Supercomputer Question

[Gnexus]

I was watching person of interest, and I wondered what kind of computing power would you need to monitor all the camera's etc etc that it does in the show. That is of course, if we could even build a AI that would be smart enough to be able to do that, and predict crimes. Currently our computers only understand yes and no, black and white. Possibly isn't something they would understand. Let's say for moment they could though. How large would the server farm have to be to achieve this ? The size of New York City ? Bigger ?  

[Electronics Wizardy]

Depends on what you want to analize for. We probably already have enough power.

[FloRolf]

I guess the Titan would even be enough for that.

Just consider that, nvidias Driver module that they want to put in automated cars can already easily detect cats and Dogs, other cars and whatever.

Pretty sure if we had the right Software it would work pretty easily.

[Dash Lambda]

Monitoring Cameras: A couple of server racks.

AI: I'll get back to you on that when we actually have one.

 

At the moment, it looks like a functional universal neural network could be run on a modestly sized supercomputer array, the problem is programming. Half the battle for self-driving cars is just teaching the network, and that's a far more systematic and robotic process than criminology.

[LukeTim]

7 hours ago, Gnexus said:

I was watching person of interest, and I wondered what kind of computing power would you need to monitor all the camera's etc etc that it does in the show. That is of course, if we could even build a AI that would be smart enough to be able to do that, and predict crimes. Currently our computers only understand yes and no, black and white. Possibly isn't something they would understand. Let's say for moment they could though. How large would the server farm have to be to achieve this ? The size of New York City ? Bigger ?  

I can't see how that is true at all.

What do you mean by that?

[positivePXL]

IBM just open-sourced their quantum computer.

 

*wink* *wink*

[Gnexus]

10 hours ago, LukeTim said:

I can't see how that is true at all.

What do you mean by that?

I've yet to hear of a computer that can predict patterns of behaviour as a AI, on it's own. Computers can do it now, but we have to give them the data to do it. Having a AI do it, is different. It has to think on it's own and study human patterns, and habits.

 

As for what it would monitor ? I would say phone calls, both landline and cell phone, as well as the camera's all in a city the size of New York. 

 

[LukeTim]

8 hours ago, Gnexus said:

I've yet to hear of a computer that can predict patterns of behaviour as a AI, on it's own. Computers can do it now, but we have to give them the data to do it. Having a AI do it, is different. It has to think on it's own and study human patterns, and habits.

 

As for what it would monitor ? I would say phone calls, both landline and cell phone, as well as the camera's all in a city the size of New York. 

 

What has predicting patterns of behaviour got to do with the statement:

Quote

Currently our computers only understand yes and no, black and white. Possibly isn't something they would understand.

This is what I was confused about. Since when did computers only "understand" yes and no?

Aside from the fact that computers operate on numbers (not just individual bits) which can have many different values, Fuzzy logic has existed for half a century now.

[Gnexus]

It was a assumption. I see now, I was wrong.

[2FA]

2 hours ago, Gnexus said:

It was a assumption. I see now, I was wrong.

You can make a program that will hypothetically "think" about it and not just see it as yes or no. There is currently simple AIs developed with much more complex AI in development. There is also the existence of early quantum computers which are supposed use more than just a 0 and 1. I'm not going to pretend to understand them but they will likely be key for future of AI.