RE: Paths to Uploading

Billy Brown (
Tue, 5 Jan 1999 16:10:41 -0600

Bryan Moss wrote:
> If you make a high-level abstraction (that is, a piece of code that
> recreates the exact input and output) of a neuron do you
> preserve subjective
> experience? If you make a high-level abstraction of the
> entire brain do you
> preserve subjective experience? What is identity? How much is
> different?
> Even if you do think uploading is possible you're still faced
> with hundreds
> of currently unanswerable questions..

Oh, no, not the dreaded identity question! :-)

I'm not getting into this one - in my experience, everyone who cares has already made up their mind.

> Why do we have such small brains? To me it suggests that the level of
> complexity achievable is *very* close to the achieved level.

I'm not sure what you're getting at here.

> Since we're talking about plausible future scenarios it might
> be fun, being
> in the midst of millennium fever, to come up with some. No dates or
> predictions, just how you think the next few major technologies will
> pan-out. How about it? (And fifty years from now, when we're all six
> centimetres tall and living in habitat domes on the moon, we
> can have a good
> laugh at them.)

OK, I'll bite. Here's my current best guess:

Computer speeds undergo a very fast exponential growth over the next few decades, as the doubling time continues to decrease. Non-sentient AI turns out to be pretty easy once you have a fast enough machine, so automated engineering devices can be built. Limited forms of nanotech arrive pretty soon, but Drexlerian wondertech takes another decade or so - we have to wait for computers to get fast enough to design this stuff.

That gives us two major branches:

If non-sentient seed AIs are relatively easy to make, Eliezer's program (or one like it) becomes an SI between 2010 and 2030. In this case we'd better make sure it grows up sane, because its going to decide all of our fates.

If not, we have a period of about a decade in which our entire technology base improves with the speed the computer industry displays today. Real intelligence enhancement should become possible towards the end of this time frame, using neural interface hardware and fancy software. The conquest of aging and disease should also fall into this period - neither problem is hard enough to require full nanotech.

After that, the rate of change becomes too fast for unenhanced humans to keep up with. Advanced nanotech and uploading arrive within a few years of each other, and megascale engineering projects (Jupiter brains, Dyson spheres) become practical a year or two later. By then there are lots of Powers around, and I have no idea what they will be doing.

Billy Brown, MCSE+I