[00:07:15] *** ChanServ has quit IRC
[00:07:19] *** SeH| has quit IRC
[00:07:19] *** archels has quit IRC
[00:07:19] *** flamt has quit IRC
[00:07:40] *** SeH| has joined #spacenet
[00:13:35] *** ChanServ has joined #spaceNet
[00:13:35] *** flamt has joined #spaceNet
[00:13:35] *** archels has joined #spaceNet
[00:13:35] *** irc.freenode.net sets mode: +o ChanServ
[00:18:44] *** archels has quit IRC
[00:18:44] *** flamt has quit IRC
[00:23:35] *** flamt has joined #spacenet
[00:23:36] *** archels_ has joined #spacenet
[00:23:43] *** archels_ has quit IRC
[00:23:43] *** flamt has quit IRC
[00:29:19] *** archels_ has joined #spacenet
[00:29:19] *** flamt has joined #spacenet
[00:30:46] *** flamt has quit IRC
[00:30:46] *** archels_ has quit IRC
[00:31:16] *** archels_ has joined #spacenet
[00:31:29] *** flamt has joined #spacenet
[00:47:36] *** archels_ has quit IRC
[00:50:26] *** archels has joined #spacenet
[00:53:15] *** archels has quit IRC
[01:22:42] *** archels has joined #spacenet
[01:58:35] <SeH|> archels: no worry. all directions as far as it can go. anything is better than the current situation. my current stuation at least
[02:00:48] <SeH|> archels: i dont know about your current situation. but i am majorly disappointed in this world.
[10:21:41] <archels> In what regard?
[10:25:09] <SeH|> :)
[10:25:45] *** SeH| is now known as SeH_
[10:26:18] <archels> I mean, generically? Or this world's OSes?
[10:27:02] <archels> You cannot expect to tackle all the world's issues at once, right?
[10:29:41] <SeH_> ideally all issues, yes
[10:30:09] <archels> Through SpaceNet?
[10:31:39] <SeH_> or whatever it becomes, including some other systems
[10:32:49] <SeH_> btw (i dont know if i was clear about it) but zoran's decision to use netbeans does not interfere with our spacenet plans
[10:33:17] <SeH_> in fact we may use netbeans platform in spacenet too
[10:33:21] <SeH_> for example, plugins
[10:34:44] <archels> right
[10:34:54] <archels> I'm just wondering at this point what exactly SpaceNet's goals are.
[10:35:29] <SeH_> im wondering too. its open-ended and room for anyone to suggest some
[10:36:22] <SeH_> http://automenta.com/node/8
[10:36:38] <archels> heh I was just browsing that
[10:37:07] <SeH_> it should be clear that i think spacenet has the potential to be a competitor to windows, gnome, os-x, etc
[10:37:18] <SeH_> (or something like spacenet)
[10:38:04] <archels> Right, but is it going to go in that direction if you start focusing on evolutionary games rather than, say, GUI development?
[10:38:46] <SeH_> it would shift the focus, right. but i dont intend to concentrate on evolutionary games
[10:38:55] <SeH_> i do intend to support more physics and physical interaction
[10:39:49] <archels> sure
[10:39:50] <SeH_> and to add more intelligence into the interface
[10:40:11] <SeH_> if that requires neural networks, or atleast attention dynamics across semantic networks
[10:40:15] <archels> It's just unfortunate that you can't do everything, so you have to restrict your effort and time to a reasonable subset of goals.
[10:40:39] <SeH_> http://opencog.org/wiki/OpenCogPrime:AttentionAllocation
[10:40:52] <SeH_> agreed. i havent been following a specific plan for some months now
[10:41:22] <SeH_> if there were more developers anchoring the path then thered be more need for planning
[10:41:31] <archels> Well, it seemed to be going pretty well until this Critter craze. ;)
[10:41:42] <SeH_> when did the criter craze start?
[10:42:10] <archels> With the snake, I suppose.
