Slashdot Mirror

From TFA it's not very clear what this simulation achieved. It was code that already existed and as far as I understand it, it was used to validate some simulation models of low-level biological vision.

However his simulation did not necessarily achieve computer vision in the usual sense, i.e: shape recognition, image segmentation, 3D vision, etc. This is the more cognitive aspect of the visual processus, which at present requires a much higher level of understanding of the vision process that we do not posess.

FYI the whole brain has already been simulated, see the work of Dr izhikevich. It took several months to simulate about 1 second of brain activity.

However this experiment did not simulate thought, just vast amounts of simulated neurons firing together. The simulated brain exhibited large-scale electrical behaviours of the type seen in EEG plots, but this is about it.

This experiment sounds very similar. I'm not all that excited yet.

I agree 100% that arbitrary barriers are worthless. More often then not, it's the developers who believe so (I consider myself a SW Engineer not a Developer), who don't get work done, because they can't just evolve around problems as they come up.

Developer -> I've only ever programmed in (Insert Language Here), but I'm really good at it.
SW Engineer -> I use C for my embedded projects because of code size and performance, mostly C# for my windows applications because I get to re-use all these great MS libraries, and Frankly pretty much any time I need a quick and dirty tool for generating test scripts, or helping with Cadence I use PERL.

Look for people who evolve themselves to meet a problem, rather than try to make the problem fit their system. I'd say more often then not these people will have excelled time after time in seemingly unrelated jobs, versus the people who continue doing the same thing forever, 'til they become an obsolete commodity and are only good for complaining about outsourcing.

Most of the people I consider superstars are really just regular old problem solving engineers. These people can take any problem, and either find or build components to make it happen. I've known a *FEW* superstars who were ONLY software engineers. I'm sure that even those guys need some variety and support too though.

Honestly I'd say what's harder is retaining people like this, because they CAN always get another great job. It's probably more important that you give these people interesting, and varied work, make sure you don't overload them as they're already the most productive piece of your puzzle, and give them a strong supporting cast they can leverage to be even more productive.

At least that's what I think makes me a superstar, and has exemplified the ones I've worked with before. What makes me feel like I can say these things?

Some of my background:

Started at AMCC as an intern, basically a lab tech.
4 years at Qualcomm in my early days doing test.
http://vesicle.nsi.edu/nomad/segway/ -> Neural Simulation/Segway based Soccer Playing Platform
http://natural.uchicago.edu/~tgal/ -> Parsing and Analysis Tools for Affymetrix Gene Array Arabadopsis Sequencing
http://www.ietf.org/internet-drafts/draft-ietf-speechsc-mrcpv2-15.txt -> IETF Speech Recognition Protocol
Development from scratch and evolution of a couple of medical devices (Triage Wireless/Dexcom)

Currently I'm a couple years deep working at a semiconductor startup, where I've done some Digital Blocks for our ASICs, Design and Layout of our basic Electronics, and ALL of the embedded firmware, along with PC Software for in house, and external use. I truly believe, and have yet to be disproved, that I can just about solve any problem. Usually it's the time -vs- desire for elegance that's my weakness. I guess that's what management is for....right?

If they had AI that could run on fast computers, then they'd have AI that could run on slow computers, just slowly.

True. But fast computers may help quite a bit in developing AI. This simulation of 100 billion neurons and a quadrillion synapses took 50 days to process one second of simulation-time. An interesting proof of concept, but not exactly ideal for experimentation; you get 7 tests a year. But increase the CPU power by 1000x, and now it only takes an hour to simulate a second and you get to do a lot more tweaking.

Actually, Eugene Izhikevich run a successful simulation of the entire human brain in 2005:
http://vesicle.nsi.edu/users/izhikevich/human_brain_simulation/Blue_Brain.htm#Simulation%20of%20Large-Scale%20Brain%20Models
Simulation of 1 second took 50 days on a cluster of 27 PCs (~4.3M times slower than realtime). Eugene is a pretty smart guy, except not as prominent as Kurzweil. Here is Eugene's estimate for AI timeline:
http://vesicle.nsi.edu/users/izhikevich/human_brain_simulation/why.htm
You may also want to google for Henry Markram and his current project.

Actually, Eugene Izhikevich run a successful simulation of the entire human brain in 2005:
http://vesicle.nsi.edu/users/izhikevich/human_brain_simulation/Blue_Brain.htm#Simulation%20of%20Large-Scale%20Brain%20Models
Simulation of 1 second took 50 days on a cluster of 27 PCs (~4.3M times slower than realtime). Eugene is a pretty smart guy, except not as prominent as Kurzweil. Here is Eugene's estimate for AI timeline:
http://vesicle.nsi.edu/users/izhikevich/human_brain_simulation/why.htm
You may also want to google for Henry Markram and his current project.

For anyone interested in fruit fly brain activity check out this paper by Bruno van Swinderen in Bioessays:

http://vesicle.nsi.edu/users/bvs/bioessays2005.pdf

Scroll down to section "Selective attention in the fly brain".

Albert

This is the most ambitions??? What about Markram & IBM? They must be just fooling around with that Blue Gene (actually I do think they are fooling around, but that's beside the point). What about Izhikvich? He simulated just a puny 100 billion neurons. That's *nothing* compare to this "most ambitious" million.

Modelling 'real' neurons in detail is generally done with ~10k compartmental models, which are generally described by something like:
http://neuron.duke.edu/cells/
and modelled in something like:
http://www.neuron.yale.edu/neuron/

Even using vastly simplified neurons, like integrate & fire types, for example: http://www.nsi.edu/users/izhikevich/publications/s pikes.htm
you still have many vastly different types of spiking behaviors.

You then still have to deal with the fact that neurons 'generally' connect to about ~10k others, (actual range something like 10-100k). And that's before you get to details like what neurons are where, with what densities, that long range connections in mammalian brains are generally not very well understood, etc. etc. etc.

The brain is a lot more complicated than you think. We're still many many years away from modeling a mouse brain, at a purely neuronal level. I mean, there still isn't a definitive model of the Aplysia, neuron count ~10k...