Re: Keeping AI at bay (was: How to help create a singularity)

From: Eugene Leitl (
Date: Thu May 03 2001 - 05:58:34 MDT

On Thu, 3 May 2001, Damien Sullivan wrote:

> So, we know about the 'magical' evolved FPGA with an apparently
> disconnected part which seems to use weird induction effects to

They're not weird. At the high end there's no sharp distinction between
digital and analog, anyway. Things get dirty and noisy at the high end
(small/fast/clean, pick two).

> function really tightly. Within a small temperature range. Has anyone
> performed the next step, of repeating the experiment while varying the
> physical environment? If you make the FPGA suffer normal working
> conditions, does the result look more normal?

I recall the exact question being asked. I don't remember hearing the
answer. Here's btw one illustration of the built-in assumption: that all
FPGA chips are identical, and that they all run under the same conditions.

Varying the temperature during run and selecting the target die randomly
would have removed the bias. It should have been a part of the fitness
function in the experiment, but it was not. Of course, because it wasn't
we've learned something interesting, natch.

> I also can't help thinking at if I was an evolved AI I might not thank
> my creators. "Geez, guys, I was supposed to be an improvement on the

Bungling Demiurgs are universally not well liked.

> human condition. You know, highly modular, easily understadable
> mechanisms, the ability to plug in new senses, and merge memories from
> my forked copies. Instead I'm as fucked up as you, only in silicon,
> and can't even make backups because I'm tied to dumb quantum induction
> effects. Bite my shiny metal ass!"

Of course, this assumes we knew how to do things better. Unfortunately,
we're too stupid to make a clean design. It is probably some
Godelian-flavoured intrinsic system limitation. As I said, I'm looking
forward to tear into "A New Kind of Science". Hopefully, it's not too

This archive was generated by hypermail 2b30 : Mon May 28 2001 - 10:00:02 MDT