Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Talvieno]

Copper??  Are you kidding me? Make it out of candy! =D Mod it in! Imagine... An AI made of candy... Undefeatable in every way.

No, really, though. BloodBeard, you've done an incredible job getting this far. If you get it to recognize even one word, you've more or less completed your goal, and shown up all the naysayers who said it couldn't be done. You are awesomeness, personified. 

[BloodBeard]

"sleeping on it" really seems to work for me when I run into these kinds of problems. Regardless of how I proceed I don't think i'll be needing to start over. If I intend to build a new text display for the user, i'm pretty sure I can repurpose the existing one to be usable by the AI. I've had an idea on how I wanted the AI's display to work and the one now is almost perfect for the job, it just needs some tweaking. Relinking the hatches on the display would be pretty easy, and since the AI has no need for a backspacing feature I only need to rebuild certain minor gears a bit to get rid of it and allow the clearing option to clear the display in one fell swoop instead of the fully automatic backspacing that it is now.

This all assumes that i'll be able to figure out a way to break down what the AI wants to say into individual characters so it can type them out.

[BloodBeard]

Option 2 where I make a new user display control and repurpose the existing one is out. I was really hoping it would work, I'd have so much extra space but there's just no way I can add support for words in a space-friendly manner, if it was even possible to begin with.

Unless I can find another way to do it, after laying the (massive) groundwork i'll only need a 9x5x5 area of space for each word programmed into the AI (not including the logic that tells it the words meaning and purpose). They would have to be built aboveground because I will have run out of underground, and you can't build right to the edge, but the entire 3x3 embark would still be enough space to support over 400 words. I hardly need that many, it's just an example.

Pretty much all the mechanics of this need 5 z-levels to work and if I were to take up all the underground and sky I could stack 5 sets of machinery like what I have uploaded on top of each other. With the word recognition groundwork and the word support built i'll have taken up 2 of them. The question becomes will 15 z-levels of a 3x3 embark be enough to build the decision making logic of the AI and all the other things it needs, I should find that out before putting my dwarves to work.

[MaximumZero]

If you're running out of CPU power, you could always come and use mine.

[shadenight123]

bloodbeard, space isn't an issue.
use dfhack and make the fortress a "lair" so there will be no scattering.
abbandon and then reclaim as a bigger area. you just got more space! and without losing previous work!

[BloodBeard]

Very clever... but my FPS is hurting already, I doubt it could take that kind of punishment. Even though some of the people who have downloaded 0.1 have gotten decent FPS, I doubt very much there would be many people with computers that could handle 25 z-levels of a 3x3 embark filled to capacity with working mechanics let alone even more. I should probably accept that if I can't do it in the amount of space I have it's not going to happen without downgrading to a more simple AI with preprogrammed responses (or that markov chain thing talked about a while ago).

Unless I can find a good reason to keep them in i'm considering scrapping numbers and operators. It would shrink the giant word recognition mechanics by a huge margin, cut several thousand linkages from the AI display (and several thousand mechanisms from the existing display), and makes things so much simpler without damaging the prime focus of this project.

[Quietust]

bloodbeard, space isn't an issue.
use dfhack and make the fortress a "lair" so there will be no scattering.
abbandon and then reclaim as a bigger area. you just got more space! and without losing previous work!

You cannot reclaim a fortress and increase the region size - the reclaim is always exactly the same size as the original embark.
You also cannot create a new embark that overlaps an existing site - if you force it, you're bound to run into mysterious glitches such as tunnels being spontaneously "un-dug".

[BloodBeard]

He said to turn it into a lair and that would then let you freely embark on top of it with a larger area, I think. I don't know if it would work then.

[Mego]

Implementing a markov chain algorithm in DF will take a bigger hit on your FPS than preprogrammed responses. It would probably be just as bad as whatever you had planned.

[BloodBeard]

The FPS will likely be shitty regardless how it's done. Everything requires endless banks of pressure plates and pumps, the only thing different is the gearing and linkages.

The space and linkage reduction from removing numbers is very tempting... It's practically half the size it was before.

I'll leave you with this frightening image:

Spoiler (click to show/hide)

[Mego]

My brain just turned to vapor.

[MaximumZero]

My brain just turned to vapor.

...How did you reply with MAGMA in your head? 

[Urist McDwarfFortress]

...Portal... AI...

Uh oh.

[BloodBeard]

Balls to it, i'm digging.

[Loud Whispers]

Balls to it, i'm digging.

And so begins Dwarfputing AI PT. II