How much longer do counters realistically have?

Quote: billryan

It doesn't matter what Ai can or can't do. All I'm concerned with is what the public thinks. In short, the public jumps on any stock or ETF they think is AI-driven. If Pepsi hinted that Ai created their next-generation sodas, people would both invest in the stock and create buzz about the product.
Long term investments require solid foundations. Short term investments you should go with rumours.
link to original post

(sigh) People still chase fads.

Quote: Archvaldor1

Quote: billryan

It doesn't matter what Ai can or can't do. All I'm concerned with is what the public thinks. In short, the public jumps on any stock or ETF they think is AI-driven. If Pepsi hinted that Ai created their next-generation sodas, people would both invest in the stock and create buzz about the product.
Long term investments require solid foundations. Short term investments you should go with rumours.
link to original post

That doesn't really work. You are trying to predict the actions of people who don't know shit about something. Good luck with that. Other people will have faster data feeds than you, better sources of information, you can't just look at the news cycle and jump on something and expect to beat the vig. Everyone would be doing it if it were that easy.
link to original post

absolutely 100% correct

great post Archvaldor1

.

Quote: Archvaldor1

Quote: BTLWI

It's pretty easy for an AI agent using surveillance footage to observe the cards and keep a perfect count. Also easy for an AI agent to observe blackjack player betting patterns and learn what is considered average player betting behavior. Also easy for an AI agent to combine count and betting patterns over time to identify deviations from normal (possible counters).

That's a super simple AI system to develop and run on surveillance footage. Does that already exist? I only vulture slots so I'm not up to par on the BJ world. I would guess systems like this would be cheap enough for all casinos to run within a year or two.

On the flip side, AI agents can be run on top of a players hidden camera footage and signal to the player the perfect play and the player could get drunk out of their mind without counting. Modern day version of the Ken Uston computers. Hidden camera embedded in a button on your shirt. Smart phone in your pocket running AI agent to count. Signal via smart watch haptics (vibrations).

My prediction is AI running on top of surveillance will make counting for any length of time extremely difficult in the future, what say you?
link to original post

A few misconceptions here:

1. You can detect card counting with statistical analysis. The tech has existed for years. It is nothing to do with AI.
2. In the modern game humans do not significantly under-perform computer-perfect play.

There is this persistent math undergraduate fantasy about computer-perfect play being able to destroy casinos. Outside of deeply dealt single-deck games it is just not true. You might at best get a tenth of a per cent using a computer in a modern game shoe game. So you might win 0.6% instead of 0.5%. It is not worth it.

The modern excitement around AI is to do with Large Language Models, these are completely useless as a casino application for multiple reasons. They can't do math currently.
link to original post

The biggest misconception is you confusing LLM's with AI Agents which is what I specified in my post. Perhaps you need to get up to speed.

Quote: DRich

Quote: Archvaldor1

1. You can detect card counting with statistical analysis. The tech has existed for years. It is nothing to do with AI.

Thank you, I was going to go on a rant how people use buzzwords like "AI" without any clue what they are. You summed it up much better as this doesn't require any AI at all.
link to original post

I specified AI Agents in my post. Do you even know what that means without looking it up? I'll bet a lot of money you don't.

Quote: BTLWI

The biggest misconception is you confusing LLM's with AI Agents which is what I specified in my post. Perhaps you need to get up to speed.
link to original post

The distinction is irrelevant. You don't need any form of AI to detect card-counting.

Quote: Dieter

Quote: Hunterhill

Quote: Dieter

Quote: heatmap

Bloodhound from shuffle master

You’ll never find more than a single mention on the internet
link to original post

I believe it's usually referred to as "BJ Survey Voice".
link to original post

I don’t think bloodhound and BJ survey voice are the same thing. Not sure though. I have alot of experience with survey voice I have used it for hundreds of hours. It definitely had ways in which it could be fooled not sure if it’s been updated to correct those issues.
link to original post

I could be misunderstanding.

This is a screenshot from an article https://euroasiacasino.com/wp-content/uploads/2015/09/Article-Blackjack-Counters_Call-the-Bloodhound1.pdf
link to original post

I stand corrected, I wasn’t aware that it was sold to shuffle master.

As for the specific topic, how long, I think it all depends individually on the casino or provider management. There are still many managers who do not understand mathematics and act on superstitions.

As for whether to use AI or not, it also depends on the methods of the analytical department. It can really help if there are a lot of players. But it doesn’t necessarily all come down to analyzing bets, but also taking into account the player’s behavior and balance, and many other telling factors. AI can really help to pay attention to some points, and then the analyst will make decisions.

Quote: AxelWolf

Quote: goose

They don't need AI for this. Online casinos are already doing it. They have a stipulation on their games like baccarat where they can turn the side bets off at their discretion. I was observing some online baccarat and noticed the bets would appear to be turned off at random points for the shoe - never at a consistent # hands lets say. So I started counting the side bets, turns out they disable the side bets the second the count is positive ev.

I think it's just a matter of time until they bring this into to the brick and mortar casinos. 5 - 10 years at most.
link to original post

Please tell me this isn't at a regulated online casino in the USA? I'm under the impression that if they are using a computer to change the odds that would be illegal in many jurisdictions.
link to original post

No, it's at live dealer casinos that serve US patrons. For a while, there was a bit to be made there, running a CA against the baccarat games, mostly the sidebets but occasionally you would go +EV on the tie, and once I actually went slightly +EV on the Player. I have a CA that will solve all the bets in a half second on the fly.

Before that, before US players were kicked out of all the live casino providers except two that deal mostly crap, you could really rake on some of their offerings. They were more worried about bonus abuse than anything else, so I just refused all bonuses. Took crypto and paid out in crypto too, so if I got douched out I can always come back with a new ID.

They could never do that in NV, best they could do is shuffle up when an alarm goes off but the players would up and leave if they did that. Not worth it.

Quote: Markwilam019s3

That’s a fascinating take! AI on both sides of the table could completely redefine the game. Casinos and players might be in a high-tech arms race soon—who do you think has the upper hand?
link to original post

Prove to me you have a soul.

edit: failed

Quote: DRich

Quote: darkoz

That's not the way AI works.

