Artificial Intelligence

[WALMART]

Be prepared to find robots involved in driving accidents and court cases.

http://www.newscientist.com/article/mg21628925.100-brainlike-chip-outstrips-normal-computers.html

 

[Tyrinth]

With news like this I'm always torn between being incredibly interested and incredibly apprehensive. It will, however, be interesting to see how artificial intelligence technology develops in the near-future.

 

[sprinkles]

Analog FTW.

We've been stuck with digital for so long that some people are starting to think that it's the only way to compute something electronically.

It's hard (practically impossible?) to get intelligence out of 1's and 0's so the key is - not surprisingly - to stop trying to use 1's and 0's.

 

[Rad3k]

I think it's entirely possible to create a good AI with only digital circuits, you'd just have to choose reasonable precision for representing real values. Doing it with analog could be more efficient though, as it requires less emulation and just lets the laws of physics calculate stuff. But there's a catch - analog circuits are much more sensitive to electromagnetic interference. What about that?

On another note - these "brain chips". I wonder if it will be possible in our lifetime to replace parts of your own brain with these. It would be awesome - you could then replace your entire biological body with a robotic one, and effectively become an android. I would probably volunteer to be a test subject, if there was no other way to get that.

 

[sprinkles]

I think it's entirely possible to create a good AI with only digital circuits, you'd just have to choose reasonable precision for representing real values. Doing it with analog could be more efficient though, as it requires less emulation and just lets the laws of physics calculate stuff. But there's a catch - analog circuits are much more sensitive to electromagnetic interference. What about that?

Interference is part of the real deal anyway.

I'm actually referring to something that could truly end up being dynamically intelligent (and perhaps self aware) rather than something that just emulates really complex behaviors on an abstracted layer.

On another note - these "brain chips". I wonder if it will be possible in our lifetime to replace parts of your own brain with these. It would be awesome - you could then replace your entire biological body with a robotic one, and effectively become an android. I would probably volunteer to be a test subject, if there was no other way to get that.

I don't know about my life time but maybe someone else's...

 

[INTP]

Pretty nice, but they need a way to grow these artificial neurons directly on the chip when ever needed similarily to how humans grow neurons, also they need billions, 400 isnt really enough for anything else than a brains of a bee.

Ps. Artificial intelligence is so 2011, artificial consciousness is what needs to be done, as its needed for true intelligence.

 

[Cellmold]

Cant be any worse than the so called 'actual intelligence' we have to put up with currently. And if a Skynet situation evolves from this; at least then it's still technically us destroying ourselves.

 

[sprinkles]

Cant be any worse than the so called 'actual intelligence' we have to put up with currently. And if a Skynet situation evolves from this; at least then it's still technically us destroying ourselves.

I don't see that happening. Strong AI wouldn't be useful for computing - it will be too smart to just do what you tell it to, not very different from a human expert. May as well stick with humans. Strong AI is just for study and academic purposes.

Weak AI is what we already have, and is "dumb smart". It can run certain types of processes more efficiently than a human, but won't think for itself or disobey orders. These are usually specialized AI as well - it's only smart in the field it is designed for, such as driving a car through a city autonomously while obeying the traffic laws. In the end it isn't really intelligent and has already been done (a robot has driven through a desert, and another one successfully driven through a city)

 

[Cellmold]

I don't see that happening. Strong AI wouldn't be useful for computing - it will be too smart to just do what you tell it to, not very different from a human expert. May as well stick with humans. Strong AI is just for study and academic purposes.

Weak AI is what we already have, and is "dumb smart". It can run certain types of processes more efficiently than a human, but won't think for itself or disobey orders. These are usually specialized AI as well - it's only smart in the field it is designed for, such as driving a car through a city autonomously while obeying the traffic laws. In the end it isn't really intelligent and has already been done (a robot has driven through a desert, and another one successfully driven through a city)

My comment on actual intelligence was meant towards people .