Re: How hard a Singularity?

From: James Higgins (
Date: Sat Jun 22 2002 - 19:16:32 MDT

At 08:02 PM 6/22/2002 -0700, you wrote:
>James Higgins wrote:
> >> Now, can either of you explain to me why a human-equivalent intelligence
> > will, all of a sudden, be capable of creating bounds & leaps of technology
> > that were otherwise impossible, just because it is running on
> > silicon???
>Why? Here's a few reasons.
>1) Totally infallable memory

Not necessarily true. The AI architecture determines memory storage and
retrieval, not the hardware. It is most likely that they will NOT have
infallible memory because they will need to delete (forget) details
continuously (like we do) in order to stay within their allocated storage
capacity. If at a later time the AI needed some of those details it would
either come up blank or attempt to deuce them based on other related
details it retained. Thus their memory would resemble ours much more than
that of a computer. This could, and most likely would, improve over time
as more storage capacity became available and as its algorithms
improved. It could have the ability to decide (consciously) what to
forget, which *may* be an improvement (or maybe not).

>2) 24 x 7 operation, no breaks for body maintenance

I account for this by saying it would be equivalent to 2-3 people. This
could possibly be as many as 10 (since inter-communication would be vastly
reduced, which takes much time on large/complex projects). Or could be as
little as 1 if it works more like us and gets burnt-out or bored if working
on the same thing 24x7. No way to tell in advance.

>3) Improved speed of access to 'canned' facts (Encyclopedias, Google, etc)

This will be the case to some degree, since (for example) I would benefit
if I could simply think words instead of typing them. But this would
contribute much less than imagined since most (at least of my time) is
spent thinking instead of typing (I've learned to type nearly as fast as I
can code, which is somewhere around 75WPM). Plus I must spend some time in
advance (away from a keyboard typically) thinking about architecture and
design issues. Two years ago I produced 350,000 lines of code in one
year. If I didn't have to type I'd guess I could have done 450,000-600,000
(at most) since much of my time was spent thinking, designing and
communicating (not typing). So while some speed will be gained here it
won't contribute much when it comes to thinking & problem solving.

>And by FAR the most important reason for "leaps & bounds" is:
>4) Ability to reprogram ones own cognitive processes.

Long term, yes, short term no/maybe.

See my post: "How to gauge positive progress (was How hard a Singularity?)"

>If this isn't enough... then... nuts! --> Its plenty enough!!

Not when you take my above points into account. Note that I would *like*
this to be enough and I would *love* for all of your points to be
absolutely true. Unfortunately, I don't find it likely.

James Higgins

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:39 MDT