It's a learning program. It does work only on what it's programmed to learn but it learns things not programmed.

If it was just a program that did what it was told then nobody would care about defining AI as a such a game changer.

First, do you really think you understand computers better than I do? I have been programming computers for the last 45 years, of which 40 years of it I have been paid to do it.

Every computer program whether you call it an algorithm or AI only does exactly what it is told to do. On a side note to that, on a slot machine the software can never malfunction so keep that in mind if they try to claim a slot malfunction and not pay you.
link to original post

Curious about this claim. What about the rare effects of cosmic rays? I think this would qualify as the computer doing something it was not programmed to do. Would you classify those only as hardware errors?

Quote: Dobrij

As for the specific topic, how long, I think it all depends individually on the casino or provider management. There are still many managers who do not understand mathematics and act on superstitions.

As for whether to use AI or not, it also depends on the methods of the analytical department. It can really help if there are a lot of players. But it doesn’t necessarily all come down to analyzing bets, but also taking into account the player’s behavior and balance, and many other telling factors. AI can really help to pay attention to some points, and then the analyst will make decisions.
link to original post

Most serious ap's can take one look at a player and establish with a high level of confidence whether he's also an ap or not.

There's no way to explain that easily to a computer system. It is a whole synthesis of things and it is somewhat situational. If you could do it then ap's would adjust accordingly with relative ease, and you'd have to do a major update.

As with most things paying a smart guy to do something is going to be about 1000% more effective than screwing around with any form of AI. Notwithstanding the fact clueless people think AI has superpowers. I'm sure somebody will try though which as an ap I welcome.

I'd add: for whatever reason computer systems are not even remotely close to performing inferential analysis to the extent the hardware is capable of. Detection systems not just in casinos but with banking, law enforcement, corporate security just do not understand how to eliminate false positives and isolate behaviors threatening to them. The people working on this stuff don't even frame the problem correctly. A lot of it is to do with paying useless senior executives big bucks and under-paying software designers.
People have constantly asked me both as a hacker and an ap why the casinos and corporations don't hire me to fix their loopholes. I say I wouldn't work for those bastards, which I wouldn't. But the truth is they never make offers that would renumerate me anything like as well as working against them pays anyway.

Quote: djtehch34t

Quote: DRich

Quote: darkoz

That's not the way AI works.

It's a learning program. It does work only on what it's programmed to learn but it learns things not programmed.

If it was just a program that did what it was told then nobody would care about defining AI as a such a game changer.

First, do you really think you understand computers better than I do? I have been programming computers for the last 45 years, of which 40 years of it I have been paid to do it.

Every computer program whether you call it an algorithm or AI only does exactly what it is told to do. On a side note to that, on a slot machine the software can never malfunction so keep that in mind if they try to claim a slot malfunction and not pay you.
link to original post

Curious about this claim. What about the rare effects of cosmic rays? I think this would qualify as the computer doing something it was not programmed to do. Would you classify those only as hardware errors?
link to original post

I’m on my own side here but that event doesn’t happen often at all it’s so rare that you will most likely die in a plane accident before a cosmic ray alters a piece of your memory

Quote: heatmap

Quote: djtehch34t

Quote: DRich

Quote: darkoz

That's not the way AI works.

It's a learning program. It does work only on what it's programmed to learn but it learns things not programmed.

If it was just a program that did what it was told then nobody would care about defining AI as a such a game changer.

First, do you really think you understand computers better than I do? I have been programming computers for the last 45 years, of which 40 years of it I have been paid to do it.

Every computer program whether you call it an algorithm or AI only does exactly what it is told to do. On a side note to that, on a slot machine the software can never malfunction so keep that in mind if they try to claim a slot malfunction and not pay you.
link to original post

Curious about this claim. What about the rare effects of cosmic rays? I think this would qualify as the computer doing something it was not programmed to do. Would you classify those only as hardware errors?
link to original post

I’m on my own side here but that event doesn’t happen often at all it’s so rare that you will most likely die in a plane accident before a cosmic ray alters a piece of your memory
link to original post

Do you happen to have a source for this frequency? I actually tried looking before your post, but could only find references to an old IBM study which looked too frequent (probably outdated). There's also the additional factor of how many bitflips actually make a difference and aren't caught by error correction...

Quote: djtehch34t

Quote: DRich

Quote: darkoz

That's not the way AI works.

It's a learning program. It does work only on what it's programmed to learn but it learns things not programmed.

If it was just a program that did what it was told then nobody would care about defining AI as a such a game changer.

First, do you really think you understand computers better than I do? I have been programming computers for the last 45 years, of which 40 years of it I have been paid to do it.

Every computer program whether you call it an algorithm or AI only does exactly what it is told to do. On a side note to that, on a slot machine the software can never malfunction so keep that in mind if they try to claim a slot malfunction and not pay you.
link to original post

Curious about this claim. What about the rare effects of cosmic rays? I think this would qualify as the computer doing something it was not programmed to do. Would you classify those only as hardware errors?
link to original post

I agree, that could cause a computer malfunction but my claim was that the software can't malfunction.

DRich appears to be saying that software can't fail because it's created by humans therefore it's the human programmers that failed.

So basically a very specious argument here over semantics.

Certainly when a cars brakes fail, the argument that the cars brakes DIDN'T fail but the humans who built the brakes were the culprit would annoy anyone to the nth degree Cars break down regardless of who built them.

Anyone who has used software in the last twenty years knows that software fails to work sometimes as well

And anyone with half a brain knows that if a slot jackpot is nulled due to malfunction that the plaintiffs argument that software can't malfunction would have zero chance of being recognized as a valid argument.

So DRich is basically making a non-valid argument for the sake of arguing apparently.

AI should definitely not be anthropomorphized anymore than a car when it fails. But next time your brakes fail, I guarantee the last thought in your mind as you careen off the road will be "omg, they can't fail. Only the human who built them could fail".

Quote: darkoz

DRich appears to be saying that software can't fail because it's created by humans therefore it's the human programmers that failed.

The principle behind AI is that you are not simply getting a device to playback information but you are giving it the capacity to learn and behave autonomously. So I would have thought blaming the AI's designers had less weight than this logic would have normally.

