From: James Higgins (jameshiggins@earthlink.net)
Date: Tue Mar 20 2001 - 00:55:47 MST
At 03:00 AM 3/19/2001 -0500, Brian Phillips wrote:
> Something else I would like to point out....
> It's obvious reading the writings of the "singularity's friends"
>that they tend to think in terms of "Before Singularity" and
>"After Singularity". What's motivating this "discrete point"
>vision of the transhuman future? shouldn't the perception be
>a steepening CURVE?
> The only thing I can think of which might produce a "single
>point" style "event horizon" is the invention of Eli's hyper-evolving
>transhuman AI, and only then in a world lacking competitors.
> The inclusion of infosystems directly into the cortex of humanity
>will be a gradual process (taking at least a few years, at least)
>(imagine that as a nanoprank...everybody wakes up with neurojacks!).
> Similarly nano-heaven (a la Diaspora or what have you) doesn't arrive
>the moment the first general purpose assembler comes on line.
> All these things may happen extremely quickly by the standards of
>past human history...but the curve is still extant, just abbreviated into
>decades or even years producing unrecognizable change.
>
>just a thought,
>brian
The singularity as a point in time comes from self-improving AI. That is
why it is essentially called the singularity, in fact. If you really sit
down and start looking at the numbers, it all pretty much happens in a
couple of seconds. I did a fairly complex spreadsheet on this because I
had a hard time believing it. Espicially when you consider that it takes
time to actually manufacture new hardware & upgrade the physical computer
the AI is running on. I did this purely based on "faster" AI (not
smarter), that essentially just continuously makes faster and faster CPUs
for itself. Assuming a subjective 18 month development time, 6 month
real-world manufacturing & 1 month real-world upgrade time. I then reduced
the manufacturing & upgrade time per iteration by like 5% (assume that some
research had been spent on improving this process also). With all of this,
it took exactly 2 years to go infinite, which happened in the final seconds
of the last year. When I graphed it out it is near-flat until the end, at
which point it goes vertical. I realize this will never be truly accurate
and does not prove anything, but it is interesting in any case.
The reason it goes infinite, even considering the manufacturing delays, is
due to iterations. In later stages the AI is able to churn through many
iterations that never even make it to manufacturing. I assume that the
last completed version whet to manufacturing every 6 months.
So this single AI goes infinite in a single second, but how does that
effect the world, right? Well, I believe it is further assumed that this
AI will have control over Nanotech (which may reduce the manufacturing
time, btw). If this is the case, this single AI could essentially upload
all of humanity within a very short period of time (a week or two?).
However, I believe I have thought of a good way to actually make the
upload. But I'm certain this group can put some holes into it, so here goes.
FYI: I assume that the AI goes infinite when it has access to Nanotech.
The AI would start by undertaking a project to move all of humanity into
VR, producing nanobots specifically for this purpose. It would use
whatever method it decided was best, of course. But for my purposes, I'm
thinking about the emulation model. A nanobot attaches it self to every
neuron in a person. Once the person can be modeled exactly, they can be
uploaded to VR. When everyone is in this state, the AI would "move"
everyone into VR at the exact same time. The VR world would then exactly
mirror our current one, so they would not realize anything had
occurred. Step 1.
Now that everyone is virtual, the rules are changed a bit. It becomes
impossible to either die or become pregnant. The human race would be
informed (I am not clever enough to write this speech) of the current
state. A time line would also be established for allowing transcendence,
say 20 years. This gives enough time for everyone to be at least 20 years
old. Maybe 30, 40 or even 50 years would be better to allow for more
psychological adjustment (that is not for me to decide). In the VR time
is, of course, subjective. So these 20+ years could occur in a much
shorter "real" time. Once the time had come, it would be possible to
transcend. Maybe phone booth type structures would appear that, when
entered, uploaded the individual. This provides people to progress of
their own free will beyond this point.
Eventually, I imagine, everyone would transcend and the simulation would
vanish.
So, any comments on that?
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:36 MDT