variations in table size & construction, where the robot is positioned, and subtle environmental issues not captured in the sim model (e.g., how humidity might influence the bounciness of the table). To handle that the CRAP-BOT needs sensors (i.e., "eyes") so it can adjust in real-time. (BTW, that's why I'm calling it CRAP-BOT instead of just CrapArm as rxwine suggested).

Phase 3 requires finding a casino willing to let you bring the whole damn rig in for testing at one or more of their tables. My guess is that if you scheduled it to coincide with COMDEX you would easily find a casino willing to sign-up just for the free media exposure.

Quote: AlanMendelson

The question isnt about the possibility-- the question is who has the ability?
link to original post

Pretty much any university in the world with a decent grad program in robotics. This would make a great Capstone project for a MS in robotics or AI/ML. I'm guessing a team of 5 to 8 working for a year would be in the ballpark. A more important question is who has the budget. For example, to train the s/w that played GO:

Quote: REF-1

according to DeepMind’s AlphaStar blog post, the company used 16 Google TPU v3 to train each of its agents for 14 days (and this is just one of the several phases of developing the AI). At current pricing rates ($8.00 / TPU hour), the company spent $43,000 to train each AI agent, and according to the paper, there were at least 18 agents, which amounts to $774,000—just for training!

Now add in the cost of developing/debugging all that hardware and you're easily pushing into the low millions.

So to summarize, yeah it's do-able and would make a great research project but you would probably need to get NVIDIA or Google to fund it.

REFERENCES:

REF-1 https://bdtechtalks.com/2019/05/28/what-is-reinforcement-learning/
REF-2 https://www.youtube.com/watch?v=jwSbzNHGflM
REF-3: https://www.youtube.com/watch?v=8sO7VS3q8d0

Quote: TumblingBones

So to summarize, yeah it's do-able and would make a great research project but you would probably need to get NVIDIA or Google to fund it.

link to original post

On the flipside, anyone who straps an arduino or a pi and a handful of electromagnets to one of those springy-arm lamps and makes it work in six weekends deserves the bragging rights.

Probably best to just figure out just the distance at which it's no longer possible. Let alone spend millions on a precision machine to throw the dice at a $5 piece of foam you can't thus know anything about. Perhaps, such foam has no guaranteed or "sweet" spot, and, say, takes some time to thus reset itself locally on various levels.

/questions/117144/is-it-really-impossible-to-calculate-in-advance-the-result-of-throwing-dice

/profile/Marcin-Kapitaniak/publication/234029725_The_three-dimensional_dynamics_of_the_die_throw/links/00b49528f8de4c42cb000000/The-three-dimensional-dynamics-of-the-die-throw.pdf