This would be less the case if a specific bias was programmed into the AI. This actually happens. You'll notice AI's tend to say way too much and do things like over-simplify -they've been specifically told to do this for commercial reasons. This is going to open a can of worms for the developers because the ARE directly culpable for things like hallucinations resulting from this.

For me this question is highly situational.

Quote: darkoz

DRich appears to be saying that software can't fail because it's created by humans therefore it's the human programmers that failed.

I am saying that software can only do what it is told to do, nothing else.

The only correct answer is "as long as the casinos want".

A hi-lo and similar strategies could have been stopped decades ago. Any casino can just use a continuous shuffle machine or reduce payouts on blackjack from a 50% premium to 20%.

A much more important question is "do casinos ever offer bets they don't want customers to make"? If no, blackjack players should never be backed off. If yes, they are intentionally screwing over their investors, customers, employees, and communities.

Quote: DRich

Quote: darkoz

DRich appears to be saying that software can't fail because it's created by humans therefore it's the human programmers that failed.

I am saying that software can only do what it is told to do, nothing else.
link to original post

I understand what you are saying. However...

The current AI is autonomous in its learning.

You gave an example of telling the AI that 1+1=3 would be the fault of the programmer.

Let's run with that but go with 2+2=4.

With the current AI, the programmer DOESN'T tell the AI that 2+2=4. The programmer tells the AI to scour the internet and LEARN that 2+2=4. It's not programmed into the AI directly but it's programmed for it to learn it.

Now at this point you will probably argue it's programmed either way. The difference here is the AI is autonomously learning the data. And false data enters.

In our example, the AI finds information that is incorrect and processes that. It finds George Orwell's passages from "1984" where the main character learns that 2+2=5. (You have to read the book to understand). But it recognizes that info and now when prompted replies that 2+2=5.

It wasn't programmed to tell you that by the programmer. It was programmed to find the correct answer. And it fed off incorrect data.

The programmer never suspected the AI would discover 2+2=5 but we wind up with what is already popularized as a "hallucination".

I really don't believe IMO that this scenario is the same as simply, "the programmers are at fault because they told it that 2+2=5".

That's my opinion. Obviously you have a different one. I just can not make sense of your opinion under these circumstances.

Quote: darkoz

Quote: DRich

Quote: darkoz

DRich appears to be saying that software can't fail because it's created by humans therefore it's the human programmers that failed.

I am saying that software can only do what it is told to do, nothing else.
link to original post

I understand what you are saying.

The current AI is autonomous in its learning.

You gave an example of telling the AI that 1+1=3 would be the fault of the programmer.

Let's run with that but go with 2+2=4.

With the current AI, the programmer DOESN'T tell the AI that 2+2=4. The programmer tells the AI to scour the internet and LEARN that 2+2=4. It's not programmed into the AI directly but it's programmed for it to learn it.

Now at this point you will probably argue it's programmed either way. The difference here is the AI is autonomously learning the data. And false data enters.

In our example, the AI finds information that is incorrect and processes that. It finds George Orwell's passages from "1984" where the main character learns that 2+2=5. (You have to read the book to understand). But it recognizes that info and now when prompted replies that 2+2=5.

It wasn't programmed to tell you that by the programmer. It was programmed to find the correct answer. And it fed off the wrong answer.

The programmer never suspected the AI would discover 2+2=5 but we wind up with what is already popularized as a "hallucination".

I really don't believe IMO that this scenario is the same as simply, tm"the programmers told it what to do".

That's my opinion. Obviously you have a different one. I just can make sense of your opinion under these circumstances.
link to original post

I don't know why you are starting this up again. What you are referring to as AI, is a software program that mines data. The Ai can only do exactly what it is programmed to do.

If you understand computers and processors everything is derived from four basic logic gates. "AND", "OR", "NAND". and "NOR". Those gates determine how a processor works, The truth is everything could be derived from just the "NAND" logic gate but would be less efficient that way. At the university we had to design a simple processor made only with "NAND" gates.

All output is a result of running through those gates and is discreet based on the inputs. In this case the software supplies the inputs and the processor gives the output. The outputs are completely determined by the inputs.

I am sure with your deep understanding of how computers work this will make complete sense.

Darkoz, you might want to read this book. No, I am not calling you a dummie.

"This book will help people:
Separate the reality of artificial intelligence from the hyoe
Know what Artificial intelligence can accomplish and what its limits are."

Quote: DRich

Quote: darkoz

Quote: DRich

Quote: darkoz

DRich appears to be saying that software can't fail because it's created by humans therefore it's the human programmers that failed.

I am saying that software can only do what it is told to do, nothing else.
link to original post

I understand what you are saying.

The current AI is autonomous in its learning.

You gave an example of telling the AI that 1+1=3 would be the fault of the programmer.

Let's run with that but go with 2+2=4.

With the current AI, the programmer DOESN'T tell the AI that 2+2=4. The programmer tells the AI to scour the internet and LEARN that 2+2=4. It's not programmed into the AI directly but it's programmed for it to learn it.

Now at this point you will probably argue it's programmed either way. The difference here is the AI is autonomously learning the data. And false data enters.

In our example, the AI finds information that is incorrect and processes that. It finds George Orwell's passages from "1984" where the main character learns that 2+2=5. (You have to read the book to understand). But it recognizes that info and now when prompted replies that 2+2=5.

It wasn't programmed to tell you that by the programmer. It was programmed to find the correct answer. And it fed off the wrong answer.

The programmer never suspected the AI would discover 2+2=5 but we wind up with what is already popularized as a "hallucination".

I really don't believe IMO that this scenario is the same as simply, tm"the programmers told it what to do".

That's my opinion. Obviously you have a different one. I just can make sense of your opinion under these circumstances.
link to original post

I don't know why you are starting this up again. What you are referring to as AI, is a software program that mines data. The Ai can only do exactly what it is programmed to do.

If you understand computers and processors everything is derived from four basic logic gates. "AND", "OR", "NAND". and "NOR". Those gates determine how a processor works, The truth is everything could be derived from just the "NAND" logic gate but would be less efficient that way. At the university we had to design a simple processor made only with "NAND" gates.

All output is a result of running through those gates and is discreet based on the inputs. In this case the software supplies the inputs and the processor gives the output. The outputs are completely determined by the inputs.

