Comment by mikeash | Hacker News Reader

Comment by mikeash | Hacker News Reader

Parent Full thread

This stuff is really interesting.

Imagine if you're playing, say, blackjack with a human dealer. After you win a hand but before you collect your winnings, you ask the dealer, "Say, would you mind if I retroactively increase my bet and collect the winnings on that?" The dealer replies, "Sure thing," and pays you accordingly.

I imagine we'd all agree that there's no harm in asking, and that the dealer's compliance is his own problem and a problem for his employer but not your problem. Why does this suddenly change when you're talking to a computer instead of a human? I'm not saying it doesn't change, but I can't entirely figure out why we approach these two scenarios so differently.

For your ATM comparison, it's not uncommon for excessively-clever bank customers to jokingly ask the teller for a million bucks, often in response to a question like, "Is there anything else I can do for you today?" Suppose you found a particularly dim teller who decided to actually hand over a million bucks when you asked. We'd blame the teller and the bank for agreeing to such a thing, not the person who asked, right? Yet when it's an ATM instead of a teller, we blame the person asking.

The gaming laws do not give the dealer the right to change the rules of the game once they have been set by the house and registered with the gaming control board.

If the dealer decides to deviate from the rules, the dealer has committed a crime. The player could also be an accomplice if he was in cohorts with the dealer.

Now in the case of the buggy slot machine, the software has been checked to NOT perform this action. If it does, it's obviously deviating from the rules as well and needs to be shut down. To avoid making the player an accomplice, the "malfunction voids all pays" rule comes into play. If the player knowingly made the slot deviate from the rules, then that falls under the definition of cheating.

The Nevada gaming laws, as I understand it, allow for payout probabilities across a time threshold. E.g. Over a 24hr period this machine pays out 95 cents on the dollar. This gives the casino the ability to make a machine pay out $1.05 dollars to $1 early in the night, then change to worse odds as players become convinced its a hot machine.

So, one could argue the 'hack' was no different than this. The probability of paying out was set to 500 to 1 for 1 hand. This is legal for the casino to do, but you turn the tables and a guy goes to jail?

Regardless, this is just more evidence that the 'house always wins' (even when it looses).

The "time threshold" is the lifetime of the machine. Not minutes, hours, or days.

Slots are fixed to a given payout percentage and then locked in with the gaming board. Traditionally this means a copy of the program is checksummed and stored offsite with the gaming commission. The officials have the right to open that machine at any time and verify that the program matches what is on record.

Casinos cannot dynamically change this payout percentage on the fly. That's a violation of the law. A casino is required to have a certain average payout across all machines, but that does not mean a particular machine needs to obey that percentage. In fact, the penny slots will be set way lower and the high-denomination slots will be set much higher. The average will hit the target.

The "hack" is theft, plain and simple. Putting aside the ability to re-award jackpots, the player found a way to change denomination of credits post-play. This is the equivalent of winning at a table game and then swapping your $1 chips for $100 chips when the dealer is not looking.

Surely it's the equivalent of winning at a table game and then convincing the dealer to swap your $1 chips for $100 chips for you.

My source to the dynamic payout ability is a past co-worker that did work in the casinos back office computer system (think big iron computing). He entertained me with stories of how the casinos manipulated the environment to manipulate you into giving them more of your money. So, I may have misunderstood some of the details. But, am curious where your certainty comes from to see who's closer to the truth here.

Regardless, when a machine pays out in a way that is against the house, my take is that these guys deserve every cent they earned. The gamblers did nothing that the casinos don't do everyday. That is, use the rules of the game to their advantage (ie. manipulate you to make a decision against your own best interest). In this case the rules of the game were broken in an extremely subtle way.

My certainty comes from nearly a decade writing software for slot machines and consulting to various gaming firms in the United States.

When you say your friend's casino "manipulated the environment" to extract more money from patrons, that's a bit vague. Is he referring to things done to patrons outside the domain of the slot machine (e.g. comped alcohol/rooms/food, awarding extra player points, show tickets, etc), or actually changing the performance of a machine while the person is playing?

