Open in TheBrain

2014 The Human Brain Project: Emulating the Brain's Computational Capabilities - Henry Markram

<p>Dec 24, 2014 03:35 PM</p>

<p>

Henry Markram speaks in Seoul &nbsp;on Dec 1 2014...at SDF 2014.</p>

<p>

7 B &nbsp;per year spent on brain resch worldwide.</p>

<p>

we cannot understand what we are generating.</p>

<p>

1 B euros... 100 scientists around Europe.</p>

<p>

bring all info together;</p>

<p>

digital model of human brain and simulate on super computers.</p>

<p>

deliver by 2023.</p>

<p>

build a multiscope...</p>

<p>

all... micro and macro... &nbsp;level of gene and proteins and cells;

and system level...</p>

<p>

see the magic that you saw in Sherlock.</p>

<p>

algos that connect all the data...</p>

<p>

entire struc of brain...</p>

<p>

100B neurons....

1 M km of fibers... runs on 20 watts. &nbsp;&nbsp;to the moon and back twice.</p>

<p>

1000 trillion synapses... &nbsp;smaller than a bacterium.</p>

<p>

10s of millions of proteins in each synapse.</p>

<p>

get hypotheticallly a Sherlock, if all goes right.</p>

<p>

we need super comps... &nbsp;billion billion calcs per sec

200 PBytes of data in memory &nbsp;with IBM;</p>

<p>

how does it func?</p>

<p>

go inside a cell... you a billion proteins in each cell.

10s of M of interactions in each cell...</p>

<p>

begin with the mouse...</p>

<p>

put brain into virtual mouse....</p>

<p>

in virt enviro...</p>

<p>

learn all that mice know how to do...</p>

<p>

step inside a brain....</p>

<p>

how to create perception...

we see w our brains... &nbsp;you are imagining me...it is an imaginarium;</p>

<p>

simulate diseases...

what chains are there; &nbsp;at molec, at cell level...</p>

<p>

gather data from hosps...</p>

<p>

complete map... abt 600 difft brain diseases...</p>

<p>

brain learns... and highly robust... can lose half your brian before someone notices...</p>

<p>

how can virt robots etc learn... implement on neuromorphic processors...

2 kinds:

1) &nbsp;the ARM proc based... attach to robots...

2) other &nbsp;10K &nbsp;faster than real time...</p>

<p>

will not be one person.. one einstein....</p>

<p>

build software, med know....</p>

<p>

train future doctors and reschers.</p>

<p>

involve the whole public... science museums... future of med and future of comps...

crowd source component.</p>

<p>

2 trillion per year problem &nbsp;(10% of world gdp is reduced by brain disease)...</p>

<p>

&nbsp;</p>

<p>

&nbsp;</p>

<p>

&nbsp;</p>

Attachments