OpenCog Brainwave - Latest Comments

Re: Genetic Crossover in MOSES

Linas Vepstas — Fri, 23 Mar 2012 17:15:22 -0000

Ahh, but I didn't want to ...

"1+2+3 every 10" uses the simplex extrapolation for 9 steps, and then a full neighborhood rescan every tenth generation. As you can see, it advances rapidly then stalls.

"1+2+3 rescan" uses simplex, until there is no improvement, and then it does a full neighborhood scan.

rev6654 is whatever bzr revision 6654 does.

nn16 is bzr rev 6600 but with a "last chance" rescan disabled.

nn18 is same as nn16, but with a different random seed. Notice how dramatically different they are.

nn28 is bzr rev 6666, i.e. the "final" code.

Because of the sharp dependence on the initial random seed (and thus, implicit random seed changes when algos access the RNG in a different order) makes it hard to compare results. Ideally, one should repeate measurements, averaging over at least 20 or better, 100 different random seeds. This is hard to do...

More details in the text file "diary-performance.txt" in the source code tree.

Re: Genetic Crossover in MOSES

Nil Geisweiller — Wed, 21 Mar 2012 02:47:19 -0000

Awesome work, Linas.

Your last remark "the algos which advance the fastest initially seem to have trouble advancing later on" is indeed pretty interesting. Maybe MOSES could do the tweaking itself depending on the target score or number of evals, etc, possibly via some form of metalearning.

Could you briefly detail the legend of that last graph?

Thanks...

Re: Sentence Patterns

web designing company — Fri, 09 Mar 2012 06:09:52 -0000

The good source is presented here..nice to read this information.Really too knowledgeable.

Re: Tuning Metalearning in MOSES

chhean — Mon, 20 Feb 2012 17:12:32 -0000

Well done Linas

Re: Determining word senses from grammatical usage

Rob — Tue, 14 Feb 2012 05:19:45 -0000

Hi,

So how have you progressed with this model?

Re: Tuning Metalearning in MOSES

J P — Sun, 12 Feb 2012 19:26:34 -0000

Very nice write up Linas - thanks :-)

Re: Hacking on Link-Grammar

Website Hosting Companies — Thu, 02 Feb 2012 07:56:21 -0000

This is what I have been searching in many websites and I finally found it here. Amazing article. I am so impressed. Could never think of such a thing is possible with it...I think you have a great knowledge especially while dealings with such subjects.

Re: Preview of a virtual learning environment

MDude1350 — Tue, 10 Jan 2012 00:11:24 -0000

That Bomberman Hero music.
This virtual learning environment sounds like something I've been wanting for a while. The bots you have seem nice, but I think it'd be fun to see what I can make on my own. Would I be able to use the environment to embody other AI systems without them being a part of OpenCog, or would I need to roll my own system for that?

Re: Preview of a virtual learning environment

Affordable cheap web hosting — Fri, 02 Dec 2011 08:03:35 -0000

Great post on virtual learning environment.. You made certain good points there. I loved it.

Re: Preview of a virtual learning environment

J P — Sat, 17 Sep 2011 03:55:58 -0000

There is definitely an aspect like curiousity. In OpenPsi (the emotion and motivation system), there are demands for "competence" and "certainty". "Certainty" will lead the character to try to understand the state of environment, "competence" will lead to the character trying to get better at influence and predict the result of its actions.

Re: Preview of a virtual learning environment

Sa3vis — Fri, 16 Sep 2011 10:15:24 -0000

I'm just wondering if curiosity and memory will be added in this project. Curiosity in that if there is a new object or button it will go observe it and then depending on what that does it will either ignore, stay away from, be cautious with or use. Maybe even hoard. And with memory then if it sees something it doesn't need at the moment it would keep a virtual memory map of where it is located and then find it when needed. Those are just a few things I was thinking would be neat. Keep up the great work and good luck.

Re: OpenCog Hong Kong Project

Matt — Wed, 24 Aug 2011 16:12:40 -0000

Wow, I have used many of the tools that you have (festival, zeromq, [amqp], [rabbitmq], [xmpp], protocol buffers, [raw sockets in raw php or etc], .. etc;;;.

Re: Preview of a virtual learning environment

J P — Thu, 11 Aug 2011 00:45:45 -0000

Hi - My aim is to be as transparent as possible about what we're doing. We're not interested in trying to deceive people with narrow scripted behaviours.