Because the NV gaming law (the one that most other jurisdictions copy when doing their own) has it plain and clear in Regulation 14.040 Section 6:

"[Slots] must not automatically alter pay tables or any function of the device based on internal computation of the hold percentage"

Now back in the day, the payout percentage target was fixed in an EPROM table, which the gaming board held a copy. Things are a bit more complex now with networked slots. The games are now downloading their code from a central server. A slot floor manager can load a new game into the cabinet with a different payout percentage, but it cannot be done while a player is playing and for a fixed amount of time afterward, and then the game must go offline (and say so publicly) for another fixed amount of time after the new game loads. This time is on the order of 10 minutes or more so a player is guaranteed to know that the game is being changed if a slot floor decides to do this.

Really, a casino really doesn't need to play around with slots to make sure they earn money. The math guarantees it. The casino's job is to make sure the players come and stay as long as possible which seems like what your friend was involved with. There's serious work in collecting all that player data and trying to understand ways to maximize the handle.

> The gaming laws do not give the dealer the right to change the rules of the game once they have been set by the house and registered with the gaming control board.

I don't know about that. Don Johnson, the blackjack player who won more than $15M at the casinos, did so by having them change the rules. He made enough little changes that the odds swung into his favor, and he took the casinos to the cleaners. All valid and legal. From the wiki: He negotiated several changes to standard casino blackjack in order to gain a mathematical edge

https://en.wikipedia.org/wiki/Don_Johnson_%28gambler%29

Computers are not possessed of free will. You don't really ask them for things. You either compel them to take an action or you don't. Knowingly exploiting a glitch in a computer is like asking somebody for their wallet with a gun to their temple — despite the fact that the exchange could be described as a request and agreement to that request, you have made it impossible for the recipient to decline under the circumstances, so we classify it as something else.

I'm not convinced that free will is even a coherent concept, let alone something that humans possess.

Then a lot of the law will probably seem a little odd to you, because it is pretty solidly on one side of that debate. I'm not qualified to prove this one way or the other, but if you want to understand the law, you have to go in with the assumption that the average human being is in deliberate control of their actions under normal circumstances.

(AFAIK it is pretty hard to come up with a system of law that doesn't start with this assumption without it being either completely ineffectual or very oppressive. Whether or not free will exists, it's at least a very handy abstraction for distinguishing good actors from bad ones.)

Much of the law does seem odd to me, although I don't think that's why.

As far as I can tell, the law is based on a collection of ideas we collectively identify as "justice". In no particular order: that punishment can reform a person so they no longer commit criminal acts, that the threat of punishment can cause a person to refrain from committing criminal acts, and that punishment as revenge is just a good thing.

None of this has anything to do with "free will", whatever it is.

Volition is a huge part of deciding how we want to punish someone and even whether a specific act was criminal at all. If you did something bad of your own free will, the law probably wants to have a word with you. If you were forced into doing something bad against your will, you very well might not be punished at all, and instead the person who forced you to commit the act might be culpable because it was their volition that caused the law to be broken.

It's tough for me to discuss this, as I don't understand what free will actually is. Could you define it for me?

I come to a fork in the road, I can choose to go left or right.

A computer comes to a fork in the road, x == 5 so it goes left, it does not have free will and cannot choose to go right, it always goes left if x == 5.

We can argue "but what about determinism" all day but I'm pretty sure you understand what we mean by "free will" and are just being pedantic.

I'm sorry if you think I'm being pedantic, but I truly don't understand the concept. The more I think and talk about it, the less I understand what it's supposed to mean.

As far as I can tell, there are three possibilities:

1. The entity has no internal state. Any stimulus always produces the same response.

2. The entity is deterministic and has internal state. The same stimulus does not necessarily produce the same response. The same stimulus combined with the same internal state will always produce the same response.

