[extropy-chat] Wetware vs. Hardware (was IQ vs Upload)

[Dirk Bruere]

The Avantguardian wrote:

>     I have done some back of the envelope
>calculations to try to answer my own questions on this
>matter. If we have 10^12 neurons with 10^4 connections
>per neuron, the total connectivity of the human brain
>  
>
A better estimate would be 10^11
And we can probably say that less than 10% of those are involved in 
'intellectual' processes.

>is (10^12)*(10^4)/2= 5*10^15 total connections. I
>divide by 2 because the connection from neuron A to
>neuron B is the same connection as from neuron B to
>neuron A. This number alone is some 5000 times higher
>  
>
Which is now 50x

>than the approximate figure of 1 terabyte of storage
>that Moore's Law has currently yielded us, but it
>isn't even a map of the SPECIFIC neuronal connections
>(i.e. neuron A is connected neuron B and so on) but is
>instead merely an aggregate figure of the total number
>of connections.
>
>     To be a specific map, one would have to have
>allocate approximately 10 bytes to address each
>connection, 5 bytes to address the input neuron and 5
>bytes to address the output neuron. This brings us to
>a total of about 5x10^16 bytes to have a virtual map
>of the human brain with near reality level resolution.
>This is 50,000 times more data density than we have
>currently achieved. So if Moore's Law keeps chugging
>away as expected then approximately 30 years from now,
>we should have our first human level AI.
>  
>
20 or less
I also imagine that data compression (you are assuming no structure) 
might shave a few years of that as well.
So we might bet it down to 15

>    Now from a human-level, the A.I.'s intelligence
>would quickly rise within a span of a decade to
>super-human levels, if we keep giving it hardware
>upgrades. But the Flynn effect would also be operative
>over those years. Since considering that my 30 year
>time-line is about a single generation of humanity,
>then Robin's figure of 1 standard deviation per
>generation would mean that the average human born then
>would be about 33% smarter than someone born today. 
>     So by my back of envelope analysis of Moore's Law
>versus the Flynn Effect, the Singularity would still
>happen but it is still at least 30 years away.  I
>actually think I might be more prepared to deal with
>it on such a timetable than if it happened tomorrow.
>Consequently I will not lose any sleep over it. Ciao.
>:)
>  
>
The calcs suggest between 15 and 20.
Maybe a lot quicker if we could scan in a *real* brain.

-- 
Dirk

The Consensus:-
The political party for the new millenium
http://www.theconsensus.org



-- 
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.323 / Virus Database: 267.6.9 - Release Date: 11/06/2005