Henry Markram of EPFL’s Blue Brain Project: IBM’s Cat Brain Claim is a “HOAX”
Over at Next Big Future, BoingBoing, and many other venues, Henry Markram of the EPFL's Blue Brain Project has a comment up on the recent IBM cat brain simulation announcement.
IBM's claim is a HOAX.
This is a mega public relations stunt - a clear case of scientific deception of the public. These simulations do not even come close to the complexity of an ant, let alone that of a cat. IBM allows Mohda to mislead the public into believing that they have simulated a brain with the complexity of a cat - sheer nonsense.
Here are the scientific reasons why it is a hoax:
He also sent a letter to IBM's CTO and CCd the media.
New Zealand PC World has an article that summarizes some of the points.
IBM responded by issuing a statement:
IBM stands by the scientific integrity of the announcement on cognitive computing led by IBM in collaboration with Stanford University, University of Wisconsin-Madison, Cornell University, Columbia University Medical Center, University of California-Merced and Lawrence Berkeley National Laboratory," the IBM statement reads. "The cognitive computing team has achieved two milestones that indicate the feasibility of building a computing system that requires much less energy than today's supercomputers, and is modeled after the cognition of the brain. This is important interdisciplinary exploratory research bringing together computational neuroscience, microelectronics and neuroanatomy, and this work has been commented on favorably by others in the scientific community.
If Markram is telling the truth in his allegations (I don't know about all of them because many of the details he mentions are not addressed in the IBM paper, but some of the claims seem obviously true to me), then IBM has lost all credibility.
IBM says that it is "modeled after the cognition of the brain", but what the hell does that mean? Point neurons, like Markram says, most likely. It also seems like Modha's web page and the text of the press release are explicitly designed to further the delusion that they have created a cat-complexity brain.
"Whole Brain Emulation: a Roadmap" has a more realistic and comprehensive estimate of the complexity required to simulate a brain.
November 24th, 2009 - 19:29
Most computer scientist considers neural networks to be modeled after the brain.
November 24th, 2009 - 19:36
http://en.wikipedia.org/wiki/Neural_network
As wiki says, you have two kinds of networks
1: Artificial neural networks
2: Biological neural networks
That said IBM sounds like its marketing something.
November 24th, 2009 - 19:43
I know what there is a big dispute about neural networks. If you have every been in an AI class you will be told jokingly never to tell what you are about to hear to a biologist.
November 24th, 2009 - 23:52
Why did the chicken cross the road?
Because chickens don’t know what roads are.
Neural may be over used.
November 25th, 2009 - 06:45
The cat brain simulation news item got me enthusiastic. I was disappointed to hear this.
It’s a major letdown from IBM. But perhaps they have a rational explanation for the way things happened.
Let’s wait and see.
November 25th, 2009 - 08:02
I’m not totally up to date with what IBM are doing, but I think the neuron models they’re using are very simple and fairly traditional ANN stuff compared to Markram’s simulation. Since nobody really agrees upon what the “essential function” of a neuron is it’s open to some debate whether they have really run a “cat scale” emulation or not.
I agree with Markram that the IBM work has been reported in the media in a misleading way as “mouse brain” or “cat brain” simulations, and that the relevant researchers appear to have made no attempt to correct misunderstandings.
However, describing what IBM has been doing as a “hoax” is probably going too far, and what this looks like to me is a battle for media attention and continued research funding between rival brain building projects. The recession must be putting budgets for this sort of fundamental research under severe pressure.
November 27th, 2009 - 06:47
Markram is right. Modha made some strange claims before. Didn’t he said that he found artificial intelligence? He doesn’t look mentally unstable, even though at the Singularity Summit I had a weird feeling about him.
Markram is more balanced in his approach. He has discovered lots of things and wrote tons of insightful papers in peer reviewed journals. He has integrity.
Modha on the other hand has only a cat in a bag.
December 18th, 2009 - 01:56
Markham claims that 10′s of thousands of differential equations are required to simulate one synapse. No particular justification is offered as to why that particular level of detail is required.
I think the relevant level of detail is somewhere between IBM’s simulation and Markham’s. The only way to actually find out is to run simulations and compare the dynamics.
I wrote this up in more detail here:
http://hyper.to/blog/link/brain-simulation-level-detail/
December 18th, 2009 - 02:04
Hm… That said, it does bother me that IBM is running larger simulations rather than trying to get smaller ones to produce interesting behaviors.