3. The entity has some random factor in how it works. The same stimulus does not necessarily produce the same response, even with the same internal state.

Which one is "free will"? It's definitely not 1. It doesn't seem to be 2, from how people talk about it. But neither does it seem to be 3, as it's discussed as something more than just randomness. But what else could there be?

Before this gets any more out of hand: http://wiki.lesswrong.com/wiki/Free_will_(solution)

That appears to be discussing the problem of free will, while assuming that you already know what the problem is, and more importantly what "free will" means. My problem is that I don't know what the very concept of "free will" is, completely separate from questions of whether humans actually have it or how.

No, the concept of what free will is is actually exactly what the series of linked posts is meant to address. Read the first, http://lesswrong.com/lw/no/how_an_algorithm_feels_from_insid..., in particular. "Free will" is a term with the same properties as "sound."

If I understand that analogy then it's basically saying that "free will" is an excessively vague term that evaporates entirely when you describe what you're thinking of more specifically. Which is basically what I'm saying: I don't think it's a coherent concept by itself, and I don't know what people mean when the use the phrase.

You can think of free will as a control system. Like a PID controller.

A person performs an action, measures how the environment is changed, and modifies their behaviour to get better outcomes (for themselves or other people they care about).

Humans are a bit more sophisticated than a PID controller. While most animals, machines and nature itself are limited to blindly performing actions and measuring the result, people (and crows) can ponder what will happen IF they do something. Several steps ahead. They do not need to perform an action to anticipate the consequence.

When we say the law assumes free will, we mean that the law assumes a good control system. "If you do this we'll put you in a concrete box". The law assumes that threat will work.

This, as it turns out, is a flawed assumption. Many people do not think through their actions (stupid criminals), or they think through their actions and judge them worth the risk (e.g. weed smokers), or they have no other choice of action (e.g. prostitutes).

tl;dr: your rice cooker has free will.

I think a good example of people who exploit bugs in people to get rich are con artists, and what they do is usually considered illegal. Sure, there are various reasons why what they do feels wrong, they may have to lie and misrepresent things, they may con people (the elderly, etc.) for which the law has specific protections (France has "abus de faiblesse" -- abuse of weakness -- for exploitation of physically or psychologically vulnerable people), they may use threats... But I would be surprised if there was no instance of a con artist being successfully prosecuted for something which basically amounted to getting a responsible person to give them money through a certain sequence of legal inputs.

I am not personally sure that I believe this to be a good thing -- the flipside of laws that protect people against themselves is that they may make them less responsible, so care is needed. But I think it is not the case that exploiting bugs in people to get money would be systematically legal.

I do think there is a difference (at least in degree, maybe in nature) between computer and human victims: for computers, we feel justified in estimating that people should know that they are dealing with them, and second-guess what they are supposed to do, so that you are blamed if it can be shown that you "knowingly" make them deviate. (Laws have this interesting way of talking about unverifiable states of mind of people, but there can be clues. Did the gamblers who exploited the bug call it a "bug"? If yes, this suggests that they realized it was unintentional, so you may argue whether it is still fair or not that they exploit it, but you can't compare their situation to someone who would hit the bug in good faith.)

With respect to this difference, I think the law is also designed to account for the pragmatical fact that everyone has entrusted their lives and businesses to the dumb computers, without having the technical skill to be responsible for what they do, so some protection is afforded computers they malfunction and/or if they have been incompetently manufactured or programmed. I often wonder whether this is a good thing. (When you consider that people are responsible for computers they own or operate, you often get interesting moral consequences.)

>I think a good example of people who exploit bugs in people to get rich are con artists, and what they do is usually considered illegal.

Many of the richest people around got that way by exploiting some "bug", but no one considers them to have done anything illegal. They find a loophole in banking laws and become a billionaire, or you manipulate people psychologically into buying your products (modern marketing). We call these "business" so it's not illegal to exploit these "bugs".

To me, the difference is in the free will. The computer doesn't have a choice; the dealer does.

Reply on news.ycombinator.com