 |
welcome, guest |
| Development Forum | |||||||
 |
| ||||||
 Geat_Masta   |
To start out "Docking Station.catalogue" contains this:
There's just one problem with this: it's wrong. It's like really wrong. To start out: the neuron that activates the deactivate script doesn't even exist, which is probably why no one uses deactivate scripts in their agents. But there's a more fundamental problem. Some of these decisions act on an agent, like activate1/push, and some of them don't like left/west, so when we list that out we get this
It may not look that bad, but this is a massive problem once we start looking at the brain. In the brain neurons 0-11 are all connected to the combination lobe, meaning that a norn needs to learn to sleep when paying attention to each category of object individually, rather than learning that sleeping is generally a good idea when tired. The same is true of dropping something, meaning that a norn needs to learn to say, drop gadget to drop a cheese, and that the fact that they're holding something in no way factors in to the decision to drop it. Similarly, norns can't learn that they can't get something if they already have something. In both cases of get and drop, because of the way the brain is wired, the norn just gets stimulated punishment, with no way of learning why that sometimes happens and sometimes doesn't. And this is a massive problem, to really fix this drop needs to be moved out of the middle of that array, and down towards the bottom. It's similar to how norns are theoretically capable of learning "there is an elevator in front of me, a vendor to my left, and a machine to my right, therefore i should go left" and other such navigational tasks, but the brain is just wired in a way where they can't. In short, quiescent/look should be at the top because it's the default action, then the actions using agents should be bunched together, followed by the unused actions (because we don't know if they need something), and then the scripts that don't need an object should be bunched together. All things done this array should look like this:
This change would be required to get the brains working at full capacity, but it would also mean all existing genomes would be incompatible because the 2 unused actions got inserted, and all existing learning would be invalidated because the actions got moved around. |
||||||
 Lurhstaap  |
Whooooo. I mean. WHOOOOOOOOO. You may have identified THE mother of all brain editing projects, my friend. Perhaps even of all genetics projects. To get the brain working at its full potential is one tempting proposition. But that is one holy heck of a lot of work. Reordering those would require repairing chunks of the nonbrain genome, too, as they would now refer to the wrong neurons, and such. To be honest, though? I say let's do it. People don't HAVE to use it, first of all. Second of all, every new major genomic advance has been largely incompatible with past genomes anyway. I can't breed a CFF to a standard Chichi and expect good results. And thirdly, I have no doubt that if doing this works, people will then naturally port over their favorite genetics even if we don't include them in the release. If someone SERIOUSLY wants (to invent something off the top of my head) a BrainOptimized Colortrue Mer-Ettin, they'll make that sucker. People will build on this work just as it builds on past work. Look at that! The program itself evolves. Funny. Anyway, I think you get my point. I wish I could contribute more on a design level, but I'm feeling like I need to sit down with a bunch of books before that'll be possible. Unfortunately, as far as I can tell, right now you're the most active person in the community rocking these kind of skills. (Which is another reason I'm eager to absorb this information - God forbid something happens and you disappear!) EDIT: with all of that said...BOY have we succumbed to major, major project creep. IT STARTED SO SIMPLY. lol I had this problem at my last job, too... I'm a real sucker for a good (sounding) idea. My coder then hated me. I think he spent more time explaining to me why things I'd come up with wouldn't work than he did actually coding. Conclude with killer catchphrase. (Lurhstaap) "This is not knowledge - this is information!" New Model Army, "Courage" |
||||||
 Intyalle |
If the change does get made, a quick guide as to what needs changing to port old genetics would be good. Also known as bab_5_freak from the old Gameware forums. |
||||||
Lurhstaap |
He came up with an alternate way to tackle this that involves making a new lobe. It may not end up being as painful to convert the old genes as in this setup. BUT with that said we will definitely be documenting all changes - not just so people can update old genetics, but also to show exactly what was done and how, to enable people to better effect their own brainwork. Conclude with killer catchphrase. (Lurhstaap) "This is not knowledge - this is information!" New Model Army, "Courage" |
||||||
Geat_Masta |
2 lobes actually, the sequence has to be rearranged coming from the verb lobe, as well as going to the decision lobe. |
||||||
Lurhstaap |
Oh, that's right. I forgot that part. (BTW, I haven't begun work on that because I broke reproduction somehow and I've got to fix that first XO) Conclude with killer catchphrase. (Lurhstaap) "This is not knowledge - this is information!" New Model Army, "Courage" |
||||||
| downloads cobs adoptions creaturelink metarooms breeds gallery art wallpaper screenshots graphics promos sprites |
dev hack shack script reservations dev resources active projects dev forum community links advice chat polls resources |
creatchi forum bookmarks general news help development strangeo survivor |
mycaves log in register lost pw |
1 online Papriko |
| creatures caves is your #1 resource for the creatures artificial life game series: creatures, creatures 2, creatures 3, docking station, and the upcoming creatures family. contact help privacy policy terms & conditions rules donate wiki |
||||