July 02, 2008 Archives

2008/07/02 05:19:06

some rambling

A significant architectural limitation of the writings of Elizer Yudkowsky is that they are damn near impossible to quote, thanks to their cumulative nature. He'll write about optimization processes, then the blind idiot god that is evolution, then a million different unrelated topics; and it's all very low level and unsurprising, but he'll then write something surprising and interesting that is built on all those previous posts, and is completely incomprehensible outside of that context.

The Standard Blog Format is: witty title, source link, blockquote, analysis/commentary. Most of Etc is of this format, with a scattering in other categories. You can shuffle the individual elements, but you can't outright axe any of them without breaking the format.

So it's damn near impossible to blog about anything he writes without putting in either mountains of context text and scaring people off, (slash boring them to death) or sticking in a few dozen links (as I just did) and hoping your readers follow them before jumping right to the quote and getting confused.

So if some topic comes up that tangently touches on an area of Elizer's work, you end up talking in uselessly broad generalizations* and abjectly failing to get the point across. This is the same mechanism by which I have apparently convinced a friend that Elizer is a Singularlity nutbag. Don't get me wrong, any word as stupidly broad as "the Singularity" will accumulate all sorts of weird shit, but Elizer doesn't fall under that category.

I got onto this topic in the first place, because I was writing something (unfortunately unpublishable, being fanfiction) a few days ago that tangentially touched on weakly-superhuman AI, and I needed to compactly describe the topic. I muddled along, of course, but it would have been nice to been able to rip off a key paragraph, say, this one:
If the AI is weak, it does nothing, because it is not powerful enough to significantly improve itself - like telling a chimpanzee to rewrite its own brain.

If the AI is powerful enough to rewrite itself in a way that increases its ability to make further improvements, and this reaches all the way down to the AI's full understanding of its own source code and its own design as an optimizer... then even if the graph of "optimization power in" and "optimized product out" looks essentially the same, the graph of optimization over time is going to look completely different from Earth's history so far.
I've got jack-all motivation to go back and polish it, of course. This, in case you care, happens all the damn time. I'll get a idea riding my ass, dash off a couple of paragraphs, then never touch it again. Occasionally I can salvage it, like with what became TWI; and sometimes I can salvage these fragments for completely different stories, but most of the time they gather dust. Publishing them on their own tends to inspire bile and vitriol; since they obviously are meant to go somewhere but don't provide closure, so I can't do that either. The worst case scenario is for me to die without fleshing any of them out, then have the executors of my will delete them.

*You can call Elizer "an AI researcher who sometimes writes about the Singularity" in much the same way you can call me "a custodian** who sometimes writes science fiction".

** Explaining this, while thematically inappropriate (it ruins the structure of the simile) is necessary to sooth my enormous ego. Plus, it's my blog, nyah ha ha ha. My job title, "Lighting fixture maintenance technician" is a resume friendly synonym for "light bulb changer", which also can be taken to mean "someone who maintains a building", which is a notational definition of the word "custodian". Feel free to laugh uproariously at the freshly explained humor; or rather, smirk appreciatively at the complicated literary device.

Speaking of the Singularity, it is amusing to note that damn near all of the "Singularity fiction" is actually "Pre-Singularity fiction", since the point of the singularity is that technology will progress too quickly to extrapolate. Thus, it is impossible for anything that even remotely takes itself seriously to be set during the Singularity. The Continuity series, like most singularity fiction, is set in the pre-singularity period; with the odd exception of stuff like Vinge's Marooned in Realtime, which is actually set after the singularity, and consists mostly of people walking around, going, "What the hell just happened?"

That Which Is, The Face Upon the Deep, and the non-canon Internation are set during the wacky transition to a MNT economy. She was just just the dry smell of gasoline and Coarse Adjustment are set in the post-war period, and thus have hardly any interesting violence, as well as much less bombastic titles.

Posted by bbot | Permanent Link | Categories: Etc