So, saying that, these videos are mostly a demonstration of the virtual world. OpenCog has specialised path-finding algorithms, but these can be constructed as part of a "action plan". So a plan might include an action to "move to the battery cube" (where battery cube is a unique id related to a previously seen object).

So we're abstracting actions at a certain level, but not confining OpenCog to only know about certain objects. We can easily add new actions to objects and it's up to OpenCog to detect the patterns of what happens when those actions are initiated.

We are building on an older embodiment system that was quite limited in what it could learn. Now a lot of time is being spent on pulling out the old parts and replacing them with general systems.

Re: Preview of a virtual learning environment

Free Thought — Wed, 10 Aug 2011 12:05:26 -0000

I'm slightly confused about whether these videos demonstrate anything that moves in a direction closer to AGI or if it's demonstrating narrow AI. Is there any chance of seeing a demonstration soon that might help to show some behaviors that are clearly solving varied and/or unique problems using a single methodology? Its hard to currently visualize how this system will be able to perform tasks outside of its narrow restraints via hard-coding requirements for the AI to try to satisfy. Even though it may be learning to satisfy these requirements in its own way, how far will one be able to ultimately expand upon the capabilities of such an AI?

Re: Preview of a virtual learning environment

J P — Sun, 07 Aug 2011 02:49:33 -0000

Not yet I'm afraid - the Unity world is under heavy development, so until it settles down we'll probably wait until we can create a decent tutorial and supporting documentation.

Re: Preview of a virtual learning environment

Matt — Sat, 06 Aug 2011 23:43:20 -0000

Cool. Is there a tutorial for setting this up like the old one for Multiverse on the cogbuntu page?

Re: Determining word senses from grammatical usage

Stuti — Wed, 18 May 2011 06:41:57 -0000

Awesome thought yaa but have you tried implementing it?
Actually I need to rank the linkages generated by link grammar.
It would be nice if you coukd give in your comments.
:)

Re: Determining word senses from grammatical usage

cece — Thu, 03 Mar 2011 11:03:01 -0000

very interesting...cool

Re: Determining word senses from grammatical usage

Jacky Alcine — Thu, 24 Feb 2011 10:09:17 -0000

Very interesting; do you think it's possible to a trigger reflexive of sorts to certain verbs then?

For example, "Eliminate all references of John Doe in my e-mail."
And it carries out the deletion of John Doe in any reference in my e-mail, per say.

Re: OpenCog navigating Nao robot

J P — Sat, 04 Dec 2010 22:20:47 -0000

Here it already "knows" the location of the ball from being informed by a rooftop camera about the position of objects. Vision processing to detect and recognise features is something we're working on...

Re: OpenCog navigating Nao robot

JB — Sat, 04 Dec 2010 15:42:20 -0000

Does the robot already know the location of the ball, or does it locate the ball and then determine a path?

Re: OpenCog navigating Nao robot

Joel Pitt — Mon, 27 Sep 2010 17:20:53 -0000

Yes, I totally understand what you mean Bob.

Here I know that speech recognition is occurring (I don't know if it's biased towards recognising "go to" statements), and then the path is worked out (probably using A*). I believe they are also using a rooftop camera to position things. The latter is obviously a hack, but it's a start.

Unfortunately I only get sporadic updates from Xiamen, but I'll likely be moving to Hong Kong where I'll be able to get a better feel for the progress and how it's implemented.

Re: OpenCog navigating Nao robot

Bob Mottram — Mon, 27 Sep 2010 13:36:44 -0000

It's always difficult to judge demonstrations like this, because they depend upon not so much what the robot does as much as how it does it and whether the methods used can generalise. This sort of demo was done long ago with Shakey, but with a lot of contrivances, out-takes and men behind curtains.

Re: Sentence Patterns

stephenk1 — Sun, 25 Apr 2010 08:54:39 -0000

any chance that Jon Barwise's stuff might be helpful?

Re: Hacking on Link-Grammar

jasonforceau — Thu, 08 Oct 2009 12:10:00 -0000

Hi, i am looking for tools for syntactic analysis for the system of my Final Year Project and found your post so interesting. But I don't know Link-Grammar. And how can I start up to using Link-Grammar to apply into my system? is it using c language? anyother language?