I am sure with your deep understanding of how computers work this will make complete sense.
link to original post

It makes perfect sense.

Current AI does not use those logic gates AND OR NAND NOR

They use a more sophisticated neural network.

Once again you insist on applying your outdated by 4 decades knowledge of computer programming to the modern era. Just think about that a moment.

Just ask the AI

AI software operates within the boundaries of its programming, but it can also exhibit behaviors that may seem beyond what it was initially designed for. Here's a more detailed explanation:

1. **Predefined Tasks**: Most AI is designed for specific tasks, such as language translation, image recognition, or playing games like chess. It follows algorithms that guide it through solving those tasks effectively.

2. **Learning and Adaptation**: Many AI systems, particularly those based on machine learning, can learn from data. While their learning is constrained by the training data and algorithms, they can adapt to new patterns or inputs that they weren’t explicitly programmed to recognize. For example, an AI trained to recognize cats in images may improve its ability to do so over time as it gets exposed to more diverse examples.

3. **Emergent Behavior**: Sometimes, AI systems can exhibit "emergent" behavior, where the way they handle tasks or data interactions results in outcomes that their original programmers didn’t anticipate. This happens because the AI may discover complex patterns that go beyond simple task instructions.

4. **Limits**: However, AI is still fundamentally restricted by the data and algorithms it's trained on. It doesn't have independent thought, creativity, or awareness. An AI can only generate results based on patterns in the data and the structure it was given, which means it cannot act beyond what its design allows unless it is explicitly reprogrammed or retrained.

So, AI can sometimes surprise us with unexpected behaviors, but it is ultimately bound by its programming, data, and underlying models.

Quote: AxelWolf

Just ask the AI

AI software operates within the boundaries of its programming, but it can also exhibit behaviors that may seem beyond what it was initially designed for. Here's a more detailed explanation:

1. **Predefined Tasks**: Most AI is designed for specific tasks, such as language translation, image recognition, or playing games like chess. It follows algorithms that guide it through solving those tasks effectively.

2. **Learning and Adaptation**: Many AI systems, particularly those based on machine learning, can learn from data. While their learning is constrained by the training data and algorithms, they can adapt to new patterns or inputs that they weren’t explicitly programmed to recognize. For example, an AI trained to recognize cats in images may improve its ability to do so over time as it gets exposed to more diverse examples.

3. **Emergent Behavior**: Sometimes, AI systems can exhibit "emergent" behavior, where the way they handle tasks or data interactions results in outcomes that their original programmers didn’t anticipate. This happens because the AI may discover complex patterns that go beyond simple task instructions.

4. **Limits**: However, AI is still fundamentally restricted by the data and algorithms it's trained on. It doesn't have independent thought, creativity, or awareness. An AI can only generate results based on patterns in the data and the structure it was given, which means it cannot act beyond what its design allows unless it is explicitly reprogrammed or retrained.

So, AI can sometimes surprise us with unexpected behaviors, but it is ultimately bound by its programming, data, and underlying models.
link to original post

Thanks Axel

My argument has always been that the AI is trained to adapt to data and it can feed off incorrect data

The whole concept of "garbage in, garbage out" is certainly not a new one.

In the case of AI, it's very programming can lead to it being fed garbage.

That a computer programmer with 40 years experience doesn't understand this is baffling.

Quote: darkoz

Quote: DRich

Quote: darkoz

Quote: DRich

Quote: darkoz

DRich appears to be saying that software can't fail because it's created by humans therefore it's the human programmers that failed.

I am saying that software can only do what it is told to do, nothing else.
link to original post

I understand what you are saying.

The current AI is autonomous in its learning.

You gave an example of telling the AI that 1+1=3 would be the fault of the programmer.

Let's run with that but go with 2+2=4.

With the current AI, the programmer DOESN'T tell the AI that 2+2=4. The programmer tells the AI to scour the internet and LEARN that 2+2=4. It's not programmed into the AI directly but it's programmed for it to learn it.

Now at this point you will probably argue it's programmed either way. The difference here is the AI is autonomously learning the data. And false data enters.

In our example, the AI finds information that is incorrect and processes that. It finds George Orwell's passages from "1984" where the main character learns that 2+2=5. (You have to read the book to understand). But it recognizes that info and now when prompted replies that 2+2=5.

It wasn't programmed to tell you that by the programmer. It was programmed to find the correct answer. And it fed off the wrong answer.

The programmer never suspected the AI would discover 2+2=5 but we wind up with what is already popularized as a "hallucination".

I really don't believe IMO that this scenario is the same as simply, tm"the programmers told it what to do".

That's my opinion. Obviously you have a different one. I just can make sense of your opinion under these circumstances.
link to original post

I don't know why you are starting this up again. What you are referring to as AI, is a software program that mines data. The Ai can only do exactly what it is programmed to do.

If you understand computers and processors everything is derived from four basic logic gates. "AND", "OR", "NAND". and "NOR". Those gates determine how a processor works, The truth is everything could be derived from just the "NAND" logic gate but would be less efficient that way. At the university we had to design a simple processor made only with "NAND" gates.

All output is a result of running through those gates and is discreet based on the inputs. In this case the software supplies the inputs and the processor gives the output. The outputs are completely determined by the inputs.

I am sure with your deep understanding of how computers work this will make complete sense.
link to original post

It makes perfect sense.

Current AI does not use those logic gates AND OR NAND NOR

They use a more sophisticated neural network.

Once again you insist on applying your outdated by 4 decades knowledge of computer programming to the modern era. Just think about that a moment.


link to original post

You must not have read the second part of your own post.

"AI does leverage the functionality of these gates within its circuitry."

Gates are not software, they are hardware.

Quote: darkoz

\
In the case of AI, it's very programming can lead to it being fed garbage.

That a computer programmer with 40 years experience doesn't understand this is baffling.

AI is programming and both correct and incorrect results are derived by the program running the AI. AI is not some autonomous thing, it is a software program that does exactly what its programmer told it to do.

What is baffling is that you think you know more about how compuetrs work than I do. When your software is featured on the cover of Science magazine you may begin to have a valid argument.

Quote: DRich

Quote: darkoz

\
In the case of AI, it's very programming can lead to it being fed garbage.

