Print

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

[Drago55577]

i tried to make an account for docking station but im not getting the email.

[Rhodan]

If you download the C1toDS addon for Docking Station, it will add an 'Offline Mode' to new world screen.  You won't be able to go online, but everything else will work just fine.

[Drago55577]

how do i ge tthat addon.

[Neonivek]

Want me to send Steve an email asking?

That's actually a good idea. He's interested in getting further funding, just not from the general populous.

Send him an email and tell him you loved creatures and want to contribute to the new project.

The problem is Kickstarter prevents him from having additional fundraisers or even allowing additional people into it.

He couldn't make another fundraiser, until the contract time runs out, if he wanted to.

[Rhodan]

how do i ge tthat addon.

There's a download link at the bottom of this page: http://creatures.wikia.com/wiki/C12DS

[Drago55577]

i went to that page seconds before you posted that....

[Moogie]

Been a while since the 'outside world' heard news from the Grandroids camp, so I thought I'd share the latest update for you all. Hope Steve doesn't mind!

The update itself was a very long post, the tl;dr of which is basically that he had to scrap a month of work on the brain chemistry/neurology because it was turning out to be too processor intensive and needs to be simplified. However, he reports that the new architecture will be a lot more efficient and even solve a few problems and make way for some improvements later down the line when it comes time for the creatures to learn how to explore and manipulate their surroundings.

Spoiler (click to show/hide)

Hi all, things are a bit up and down at the moment, hence the lack of updates. I guess I should report the trials and tribulations as well as the good stuff though, so here's the latest. Basically I've been developing the creatures' brain and in some ways it was going great, but I've decided to throw all the code away (about a month's work!) and start again with a different architecture. It means I lose a few things I was fond of, but we gain at the other end.

As those of you who are interested in the neuroscience know, my design is based on the assumption that the cerebral cortex is divided up into maps, each of which learns to perform a different function in some learned coordinate space, for example the direction in which Gollum's eyes are facing relative to where he just put down his coffee cup, and what he needs to do in order to reach towards it. I'd planned these maps to be two-dimensional, and  have them carve themselves out of a continuous sheet of neurons. As cells that recognized similar things started to migrate towards each other and form clusters, in response to signals that spread out over the cortical surface, ridges would eventually form between these basins of attraction that stop the signals from one cluster from bleeding over into another. In other words, the clusters would form into maps and the ridges would eventually isolate the maps from each other and stop Gollum experiencing synesthesia.

This was all very neat, but it has two problems:

The first is that it's very resource intensive. Two-dimensional maps require a LOT of pattern-detecting neurons. I started out with about 36,000 neurons, which I could process about a hundred times per second, but I gradually realized this was nowhere near enough. The more neurons there are, the more memory each creature takes up and the more processor time it needs, slowing down the reaction time of the brain as more creatures are added to the world and limiting how many maps I can expect to develop. Much to my dismay, the numbers just weren't working out.

The other snag is the amount of my own brainpower needed to understand the dynamics of the system. Thousands of neurons were making and breaking connections, moving their axons around, digging basins of attraction and carving ridges, while two complete learning systems were happening at the same time, each one depending somewhat on the other. And that's just the bare bones of the system. Once you add imagery, planinng, working memory and a bunch of other stuff that I haven't entirely got my head around yet it was all just getting too complex to track.

So, reluctantly I've decided to start again using fixed, one-dimensional maps. I'm sad to see some of this self-organization go, but it will pay off in the long run.

One-dimensional maps are MASSIVELY faster and smaller. At least by the square root of their 2D cousins. And having them as fixed objects, which I can process one at a time, allows me to take all sorts of shortcuts to improve the processing speed still further. In the original version, for instance, something like "gaze space" comes to be represented by a two-dimensional map in which the middle means "straight ahead of the creature's chest" and the top-left corner means "up and to the left as far as it can see", whereas the new version just has a single row of neural columns relresenting space to Gollum's left and right, and another row representing "above and below". This is more limited but it takes advantage of the fact that most things in a creature's life take place on a plane aligned with the horizon or are otherwise representable on a continuum.

So Gollum's brain is now going to consist of a stack of ready-made strips of neural columns, like matches side-by-side in a matchbox. Being fixed in width and position means I can take advantage of all sorts of optimizations when processing the neurons, and being one-dimensional scales everything down exponentially. I can understand what the hell is happening MUCH more easily and isolate one part of the system from the rest, and this gives me much more confidence about the later phases of development as I add more "cognitive" aspects to the system. It also allows the creatures to have two distinct brain hemispheres and that leads to some interesting possibilities and advantages.

