Thursday, March 11, 2010

Finished Part One

Well, more or less.  I'll probably end up having to go back and edit something. 

Part Two has about 15,000 words so far.  The reason I got so far into part two before really finishing the first part is that I dreaded going back and editing.  Not because I don't like editing, but because I lived the story the first time.  It was intense writing it, and I thought it would be as intense to edit.  A funny thing happened.  The really bad parts weren't as bad (maybe because I knew they were coming), and other parts got me.  Yah, well.


  1. Fantastic! I read it in one setting and can't wait to get my hands on part 2. I love the novel point of view and it has given me fresh perspective on some possible variations of an AI rich world.

  2. Just finished reading, and I absolutely loved it. Thanks for sharing this. :D

    It does make me wonder how we will look at the scenario you describe in five years time.

    In five years of software running "in a cloud" (like the artificials do on The Company's machines), will we still be thinking of processes as individuals or rather as a loose coupling of semantic services?

    Already now, I mused about the boot process of the artificials - aren't they software rather than hardware? Wouldn't they "initialize" rather than boot?
    Wouldn't they "hibernate" rather than boot (as they keep their memories)?

  3. I don't know why the comments aren't arriving in my inbox like they're supposed to...

    I imagine the boot process as necessary for cleaning up problems that nopping can't. I wonder if our brains do a kind of reboot when we sleep. I guess I imagined these PDAs running on some kernel that sets itself up like a virtual machine, so virtual booting not physical chips cycling power.

  4. The word "booting" comes from"bootstrapping", which in turn comes from the phrase "to pull oneself up by one's bootstraps." Depending on the workings of the container (operating system, in most cases) and how much of the work it does for the child process, I think it could make sense. Especially for something that seems to be as self-directed as an ai.

  5. @benwr, The motivation is really the Halting Problem, and the fact that the AI system will be very complex and poorly understood. So reverting to a 'known good' state occasionally is necessary, else the whole thing may crash and mangle memory beyond repair. So I imagine it like loading up the image of a virtual machine that's the template, and then adding in the incrementals that make that AI unique, including 'life history' as expressed in some gigantic state vector.