It's interesting to apply these kind of calculations to the human brain, to understand the scale of the thing.
From TFA, to simulate a single cortical column:
"You need one laptop to do all the calculations for one neuron," he said. "So you need ten thousand laptops." Instead, he uses an IBM Blue Gene machine with 10,000 processors.
Okay, so there's about 20 billion neurons in the neocortex alone, about 2 million cortical columns then assuming 10,000 neurons in each. Even the mighty Moore's law from 2005 (Blue Brain's construction) -> 2019 isn't going to cover an increase of 2 million times for the kind of supercomputer you can construct at that time. So it's already relying on things like GPGPU to supercede Moore's law.
Storage is another problem. Even simple representations of a neuron, its position, state, and all of its connections get huge when you multiply them by 20 billion.
Blue Brain is trying to do a chemically accurate simulation of the brain, which as stated could very well be useful for testing new drugs and so on. But I don't expect this kind of heavy simulation to be the first thing to gain conciousness-like properties. We need to use the data generated by Blue Brain to build simpler models of neurons and cortical columns that behave in comparable ways, and then construct our artificial brain from those. Of course connectivity is another issue..
I have to make my wife cookies every year on white day. I'm not sure how she'd react if presented with a Desperate Dan style cow pie with horns poking out instead.:)
Slashdot Mirror
It's interesting to apply these kind of calculations to the human brain, to understand the scale of the thing.
From TFA, to simulate a single cortical column:
"You need one laptop to do all the calculations for one neuron," he said. "So you need ten thousand laptops." Instead, he uses an IBM Blue Gene machine with 10,000 processors.
Okay, so there's about 20 billion neurons in the neocortex alone, about 2 million cortical columns then assuming 10,000 neurons in each. Even the mighty Moore's law from 2005 (Blue Brain's construction) -> 2019 isn't going to cover an increase of 2 million times for the kind of supercomputer you can construct at that time. So it's already relying on things like GPGPU to supercede Moore's law.
Storage is another problem. Even simple representations of a neuron, its position, state, and all of its connections get huge when you multiply them by 20 billion.
Blue Brain is trying to do a chemically accurate simulation of the brain, which as stated could very well be useful for testing new drugs and so on. But I don't expect this kind of heavy simulation to be the first thing to gain conciousness-like properties. We need to use the data generated by Blue Brain to build simpler models of neurons and cortical columns that behave in comparable ways, and then construct our artificial brain from those. Of course connectivity is another issue..
No, it's gameplay. There's something called willing suspension of disbelief.
Well I'm not having anyone staring in disbelief at my willy suspension!
I have to make my wife cookies every year on white day. I'm not sure how she'd react if presented with a Desperate Dan style cow pie with horns poking out instead. :)
Hawkins' published a book before this was implemented in code called "On Intelligence". You could do worse than starting by reading through that.
He's also done some lectures available on Google Video.
There should be news servers for the rest of us with binary attachments blocked.
text.news.ntlworld.com has materialized recently for just this purpose.