For instance, Gollum has two shoulders, obviously, one on each side. So I can split some of these strips down the middle and the maps that learn to abstract the movements of the left shoulder can appear in the left hemisphere while those for the right shoulder will appear on the right (just as they do in nature, except in real life the left side of your brain controls the right side of your body and for now it's easier to stick to left being left). "Reach space" will thus develop as a pair of maps, one representing which way Gollum's left arm is currently pointing in relation to the mid-line of his body, and the same for his right. And yet when I send signals in the other direction to control the arms I can treat the two maps as if they were two halves of the same map. So if Gollum sees a rolling ball to his right in egocentric space, he will automatically reach out for it using his right hand, but if the ball rolls past the mid-line into the left side, he'll use the other hand.

This stripy approach also gives me a lot more genetic control over how everything wires up. I was hopeful that all the wiring would occur spontaneously in the old system, but experiments suggest that Gollum might take months merely to learn to walk, much like a human baby does. I doubt any of you have the patience to wait that long! So now I can genetically specifiy exactly which maps any other map receives its signals from, massively speeding up learning. And I can even replace some of the maps with specially-coded algorithms to perform the more complex functions that I don't yet know if I can compute neurally. Ideally I'd make most of the maps specially coded at first and then replace them one by one with a more general scheme as I figure out how. But in the meantime it gives me plenty of fall-back positions so that Gollum can know things about his world that otherwise he might not be able to figure out.

So, that was a whole month's work down the drain! I learned a lot and have a much better feel for how I should have done it, but heck, it's put me behind again and I still have to unpick all the old code yet, before I even start writing the new stuff. I'm sure it's the right thing to do, though. Such is the progress of science...

Before I even commit fingers to keyboard I want to do a bit more exploration of the design yet. A few new thoughts have cropped up just today - exactly how yin and yang signals need to reflect back on themselves to create inverse mappings and the wherewithall for imagery and planning, for instance, and how I might possibly be able to multiplex the system as a way of supporting working memory, daydreaming imagery and even mirror neurons. But those are a story I'll tell later!

This last paragraph explains an accompanying screenshot.

Yet again, I have nothing to show you because it's in pieces. Thanks for being so patient. This is REALLY hard stuff, but I'll get you things to look at and put them into your hot, sweaty hands just as soon as I can. Anyway, before I cover the floor in nuts and bolts and gizmos that I'll probably forget how to put back together, here's a quick screenshot of the gene editor with Gollum learning about which direction he's looking (this is the simplest thing his brain has to learn, so it's a good test). To the left is the UI for the gene editor, with one of the peripheral nerve genes open for editing. Gollum's busy looking around himself, driven by random noise in his motor cortex, and the gray rectangle behind his head is a simple chemistry monitor. The rectangle with the green blobs on it is the brain scanner (these things will look much prettier in the actual game, obviously; I've just built the bare minimum for debugging purposes right now). The UI at the bottom of the screen appears when you get close to the scanner, so that you can control its settings. The different colors on the scanner represent the activity of different layers in Gollum's cortex. Down in the bottom left are the proprioceptor signals from his head and eye muscles (all the rest of his senses are turned off at the moment) and the other blobs are neural columns with signals bleeding out from them, which are slowly trying to come together into a map of gaze direction. TOO slowly.

(And to reiterate something we've been told many times, but others may not be fully aware of: those "Gollums" are NOT what the final grandroids are going to look like. They're just test dummies, for now. )

[Pnx]

I recognise that skybox, this must be being built in Unity! Doubtless anyone who kept up on the project could have told me that, but I feel amused with myself that I figured that out.

[GlyphGryph]

Great to hear what's been going on.

[Orb]

I recognise that skybox, this must be being built in Unity! Doubtless anyone who kept up on the project could have told me that, but I feel amused with myself that I figured that out.

It also says its Unity at the very top of the image. 

[Pnx]

I recognise that skybox, this must be being built in Unity! Doubtless anyone who kept up on the project could have told me that, but I feel amused with myself that I figured that out.

It also says its Unity at the very top of the image. 

...

So it does, let's all congratulate me on my perception again.

[Drago55577]

waiting :|,  im glad i have many pony things .

[Felius]

A curiosity from someone who didn't backed it back then: Is the project still alive? I've looked around a bit, but I can't seem to find any update that isn't dated back to the kickstarter funding date, about an year ago. I'd like to hear about someone who is in the grandroid site if the project is still going and healthy, even if not the specifics.

[Neonivek]

A curiosity from someone who didn't backed it back then: Is the project still alive? I've looked around a bit, but I can't seem to find any update that isn't dated back to the kickstarter funding date, about an year ago. I'd like to hear about someone who is in the grandroid site if the project is still going and healthy, even if not the specifics.

It is still alive it is just that UNLESS you were one of the original donators you get NO information until the game goes Beta.

[PTTG??]

It is allegedly going. Steve doesn't talk much about what he's working on. He's been building physics, then biochemistry, and currently brains, over the past year. The last time we got something that was officially a full update was, I think, three months ago.

But he does appear to be working, and the stuff he is doing is really amazing. He just won't show anyone.

EDIT: The vibe I got was that it's okay to tell people what's going on, just don't copy-paste the actual content of the updates.