[10:42:21] <SeH_> http://automenta.com/artificial_life_development_environment
[10:42:33] <SeH_> several months ago actually
[10:44:06] <archels> Hmm, maybe I'm not clear on what your goals are for implementing that snake. Do you want to experiment with evolution/neural nets, or do you want to create an environment/IDE (SpaceNet) so that others can use it to experiment with evolution/neural nets?
[10:44:51] <SeH_> both, absolutely
[10:45:03] <archels> I dig what's on that page, but it sounds very much in the line with the "OS" nomer.
[10:45:31] <archels> But actually toying with critters instead of creating the virtual world so that you _can_ play with critters, doesn't.
[10:46:01] <archels> Do you see what I mean?
[10:46:07] <SeH_> no? eventualy the entire world will be a meta-critter
[10:46:46] <SeH_> the same way i think spacenet can compete with windows, etc - i think spacenet can compete with all existing IDE's and web browsers
[10:46:57] <SeH_> essentially all existing software apps
[10:47:47] <archels> Right, but take the web browser example - there's a difference between creating a browser (and testing it on some prototypical web pages) and creating actual full-featured web pages.
[10:48:51] <SeH_> so where do you think i should focus next?
[10:49:46] <archels> Well, you've proven that you can use SpaceNet as an IDE for evolutionary sandboxing and neural nets. This is good, but I would leave it at that (for now, at least).
[10:50:16] <archels> So e.g. focus on the things described on that page, i.e. making those things possible.
[10:50:27] <SeH_> which page?
[10:50:48] <archels> /node/8
[10:51:18] <SeH_> agreed
[10:52:43] <archels> and /alife_dev as well
[10:53:17] <archels> But focus on making it possible, i.e. IDE dev.
[10:53:50] <archels> And not getting hooked on your own product and spending all your time perfecting critters _inside_ the IDE. :)
[10:54:15] <SeH_> lol i didnt even start perfecting yet it - i was testing the physics engine
[10:54:30] <archels> Alright, that's good.
[10:54:41] <SeH_> i had to test physics pointer interaction and the 6DoF constraints
[10:54:49] <SeH_> and i decided to make a video for my #critterding friends
[10:55:06] <archels> cool
[10:56:36] <archels> Did you implement a file browser yet by the way?
[10:56:54] <SeH_> not yet, that would be a good feature
[10:57:01] <SeH_> how do u think it shold work?
[10:57:56] <archels> Not sure, but trying to effectively visualize the giant hierarchical structure might be fun.
[10:58:27] <archels> e.g. you couldn't pre-load it all at once, it'd have to be dynamic
[10:58:42] <archels> Maybe zoom into a node (folder) to look at the files it contains, or so.
[10:59:09] <SeH_> http://opencog.org/wiki/MindOntology:Economic_Attention_Allocation
[10:59:20] <SeH_> archels: do u understand this?
[10:59:26] <SeH_> what do you know about opencog
[10:59:39] <SeH_> http://opencog.org/wiki/OpenCogPrime:AttentionAllocation
[10:59:41] <archels> I know that I don't like it. ;)
[10:59:47] <SeH_> what dont u like?
[11:00:24] <archels> The presumption that they can make AGI by slapping together a bunch of heavily contorted symbolic models.
[11:00:50] <SeH_> well besides the entire AGI goal, im asking about this attention allocation part
[11:01:06] <archels> Of course, applying some of its principles (like maybe attention allocation) to SpaceNet may be good, because the context is completely different.
