[extropy-chat] Desirability of Singularity (was Are ancestor simulations immoral?)

[Anders Sandberg]

Eliezer S. Yudkowsky wrote:
> Reality check: a traveling spike involves many ions being released from
> the cell membrane, traveling with the potential grade, then
> painstakingly pumped back in against the potential grade.  It's not
> firing that takes the energy, it's preparation to fire.  If I recall
> correctly, it takes one ATP->ADP reaction per ion pumped against the
> grade.

Actually, it is three sodium out and two potassium in per ATP. A very cool
little pump. And indeed, it seems to be the main energy cost of the brain
- the spike is just a release of the spring.

>  And I would expect much more than a thousand ions released per
> total synaptic spike.

Surprisingly enough, it is on the order of 100,000 ions. Much more than a
thousand, but still far far from moles. The Wikipedia gives a calculations
for a higher upper bound for a somewhat big spike, 50 million
http://en.wikipedia.org/wiki/Membrane_potential#The_number_of_ions_involved_in_generating_the_resting_potential

> So the inefficiency relative to the thermodynamic
> limit is surely more than just three orders of magnitude.

The Brillouin inequality is only about information erasure. Many of the
brain computations may be rather information-preserving. A synaptic signal
for example, if perfect, would not cost any thermodynamic cost for
erasure. In practice the release probability is 10-30% according to
Markram and Tsodyks ( http://diwww.epfl.ch/~gerstner/SPNM/node33.html ),
so that would be on average 2-3 bits of erasure per synapse and signal.

Hmm, around 8e14 synapses with an average population of 1-10% neurons
firing at 1-100 Hz. That makes 1e9-1e12 firings affecting 8e12-8e15
synapses. At a cost of 2.4e-21 - 3.6e-21 J this is 1.9e-8 - 2.8e-5 W. So I
get five-six order of magnitude for this with the most pessimistic
calculation.

Going backwards, we would have an allowance of 1.6e11 bits/s per neuron
and 20e6 bits/s per synapse if all the energy use was for computation.
Given a maximal input/output rate of the nervous system on the order of
gigabits per second, this seems rather high.

Ah, the Brillouin inequality, synaptic release probability and singularity
all in one thread. It feels good to be on the Extropians list again!


-- 
Anders Sandberg,
Oxford Uehiro Centre for Practical Ethics
Philosophy Faculty of Oxford University