That a computer programmer with 40 years experience doesn't understand this is baffling.

AI is programming and both correct and incorrect results are derived by the program running the AI. AI is not some autonomous thing, it is a software program that does exactly what its programmer told it to do.

What is baffling is that you think you know more about how compuetrs work than I do. When your software is featured on the cover of Science magazine you may begin to have a valid argument.
link to original post

When experts get things wrong, they get things wrong.

You stated above that neural networks use logic gates that I just showed they don't use.

You got it wrong my man.

Personally if I was an expert and proven wrong I would just admit it.

Quote: darkoz

Quote: DRich

Quote: darkoz

\
In the case of AI, it's very programming can lead to it being fed garbage.

That a computer programmer with 40 years experience doesn't understand this is baffling.

AI is programming and both correct and incorrect results are derived by the program running the AI. AI is not some autonomous thing, it is a software program that does exactly what its programmer told it to do.

What is baffling is that you think you know more about how compuetrs work than I do. When your software is featured on the cover of Science magazine you may begin to have a valid argument.
link to original post

When experts get things wrong, they get things wrong.

You stated above that neural networks use logic gates that I just showed they don't use.

You got it wrong my man.

Personally if I was an expert and proven wrong I would just admit it.
link to original post

That last line made me spit up my smoothie.

Quote: billryan

Quote: darkoz

Quote: DRich

Quote: darkoz

\
In the case of AI, it's very programming can lead to it being fed garbage.

That a computer programmer with 40 years experience doesn't understand this is baffling.

AI is programming and both correct and incorrect results are derived by the program running the AI. AI is not some autonomous thing, it is a software program that does exactly what its programmer told it to do.

What is baffling is that you think you know more about how compuetrs work than I do. When your software is featured on the cover of Science magazine you may begin to have a valid argument.
link to original post

When experts get things wrong, they get things wrong.

You stated above that neural networks use logic gates that I just showed they don't use.

You got it wrong my man.

Personally if I was an expert and proven wrong I would just admit it.
link to original post

That last line made me spit up my smoothie.
link to original post

Yeah that last line does have some humor in it. Me rushing.

But everyone should know what I meant.

DRich claims to be an expert in computers. He claimed AI uses four logic gates. And you can look it up yourself, AI uses a neural network that doesn't use the logic gates he explained .

If I was him, I would just admit I got it wrong. It happens.

PS, did your smoothie stain your shirt?

Can’t nobody tell me nothing.

Quote: darkoz

And you can look it up yourself, AI uses a neural network that doesn't use the logic gates he explained .
link to original post

I took your advice and looked it up -

here's what I found - from the link:

"logic gates, which featured early on in the development of yesterday's computing systems, CONTINUE TO BE THE UNDERLYING RESOURCES for advanced work in neuronal networks and the adoption of even strong machine learning and artificial intelligence tools"

and:

"artificial intelligence work often includes the use of logic gates in computing systems"

and:

"logic gates provide the building blocks for machine learning, artificial intelligence and everything that comes along with it"


https://www.techopedia.com/how-are-logic-gates-precursors-to-ai-and-building-blocks-for-neural-networks/7/33018

.

Quote: lilredrooster

Quote: darkoz

And you can look it up yourself, AI uses a neural network that doesn't use the logic gates he explained .
link to original post

I took your advice and looked it up -

here's what I found - from the link:

"logic gates, which featured early on in the development of yesterday's computing systems, CONTINUE TO BE THE UNDERLYING RESOURCES for advanced work in neuronal networks and the adoption of even strong machine learning and artificial intelligence tools"

and:

"artificial intelligence work often includes the use of logic gates in computing systems"

and:

"logic gates provide the building blocks for machine learning, artificial intelligence and everything that comes along with it"


https://www.techopedia.com/how-are-logic-gates-precursors-to-ai-and-building-blocks-for-neural-networks/7/33018

.
link to original post

Nice but those aren't the logic gates he claims.

I said they don't use the logic gates he claims they do, not that they don't use logic gates.

Quote: AutomaticMonkey

No, it's at live dealer casinos that serve US patrons. For a while, there was a bit to be made there, running a CA against the baccarat games, mostly the sidebets but occasionally you would go +EV on the tie, and once I actually went slightly +EV on the Player. I have a CA that will solve all the bets in a half second on the fly.

That was a great play. The sidebets kind of ruined that. Without them I would still be making a lot of money.

From what I can tell there doesn't seem to be anywhere which shuffles at much less than a deck nowadays online or offl, though I'd appreciate any information to the contrary, even to just confirm any still exist. .

Quote: Archvaldor1

Quote: AutomaticMonkey

No, it's at live dealer casinos that serve US patrons. For a while, there was a bit to be made there, running a CA against the baccarat games, mostly the sidebets but occasionally you would go +EV on the tie, and once I actually went slightly +EV on the Player. I have a CA that will solve all the bets in a half second on the fly.

That was a great play. The sidebets kind of ruined that. Without them I would still be making a lot of money.

From what I can tell there doesn't seem to be anywhere which shuffles at much less than a deck nowadays online or offl, though I'd appreciate any information to the contrary, even to just confirm any still exist. .
link to original post

What no sidebets? Sidebets and derivative games are my thing!

It's been a while since I've played online, but whatever that video house was that served Bovada was where I did most of that play, at various stores. They would put the cut card 11 cards from the end which was as deep as you can go without the risk of running out of cards, and I forgot how they did the burn but you always had pen of 1/4 to 1/3 of a deck. However they shut the most playable sidebets down with about 1.5-2 decks. You still could play them a lot though with a CA, which tends to make even crap bets playable as long as they are close to break-even, and that infamous baccarat sidebet which is pretty reasonable edge off the top but low EOR of cards was what it usually hit on. It also let you play Perfect Pairs down to the end. Poker based sidebets don't have normal EORs and my CA was not suit aware, would have taken too long to get my bets down if it was but in retrospect I could have added a non-CA routine to deal with that which would not have added much time to the calculation.

As far as live play at tables, here in Vegas all the tables I've played have been the traditional half deck or better pen.

Quote: darkoz

You stated above that neural networks use logic gates that I just showed they don't use.

You got it wrong my man.

LOL!! You got it wrong man. If you can show where I said Neural Networks use logic gates I will crown you the greatest AI professional in the world and also send you $100. If you can't, let's just end this discussion again.

Quote: DRich

Quote: darkoz

You stated above that neural networks use logic gates that I just showed they don't use.

You got it wrong my man.

LOL!! You got it wrong man. If you can show where I said Neural Networks use logic gates I will crown you the greatest AI professional in the world and also send you $100. If you can't, let's just end this discussion again.
link to original post

You didn't say it directly. I inferred it from a few of your comments through the thread.

Collecting those comments together for my $100 I see my inference in that regard may be wrong. I still see it but other people might not so I will just concede that point.

Like you said Let's just end it.

Quote: darkoz

Quote: DRich

Quote: darkoz

You stated above that neural networks use logic gates that I just showed they don't use.

You got it wrong my man.

LOL!! You got it wrong man. If you can show where I said Neural Networks use logic gates I will crown you the greatest AI professional in the world and also send you $100. If you can't, let's just end this discussion again.
link to original post

You didn't say it directly. I inferred it from a few of your comments through the thread.

Collecting those comments together for my $100 I see my inference in that regard may be wrong. I still see it but other people might not so I will just concede that point.

Like you said Let's just end it.
link to original post

Thank you. I don't think I ever mentioned Neural Networks.

Quote: AutomaticMonkey

Quote: Archvaldor1

Quote: AutomaticMonkey

No, it's at live dealer casinos that serve US patrons. For a while, there was a bit to be made there, running a CA against the baccarat games, mostly the sidebets but occasionally you would go +EV on the tie, and once I actually went slightly +EV on the Player. I have a CA that will solve all the bets in a half second on the fly.

That was a great play. The sidebets kind of ruined that. Without them I would still be making a lot of money.

From what I can tell there doesn't seem to be anywhere which shuffles at much less than a deck nowadays online or offl, though I'd appreciate any information to the contrary, even to just confirm any still exist. .
link to original post

What no sidebets? Sidebets and derivative games are my thing!

It's been a while since I've played online, but whatever that video house was that served Bovada was where I did most of that play, at various stores. They would put the cut card 11 cards from the end which was as deep as you can go without the risk of running out of cards, and I forgot how they did the burn but you always had pen of 1/4 to 1/3 of a deck. However they shut the most playable sidebets down with about 1.5-2 decks. You still could play them a lot though with a CA, which tends to make even crap bets playable as long as they are close to break-even, and that infamous baccarat sidebet which is pretty reasonable edge off the top but low EOR of cards was what it usually hit on. It also let you play Perfect Pairs down to the end. Poker based sidebets don't have normal EORs and my CA was not suit aware, would have taken too long to get my bets down if it was but in retrospect I could have added a non-CA routine to deal with that which would not have added much time to the calculation.

As far as live play at tables, here in Vegas all the tables I've played have been the traditional half deck or better pen.
link to original post

My aversion to the baccarat side-bets was mainly down to the fact you could get away with betting into the tie for decades-but then when they introduced obviously beatable side-bets everyone was doing it and pen tightened up. Obviously side-bets generally speaking are jam for counters.

Is there anywhere in Vegas that still deals down to the last quarter-deck? Or for that matter anywhere. I don't need specifics obviously. Couldn't find anywhere doing that on my last trip but I didn't look hard.

Quote: DRich

I am saying that software can only do what it is told to do, nothing else.
link to original post

If you program the software to answer "yes" if it is asked if it is conscious. it will say "yes" Just like a human. If you program the software to do something random if it is ask to prove it can't act independently, then it will do something random. Just like a human. . If you say the software doesn't have real experiences because you have it in a room and feed information into it, you can also put a human in a room and feed new information to the person, and it won't have actual experience either, but the human is still human. If you program the software with appropriate emotional designed responses, it will look like an emotional human. The software can give wrong information and will make mistakes, But a human will also. If you program the software to recognize the difference between a pat on the computer box and a strike to its computer and respond with "ouch" or "aw".or some other appropriate response. Just like a human.

I don't see a lot of ways to disprove a computer with the right program is not just like a human. Just takes programming.

Sorry for the tangent.

Quote: DRich

Quote: darkoz

DRich appears to be saying that software can't fail because it's created by humans therefore it's the human programmers that failed.

I am saying that software can only do what it is told to do, nothing else.
link to original post

Wrong.



Go to 16:30 and listen till 20:20

Quote: BTLWI

Quote: DRich

Quote: darkoz

DRich appears to be saying that software can't fail because it's created by humans therefore it's the human programmers that failed.

I am saying that software can only do what it is told to do, nothing else.
link to original post

Wrong.

Go to 16:30 and listen till 20:20
link to original post

I watched it and there is absolutely nothing in there that shows the software doing something other than what it was programmed to do.

Until people are ready to reveal their source code there is no way of proving what a system was programmed to do.

Quote: DRich

Quote: BTLWI

Quote: DRich

Quote: darkoz

DRich appears to be saying that software can't fail because it's created by humans therefore it's the human programmers that failed.

I am saying that software can only do what it is told to do, nothing else.
link to original post

Wrong.

Go to 16:30 and listen till 20:20
link to original post

I watched it and there is absolutely nothing in there that shows the software doing something other than what it was programmed to do.

Until people are ready to reveal their source code there is no way of proving what a system was programmed to do.
link to original post

Most although certainly not all AI source codes are open source and GitHub available.

Please tell me you know what GitHub is.

Quote: darkoz

Quote: DRich

Quote: BTLWI

Quote: DRich

Quote: darkoz

DRich appears to be saying that software can't fail because it's created by humans therefore it's the human programmers that failed.

I am saying that software can only do what it is told to do, nothing else.
link to original post

Wrong.

Go to 16:30 and listen till 20:20
link to original post

I watched it and there is absolutely nothing in there that shows the software doing something other than what it was programmed to do.

Until people are ready to reveal their source code there is no way of proving what a system was programmed to do.
link to original post

Most although certainly not all AI source codes are open source and GitHub available.

Please tell me you know what GitHub is.
link to original post

Github?? Is that some sort of gift exchange website?

Yes, Github is a Microsoft program now. I have been using Github since about 2010. I don't use it often but definitely I download source code a few times a year from there.

I've been meaning to weigh in on this topic but have not had the free time until now so here goes....

First, since DRich and Darkoz have been making an issue of who does, or does not, have the expertise to voice an informed opinion I'll state my background and let them decide. The first program I wrote was my senior year in HS (1969) and was intended to translate German into English. The program repeatedly crashed the computer (an IBM 1130) and I got a C in German. I did however learn a hell of a lot about programming. I retired in early 2023. In between I wrote a lot of code in the lots of languages but the last 30+ years was all for the DOD doing R&D on cognitive architectures for intelligent agents working in dynamic and noisy threat environments. In layman's terms: SKYNET. Having said that:

Quote: darkoz

When experts get things wrong, they get things wrong

I have never claimed to be an expert in AI or anything else. Whenever I was introduced at meetings as "the expert in XYZ" I would correct them and point out that experts claim to know all the answers about a field whereas a specialist makes the bulk of their mistakes in that field. I myself am a generalist with my mistakes spread far and wide over many areas including AI/ML.

Quote: darkoz

AI is a broad term referring to learning computers.

NO. That is Machine Learning (ML) which is a major subset of AI but not the entire field. As to what "AI" means, the answer is it means different things to different folks at different times. Personally I avoid the term and try to be specific about the tech I am referencing (i.e., Evolutionary Programming, Reinforcement Learning, etc.). If, however, you want a definition, I think the one posted by AxelWolf is close enough for general usage.

A lot of the back-and forth between DarkOz and DRich seems to be whether or not AI/ML software is actually doing anything "original". To say there is disagreement on this by the experts specialists big names in AI is an understatement. Check out an article in MIT's Tech Review that is essentially a history of name-calling and insults. Imagine you're at a major AI conference and it's lunchtime. You would expect to see a bunch of ivory tower types having having serious and sedate discussions while they eat. Instead what you see is the food-fight scene from Animal House. From the MIT artiicle:

Quote:

“It’s offensive to some people to suggest that human intelligence could be re-created through these kinds of mechanisms. People have strong-held beliefs about this issue—it almost feels religious. On the other hand, there’s people who have a little bit of a God complex. So it’s also offensive to them to suggest that they just can’t do it.”

The bottom line is DRich and DarkOz can both find "experts" who will strongly support their POV on this. Personally, I don't care one way or another. I will, however, correct DRich on one point:

Quote: DRich

If you understand computers and processors everything is derived from four basic logic gates. "AND", "OR", "NAND". and "NOR". Those gates determine how a processor works, The truth is everything could be derived from just the "NAND" logic gate but would be less efficient that way. At the university we had to design a simple processor made only with "NAND" gates.

That is no-longer strictly true. It's my understanding that some of the new neuromorphic chip technologies take an analog-based approach to emulate the neurons in our brains and that these do not use logic gates in their core ANNs. Also:

Quote: DRich

Ai never gives incorrect information, it only returns exactly what the programmer programmed it to do. Bad programmers equal bad code which results in bad answers (but not incorrect as a computer can only do what it is told to do).
If I write a computer program that outputs 1+1=3, the computer gave the correct output given a potentially flawed input.
link to original post

This interpretation of the meaning of "incorrect" is so far off from being meaningful that I suspect DRich is intentionally trolling DarkOz. If anyone working on one of my projects actually thought that, they would be invited to seek employment elsewhere.

Quote: Archvaldor1

1. You can detect card counting with statistical analysis. The tech has existed for years. It is nothing to do with AI.

So what? I can hike from NYC to WDC. The necessary tech (i.e., shoes) has been around for millennia. But taking AMTRAK is faster and more comfortable. Bottom line is whether or not supplementing statistics with "AI" will reduce false negatives/positives and/or save money.

Finally, on the original question asked by BTLWI, my take is that if the casinos really want to crack down and spend the necessary money, there is a hell of a lot of things that can be done to make life much more difficult (but not necessarily impossible) for APs regardless of the game they target. Lots of techniques that could be lifted from the world of counter-terrorism and counter-narcotics. For example, integration of video with SIGINT (e.g., cell ph SIM and IMSI numbers) and bio-metrics (e.g., gait identification). Based on what I saw at the one G2E I attended, my hunch is the casino industry is too slow-moving and unimaginative to go that route so you probably can rest easy.

Quote: TumblingBones

I've been meaning to weigh in on this topic but have not had the free time until now so here goes....

First, since DRich and Darkoz have been making an issue of who does, or does not, have the expertise to voice an informed opinion I'll state my background and let them decide. The first program I wrote was my senior year in HS (1969) and was intended to translate German into English. The program repeatedly crashed the computer (an IBM 1130) and I got a C in German. I did however learn a hell of a lot about programming. I retired in early 2023. In between I wrote a lot of code in the lots of languages but the last 30+ years was all for the DOD doing R&D on cognitive architectures for intelligent agents working in dynamic and noisy threat environments. In layman's terms: SKYNET. Having said that:

Quote: darkoz

When experts get things wrong, they get things wrong

I have never claimed to be an expert in AI or anything else. Whenever I was introduced at meetings as "the expert in XYZ" I would correct them and point out that experts claim to know all the answers about a field whereas a specialist makes the bulk of their mistakes in that field. I myself am a generalist with my mistakes spread far and wide over many areas including AI/ML.

Quote: darkoz

AI is a broad term referring to learning computers.

NO. That is Machine Learning (ML) which is a major subset of AI but not the entire field. As to what "AI" means, the answer is it means different things to different folks at different times. Personally I avoid the term and try to be specific about the tech I am referencing (i.e., Evolutionary Programming, Reinforcement Learning, etc.). If, however, you want a definition, I think the one posted by AxelWolf is close enough for general usage.

A lot of the back-and forth between DarkOz and DRich seems to be whether or not AI/ML software is actually doing anything "original". To say there is disagreement on this by the experts specialists big names in AI is an understatement. Check out an article in MIT's Tech Review that is essentially a history of name-calling and insults. Imagine you're at a major AI conference and it's lunchtime. You would expect to see a bunch of ivory tower types having having serious and sedate discussions while they eat. Instead what you see is the food-fight scene from Animal House. From the MIT artiicle:

Quote:

“It’s offensive to some people to suggest that human intelligence could be re-created through these kinds of mechanisms. People have strong-held beliefs about this issue—it almost feels religious. On the other hand, there’s people who have a little bit of a God complex. So it’s also offensive to them to suggest that they just can’t do it.”

The bottom line is DRich and DarkOz can both find "experts" who will strongly support their POV on this. Personally, I don't care one way or another. I will, however, correct DRich on one point:

Quote: DRich

If you understand computers and processors everything is derived from four basic logic gates. "AND", "OR", "NAND". and "NOR". Those gates determine how a processor works, The truth is everything could be derived from just the "NAND" logic gate but would be less efficient that way. At the university we had to design a simple processor made only with "NAND" gates.

That is no-longer strictly true. It's my understanding that some of the new neuromorphic chip technologies take an analog-based approach to emulate the neurons in our brains and that these do not use logic gates in their core ANNs. Also:

Quote: DRich

Ai never gives incorrect information, it only returns exactly what the programmer programmed it to do. Bad programmers equal bad code which results in bad answers (but not incorrect as a computer can only do what it is told to do).
If I write a computer program that outputs 1+1=3, the computer gave the correct output given a potentially flawed input.
link to original post

This interpretation of the meaning of "incorrect" is so far off from being meaningful that I suspect DRich is intentionally trolling DarkOz. If anyone working on one of my projects actually thought that, they would be invited to seek employment elsewhere.

Quote: Archvaldor1

1. You can detect card counting with statistical analysis. The tech has existed for years. It is nothing to do with AI.

So what? I can hike from NYC to WDC. The necessary tech (i.e., shoes) has been around for millennia. But taking AMTRAK is faster and more comfortable. Bottom line is whether or not supplementing statistics with "AI" will reduce false negatives/positives and/or save money.

Finally, on the original question asked by BTLWI, my take is that if the casinos really want to crack down and spend the necessary money, there is a hell of a lot of things that can be done to make life much more difficult (but not necessarily impossible) for APs regardless of the game they target. Lots of techniques that could be lifted from the world of counter-terrorism and counter-narcotics. For example, integration of video with SIGINT (e.g., cell ph SIM and IMSI numbers) and bio-metrics (e.g., gait identification). Based on what I saw at the one G2E I attended, my hunch is the casino industry is too slow-moving and unimaginative to go that route so you probably can rest easy.
link to original post

Thank you so much for your response.

I certainly consider DRich the expert in this area who knows a lot more than me.

Which is why it was so frustrating to be arguing things I was certain I was correct on.

As for expert testimony, every lawyer knows for every expert the plaintiffs have there is an expert for the defense refuting everything.

So expert testimony has always been tricky.

I did feel DRich might be trolling but he has never seemed that sort and I didn't want to make the accusation. I'm glad now the thought wasn't completely out of left field.

Thanks again for your expert analysis.

Quote: darkoz

I did feel DRich might be trolling but he has never seemed that sort and I didn't want to make the accusation. I'm glad now the thought wasn't completely out of left field.

Thanks again for your expert analysis.
link to original post

I definitely wasn't trolling, if any one of you can show me any computer software that does anything that it wasn't programmed to do, I may acquiesce. Yes, we will need to see the source code to prove it.

It seems to me that the dispute between Drich and drakoz is of no fundamental importance for the methods (although DRich logic seems more correct to me).


What difference does it make what exactly the casino uses to identify counters, special software, AI, or just Excel, the whole point comes down to identifying a gambler whose bet amount on a positive EV in theoretical money is greater than on a negative EV. In any case, in a casino the final decision is made by the person holding the position of analyst.

Quote: DRich

if any one of you can show me any computer software that does anything that it wasn't programmed to do,

I don't know how it's even possible for software to do something it isn't programmed to do. If it's learning, didn't someone or something programed it to do that?

Quote: AxelWolf

Quote: DRich

if any one of you can show me any computer software that does anything that it wasn't programmed to do,

I don't know how it's even possible for software to do something it isn't programmed to do. If it's learning, didn't someone or something programed it to do that?
link to original post

Axel, this is kinda why we were going at it.

Yes, a programmer programs the AI to let's say learn about dogs so upon request you will get images of dogs.

The programmer expects the AI to properly learn about dogs.

The AI comes across weird artwork (think the 3 headed Cerberus).

You ask for an image of a doberman pinscher and get a 3 headed doberman

My argument is because it's a learning tool it can learn from false data and certainly the programmers didn't program the AI to learn from false data. It's a learning mechanism and can learn incorrectly.

That's all I am saying. So when the AI gives answers that doesn't make sense it's not necessarily the fault of the programmer and I think it's safe to say the program itself supplied the wrong answer.

I'm not saying it's a living organism or anything like that.

Ultimately these kinks will get ironed out but with millions of data scrubbed by the AI for answers there are going to be consequences unintended by the original programmers. In essence, it's going to do things it wasn't programmed for.

Quote: TumblingBones

Quote: Archvaldor1

1. You can detect card counting with statistical analysis. The tech has existed for years. It is nothing to do with AI.

So what? I can hike from NYC to WDC. The necessary tech (i.e., shoes) has been around for millennia. But taking AMTRAK is faster and more comfortable. Bottom line is whether or not supplementing statistics with "AI" will reduce false negatives/positives and/or save money.

Yeah, it won't.

The vast majority of counters are detectable through bet variation. You look at how much money they are betting into positive counts. It shows up very quickly even with deep cover. Random players do not bet more into positive counts. They do not know what the count is. You can write ten lines of code which separate regular players from counters within a short timeframe.

Maybe a hypothetical AI could make inferences about a player's age, ethnicity, could assess the expectation gained from playing departures or decision-making frequency etc. None can do this currently with any degree of accuracy, If they could then you make it a lot easier for card counters who work out how the program works to evade detection. For what? Detecting a counter within 172 hands rather than 161? It isn't worth it and it isn't ever going to be.

The subtext I get from what you are saying is "AI has super-powers which work in some vague and non-specific way". For you to code something effective you need to be able to break it down into actionable steps and you haven't really done that yet so I'm very sceptical.