[11:01:11] <SeH_> exactly
[11:01:32] <SeH_> well the attention system basically associates each node in the memory with a value
[11:01:40] <SeH_> in opencog its several values: short term importance, long term, and very long term
[11:02:00] <SeH_> but in an isolated context (like a spacenet space), we can simplify it to one
[11:02:07] <SeH_> it can be called attention, importance, or energy
[11:02:11] <SeH_> or temperature
[11:02:34] <archels> sure, maybe
[11:02:37] <SeH_> we can use the attention value to decide geometric characteristics
[11:02:41] <SeH_> like size or color
[11:02:56] <SeH_> so, adding a burst of energy to a specific node will cause it to grow
[11:03:21] <SeH_> if the energy is spread, then evrything associated with that node will receive some of that energy
[11:03:33] <SeH_> so theres different energy dynamics that can be applied
[11:03:38] <SeH_> some continuous, some instantaneous
[11:03:59] <SeH_> energy can be spread in different ways, ex: along certain edges in the graph
[11:04:13] <SeH_> it can also be collected or filtered
[11:04:17] <SeH_> just like photoshop
[11:04:18] <archels> sure, but I'm not sure about fiddling much with the geometry of a node
[11:04:31] <SeH_> its about chanigng the relative geometry of several nodes
[11:04:35] <archels> (I mean I actually don't know, and I'd love to see some experiment videos or so)
[11:04:44] <SeH_> so that as attention shifts, so does the interface - continuously
[11:05:03] <SeH_> so, if you start at the root node of the filesystem and add energy
[11:05:03] <archels> because when you mess with the geometry, you might break the recognizability of a node or scene to a user
[11:05:09] <SeH_> then as you navigate you can push and pull
[11:05:25] <SeH_> right but it can also be constrained with limits, so that if something gets too small it disappears
[11:05:31] <archels> yeah, sure
[11:05:41] <archels> or you could indeed at least apply it to a subset of nodes, like the filesystem nodes
[11:05:47] <SeH_> right
[11:05:49] <SeH_> in isolated contexts
[11:05:57] <SeH_> attention doesnt need to be some global thing, like it is in opencog
[11:06:16] <SeH_> spacenet decouples graphs from 'ScalarMaps'
[11:07:19] <archels> yeah
[11:36:40] * archels just received Toward Replacement Parts for the Brain
[11:38:59] <SeH_> is that a book? physical or e
[11:41:12] <archels> yeah, ordered it from Amazon.
[11:41:27] <archels> tangible. :P
[11:56:47] <archels> Got it for $3.99, too.
[12:05:33] <SeH_> http://opencog.org/wiki/Hopfield_network_emulator
[12:05:54] <SeH_> archels: do you understand this? they emulate a hopfield network via the importance of nodes
[12:06:05] <SeH_> so its like a neural network superimposed on a symbolic network
[12:08:11] <archels> interesting mix of high- and low-level approaches
[12:10:02] <SeH_> ImportanceUpdatingAgent - pays "wages" and collects "rent" from atoms in the form of Short term importance (STI) and Long term importance (LTI).
[12:10:04] <SeH_> ImportanceDiffusionAgent - spreads STI along HebbianLinks.
[12:10:05] <SeH_> ImportanceSpreadingAgent - alternative way to spread STI along HebbianLinks.
[12:10:07] <SeH_> ForgettingAgent - "forgets" atoms with low LTI.
[12:10:08] <SeH_> HebbianLinkUpdatingAgent - updates the weights of HebbianLinks based on what is in the Attentional focus of the OpenCog instance at the time it's run.
[18:00:00] <flamt> http://www.youtube.com/watch?v=uMCG__YXxVk telepathic-critterdrug 300x speed
[18:00:02] <flamt> >_>
[18:01:05] <archels> You know, CritterDing is the type of simulation which would really benefit from computing power.
[18:12:06] <archels> http://www.youtube.com/watch?v=K4Hm7cdYOM0
[18:12:15] <archels> ^ holy frack, awesome
[18:15:40] <SeH_> http://www.youtube.com/watch?v=9U5_A4eNzCE
[18:28:04] <archels> nice visualisation
[19:21:12] <archels> except of course, that the dynamics are completely wrong
[19:23:15] <SeH_> not wrong, just not emulating earth-like neural nets ;)
[19:24:09] <archels> true, true
[19:29:09] <SeH_> i imagine neurons w/ flow reversal
[19:29:20] <SeH_> where synapses and dendrites are either reversible or the same thing
[19:32:42] <archels> The way I figure, nature probably had good reasons for making them unidirectional. ;)
[23:40:36] *** flamt has quit IRC
[23:44:28] *** flamt has joined #spacenet

   January 16, 2010  
< | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | >