movie question
August 12, 2009 9:19 PM
Is the Greco Player Tracker featured in Oceans 13 possible?
For those who don't know what the Greco Player Tracker is; In Ocean's 13 the casino (Shuffle Royal) Danny and the boys are planning to rob supposedly has in place this highly sophisticated security (AI) computer system called the Greco Player Tracker that tracks players emotions, pupil dilation, heart beat, body temp in real time in a server field of exabytes and determines whether a player's reaction to a win is legit or fake. I know it's a movie and things in movies most of the time can never happen in real life but is this computer system even possible?
For those who don't know what the Greco Player Tracker is; In Ocean's 13 the casino (Shuffle Royal) Danny and the boys are planning to rob supposedly has in place this highly sophisticated security (AI) computer system called the Greco Player Tracker that tracks players emotions, pupil dilation, heart beat, body temp in real time in a server field of exabytes and determines whether a player's reaction to a win is legit or fake. I know it's a movie and things in movies most of the time can never happen in real life but is this computer system even possible?
Well, the field of security biometrics grows apace. One never knows what might be technically possible with something like facial expression analysis. But this isn't really the realm of AI so much as neural networks, and training one with the proper heuristics would be quite difficult even when limited to a handful of research subjects.
The real problem with any deployment in the wild will be massive numbers of false positives. At best it could only flag certain players for further scrutiny.
I don't think that such a network would really perform better than the network of human security personnel who have experience with how legitimate and illegitimate gamblers look, act, and perform.
I just took it as an entertaining Macguffin to get us into the mechanics of e.g. climbing through ductwork.
posted by dhartung at 12:10 AM on August 13, 2009
The real problem with any deployment in the wild will be massive numbers of false positives. At best it could only flag certain players for further scrutiny.
I don't think that such a network would really perform better than the network of human security personnel who have experience with how legitimate and illegitimate gamblers look, act, and perform.
I just took it as an entertaining Macguffin to get us into the mechanics of e.g. climbing through ductwork.
posted by dhartung at 12:10 AM on August 13, 2009
Wouldn't varying levels of intoxication throw this off?
posted by mkultra at 6:48 AM on August 13, 2009
posted by mkultra at 6:48 AM on August 13, 2009
But this isn't really the realm of AI so much as neural networks
off-topic nitpick: "AI" and "neural networks" aren't exclusive of each other---neural networks are a technique that finds use in AI applications and elsewhere. Furthermore, as someone who does "AI" research, loosely construed, I would attack the "cheating gambler" computer vision problem with my weapon of choice: Bayesian probabilistic approaches that aren't very usefully characterized as "neural networks". There are other choices as well.
I agree that heartbeat measurement would be exceedingly difficult---probably impossible with cameras alone---and that the win rate tracking b1tr0t talks about would probably get you most of the way there, once you worked out how to reliably track individuals throughout the casino. Analysis of posture, facial expression, gestures, and hey, buffet food choice, why not, may also provide some very vague hints. (The nice thing about Bayesian approaches in this case is that they provide a relatively principled and robust framework for the designer to characterize and model more explicitly the relationship between the factors they are measuring and the circumstance they are trying to detect, even when the relationship is very weak. This takes some of the sting out of the "heuristics design" dhartung is talking about. I would also venture that Bayesian models are better at saying "I don't know" than neural nets are.)
I agree, though, that trained personnel would probably do better.
posted by tss at 7:08 AM on August 13, 2009
off-topic nitpick: "AI" and "neural networks" aren't exclusive of each other---neural networks are a technique that finds use in AI applications and elsewhere. Furthermore, as someone who does "AI" research, loosely construed, I would attack the "cheating gambler" computer vision problem with my weapon of choice: Bayesian probabilistic approaches that aren't very usefully characterized as "neural networks". There are other choices as well.
I agree that heartbeat measurement would be exceedingly difficult---probably impossible with cameras alone---and that the win rate tracking b1tr0t talks about would probably get you most of the way there, once you worked out how to reliably track individuals throughout the casino. Analysis of posture, facial expression, gestures, and hey, buffet food choice, why not, may also provide some very vague hints. (The nice thing about Bayesian approaches in this case is that they provide a relatively principled and robust framework for the designer to characterize and model more explicitly the relationship between the factors they are measuring and the circumstance they are trying to detect, even when the relationship is very weak. This takes some of the sting out of the "heuristics design" dhartung is talking about. I would also venture that Bayesian models are better at saying "I don't know" than neural nets are.)
I agree, though, that trained personnel would probably do better.
posted by tss at 7:08 AM on August 13, 2009
...the Greco Player Tracker that tracks players emotions
No. Humans can't even reliably tell what another human's current emotional state is. Machines don't stand a chance.
pupil dilation
Probably. That just needs a camera.
heart beat
No. You'd need to be in physical contact to track a heartbeat. I suppose jn a silent, echo dampened room you could tell someone's heartbeat by sound but in a crowded casino full of hundreds of people: No.
body temp
Yes. You can buy wireless optical thermometers at Amazon for under $100, easy.
in real time in a server field of exabytes and determines whether a player's reaction to a win is legit or fake.
No. Even if you measure this crap, you have no baseline for the individual to compare it to. There's far too many variables to do anything useful with the data. I have no idea why this would take exabytes of data.
It's also not clear if your question is: "Can you reliably measure all this junk?" or "Given you can reliably measure all this junk, can you do anything useful with it in a casino?"
posted by chairface at 1:39 PM on August 14, 2009
No. Humans can't even reliably tell what another human's current emotional state is. Machines don't stand a chance.
pupil dilation
Probably. That just needs a camera.
heart beat
No. You'd need to be in physical contact to track a heartbeat. I suppose jn a silent, echo dampened room you could tell someone's heartbeat by sound but in a crowded casino full of hundreds of people: No.
body temp
Yes. You can buy wireless optical thermometers at Amazon for under $100, easy.
in real time in a server field of exabytes and determines whether a player's reaction to a win is legit or fake.
No. Even if you measure this crap, you have no baseline for the individual to compare it to. There's far too many variables to do anything useful with the data. I have no idea why this would take exabytes of data.
It's also not clear if your question is: "Can you reliably measure all this junk?" or "Given you can reliably measure all this junk, can you do anything useful with it in a casino?"
posted by chairface at 1:39 PM on August 14, 2009
This thread is closed to new comments.
Also, remotely sensing some of those things without the subject being aware that he was being monitored would be extremely challenging. Pupil dilation is possible, albeit difficult if the person has dark brown eyes. Body temp measurement is possible, but not necessarily as rapidly or as accurately as would really be needed for this kind of thing. Measuring heart rate? I don't see how that could be done at all, in a crowded and noisy casino. Even measuring respiration reliably would be tough.
posted by Chocolate Pickle at 10:04 PM on August 12, 2009