Slashdot Mirror
A Skeptical Reaction To IBM's Cat Brain Simulation Claims
kreyszig writes "The recent story of a cat brain simulation from IBM had me wondering if this was really possible as described. Now a senior researcher in the same field has publicly denounced IBM's claims."
More optimisticaly, dontmakemethink points out an "astounding article about new 'Neurogrid' computer chips which offer brain-like computing with extremely low power consumption. In a simulation of 55 million neurons on a traditional supercomputer, 320,000 watts of power was required, while a 1-million neuron Neurogrid chip array is expected to consume less than one watt."
Think about it. Think about it like a cat.
Somehow my pet parrot now seems oddly... delicious. :O
Wouldn't power consumption grow more than linearly with neuron count? I would think the number of connections is the dominant factor - so the comparison of two data points of power consumption vs neuron count is meaningless.
All those neurons using less than 1 watt?
I know some people like that.
No brain, no pain.
If you have custom silicon to do each neuron then you are going to be hugely more power efficient that a general purpose processor simulating a neuron in software. There is nothing new there and anyone who thinks otherwise is just clueless. Given IBM have the facilities and resources to fabricate some custom silicon I fail to see the issue.
From the original FA: "The simulation, which runs 100 times slower than an actual cat's brain, is more about watching how thoughts are formed in the brain and how the roughly 1 billion neurons and 10 trillion synapses in a cat's brain work together."
So the most bad-ass computer simulation, assuming it worked, which this guy is saying it probably didn't, was still 100 times slower than a real cat's brain. A real cat's brain also fits inside a tiny furry space the size of a baseball... and it runs on a once-daily small bowl of cat food. We have a long ways to go.
So according to this guy rant letter, the "cat-brain simulation" was nothing more than the simulation of a ANN wiht X number of neurons with X equal to the average number of neurons in a cat.
However, it seems the /complexity/ of the simulated neurons is not remotely similar to that of the neurons of a real cat.
With that view, yes it seems less breakthrough. The experiment reminds me of AI researchers that thought that we could get intelligent machines using a brute-force kind of approach; this by adding /enough/ knowledge-rules, /enough/ processing power, etc...
Ubuntu is an African word meaning 'I can't configure Debian'
This IBM announcement was just ridiculous. To cite only one argument, the brain does not consist only of neurons. It contains at least as many other cells which are also involved in signal processing. Mohda would be laughed at in any neuroscience conference and he certainly doesn't help the cause of theoreticians in the neuroscience field by making such stupid announcements. Eugene Izhikevich who designed the neuron model being used for these simulations had a PNAS paper not too long ago modeling the entire human brain and he did not claim that he successfully modeled the human brain. Plus no one has any clue how the brain computes really so making a claim about the formation of thoughts is just nonsense.
Except that individual neurons have tens of thousands of possible connections to other neurons, and continually morph and change those connections. That's impossible to do on a rigid piece of hardware.
I'm no expert and I've just been reading the second link's project site on Stanford's page but your assertion to continually morph and change those connections seems to be mitigated by this strategy:
Neurogrid simulates six billion synaptic connections by using local analog communication, another choice motivated by cortical studies. Cortical axons synapse profusely in a local area, course along for a while, then do it again. Thus, nearby neurons receive inputs from largely the same axons, as expected from the map-like organization of cortical areas. Local wires running between neighboring silicon neurons emulate these patches, invoking postsynaptic potentials within a programmable radius. Using a patch radius of 6 lets us increase the number of synaptic connections a hundredfold—from 600 million to six billion—without increasing digital communication.
If they connect most (if not all) possible connections that the morphing/changing synaptic channels can take, then they use a sort of addressing technique and RAM strategy to continually morph and change:
Instead of hardwiring the silicon neurons together, as Mead did in his silicon retina, we softwired them by assigning unique addresses. Every time a spike occurs, the chip outputs that neuron’s address. This address points to a memory location (RAM) that holds the synaptic target’s address, or to multiple memory locations if the neuron has multiple synaptic targets. When this address is fed back into the chip, a post-synaptic potential is triggered at the target. An extremely efficient technique, as the same post-synaptic circuit serves all the synapses that neuron receives—virtual synapses! Encoding, translating, and decoding an address happens fast enough to route several million spikes per second, allowing a million connections to be made among a thousand silicon neurons. These softwires may be rerouted simply by overwriting the RAM’s look-up table, making it possible to specify any desired synaptic connectivity.
Although their page is really hard for a lay person like myself to get through, it's very informative. Having read it, I'm considerably more optimistic about the future of biological tissues and nervous systems being translated to hardware. At least people are starting back at the simple components and adding new twists.
My work here is dung.
I saw that story earlier and dismissed it for the crap that it was. I'd like to thank Henry Markram for vindicating my snap judgment with his flame email.
My research recently took me to some of Markram's work - the guy is brilliant and REALISTIC. His research goals are simple and attainable and any claims of success he has are *well* within the real world. He's incrementally worked his way up from a few neurons - the way a *real* scientist works; and to him, the simplest "brain simulation" of any sort is definitely possible, but far off in the future.
320kW / 55 = 5.818kW per million of neuro with a traditional supercomputer.
One watt per million of neuro with a Neurogrid chip array.
So if a cat's brain is 1 BILLION neurons, that would require 5818.182kW with a supercomputer and 1kW with the Neurogrid chip array.
A reduction of 5817.182kW.
"I used to think that the brain was the most wonderful organ in my body. Then I realized who was telling me this."
Faith is a willingness to accept something w/o complete proof and to act on it. Reason allows you to correct that faith.