[extropy-chat] question from Ted Berger interview in Pop Sci

[Jef Allbright]

On 4/10/07, pjmanney <pj at pj-manney.com> wrote:
> In the Ted Berger interview in Popular Science,
>
> http://www.popsci.com/popsci/science/0e54d952c97b1110vgnvcm1000004eecbccdrcrd.html
>
> the writer says the following:
>
> "For the past four years, Granacki has been trying to develop circuitry that could translate Berger's equations into electrical pulses.  The big mechanical hurdle has been figuring out a way to reduce the amount of heat generated by the transistors so that a chip won't damage healthy brain cells.  The solution was to create a more complex version of the same kind of digital circuit that performs computations for a family desktop, except far smaller.
>
> "Jeff LaCoss... hands me a working model of the memory chip... lighter than a feather..."
>
> What does the writer mean?  Do computer digital circuits not produce the same kind of heat as other circuits?  Or is its size the reason it doesn't produce too much heat to be placed in the deep interior of the brain?

Heat generation and dissipation in dense electronic circuitry is quite
a complex subject, but a specific answer to your question is that
digital circuits don't necessarily generate significant heat except
for the moments when their transistors are changing between states.
This means the more state changes per second, the more heat is
generated.  What they have in this case is digital signals between the
chip and the computer carrying signals in both directions.  And they
have some form of A-D (analog to digital) and D-A (digital to analog)
conversion between the chip and the brain tissue.  There are various
techniques for reducing heat dissipation in such circuits but there's
always a trade-off.  For example cutting the operating voltage in half
reduces the heat to one fourth, but with detriment to switching speeds
(probably not critical here) and noise immunity (probably more
important here.)  Also, semiconductors require a certain minimum
voltage in order to switch at all.  In this particular article,
they're really only talking about the normal engineering requirements
of converting from a prototype in the lab to an implantable device.

If your interest is in regard to a futuristic scenario of high
performance brain implants, then you would plausibly have only the
necessary interface electronics inside the cranium, connected by a
hair thin photonic cable to a processing module elsewhere on the body
where power and cooling is more convenient.  The cable could easily be
routed under the skin.  If you really needed high powered computing
within the cranium, then you would need to have either liquid cooling
piped through, or some highly effective heat sink (carbon nanotube
fibers could be very effective), possibly terminating in a Mohawk to
dissipate the heat.

Theoretically, reversible computing can be done with no heat loss at
all, but I think that's too far from practice to be worthy of
consideration here.

FWIW,

- Jef