New Math for Artificial Neurons

spectrum.ieee.org...

Floating-point processors in FPGAs make for artificial neurons quick enough to communicate with real ones . 2 September 2009—Computer hardware that can simulate brain function could bring greater understanding of how the brain develops and works and may even lead to ways of repairing brain damage caused by injury or disease. But because the activity of each neuron is so complex, it’s been difficult to simulate with great detail or in real time.
Now researchers at the University of Bristol, in England, say they’ve come up with a method to model neural activity with enough detail and speed for living cells to talk to synthetic neurons. ”We want to create an artificial brain that can communicate with a real brain,” says José Nuñez-Yañez, a senior lecturer in electronic engineering in Bristol’s Centre for Communications Research.

WOW
If this technology is developed can you imaging the things that they could do ?

reply to post by Max_TO


Opens wide the possibility for cybernetics if they can pull it off. Awesome.

the briits rock ;p
2nd

ill be baacckk lol

"Floating-point processors in FPGAs make for artificial neurons quick enough to communicate with real ones."

I must be getting old when I actually understand what this guy is saying.

FPGA or Field Programmable Gate Arrays allow one to create
"Soft Processors" that are essentially floating point math functions
coded in C/C++ and then compiled to a hardware description language
such as Verilog that allows one to create VERY FAST versions of
the original software code burned to chip-based hardware

The British scientists are trying to create array processors that work
on very large sets of numbers used in Boolean-based (i.e. YES, NO, MAYBE)
pattern matching hardware that have the ability to do advanced
acoustic and vision recognition for face and object recognition,
text-based pattern searches for advanced Google-like text searches
applied to images and sound such as "Find all Red Corvettes in list of Photos"
or "Find all sounds of Birds in list of audio files"

Once you have advanced pattern matching you can begin to
do proper Strong Artificial Intelligence. It is also not too far away
(i.e. about 5 or more years) to create full 3D processors so that
we can put hundreds of THOUSANDS of Pentium processors
in a 3D stacked box about the size of a box of cereal that has enough
horsepower to quite out-perform the approximate 100 Petaflop
(i.e. 100 Quadrillion Floating Point Operations) horsepower of the
human brain. Again we're only 5 or so years away from that
sort of horsepower. Combine a few hundred of the "FPGA Cereal Boxes"
together and we have ourselves a Super-Intelligence that goes
WAY BEYOND our ability to keep up....!!!!!!

: BUMP

Just wanted to bump this report , so there , BUMP

A hypothetical question here:
What if you could incrementally replace human neurons with artificial neurons. Would you then have a form of immortality in an artificial cyborg that housed your consciousness?

Would your spirit possess the machine?

If so, and these machines one day rule the earth, the next step is for malevolent spirits, demons if you will to possess our machines.

(We had a Buick once that was definitely demon possessed.)

Now that's freaking cool. StargateSG7 your post was very informative. This opens up some huge doors...SevenThunders, imo, if these artificial neurons could replicate everything an organic neuron could...then that's all you need to house the "soul"...the soul is in fact the result of the brains complex functioning...the same thing that gives us self awareness, gives us a soul. It doesn't matter whether the "hardware" is organic or artificial...it's whether you can make artificial hardware that can do the same things organic hardware can. Which is an extremely hard thing to do, but not impossible. I believe in my life time I will see true A.I and advanced cybernetics become a reality. Robots with self awareness will walk among humans. I believe this is a good thing and it's wrong to discriminate them, because we fear them. They are sentinel beings who deserve respect and rights like any other being. Although, one still can't help fear them, when you look at humanity, and the evil we are capable of. It's scary to think what an evil android (or multiple) would be capable of...

[edit on 10/9/09 by CHA0S]

I wonder if the IEEE journalist got this story a little bit wrong? Unlikely but possible.

Whilst it is certainly possible to run muliple virtual processors in parallel in a single FPGA as StargateSG7 describes, I would have thought that emulation of biological systems would be better in an analog FPGA or FPAA which is the same thing but works in the analog realm.

Edit: See EETimes Article for more info.

Edit2: Looks like the journalist was correct, it is just the university that needs to keep up with the times!


[edit on 10/9/2009 by LightFantastic]

Originally posted by StargateSG7
It is also not too far away(i.e. about 5 or more years) to create full 3D processors so that we can put hundreds of THOUSANDS of Pentium processors in a 3D stacked box about the size of a box of cereal that has enough
horsepower to quite out-perform the approximate 100 Petaflop
(i.e. 100 Quadrillion Floating Point Operations) horsepower of the
human brain. Again we're only 5 or so years away from that
sort of horsepower. Combine a few hundred of the "FPGA Cereal Boxes"

This made me laugh a bit. You do know how close we've come to the speed of the human brain right? Yeah not even close! I don't care what kind of technology we have, 100 Quadrillion FPO is beyond us and will be for quite some time yet. I do agree that we will start getting faster multicore processors though.

Originally posted by DaMod

Originally posted by StargateSG7

This made me laugh a bit. You do know how close we've come to the speed of the human brain right? Yeah not even close! I don't care what kind of technology we have, 100 Quadrillion FPO is beyond us and will be for quite some time yet. I do agree that we will start getting faster multicore processors though.

When attempting to emulate neural systems with digital systems then yes, we are miles off. In the analog domain however we may get closer faster thank you think.

Also the 100 Quadrillion FLOPS is an estimate of the performance required for a digital system to emulate the human brain, not an estimate of the human brains performance.

I would be happy to acheive 1 FLOPS!

[edit on 10/9/2009 by LightFantastic]