4 ms·

Emergent Neural Network Simulation System

16 points by robg 18y ago

Silentio 18y ago
I hope it doesn't attain AI, bootstrap itself onto the internet and take over the world.
whacked_new 18y ago
Hey robg, are you familiar with the general landscape of cognitive modeling/simulation implementations? Although Emergent claims to help creating models of cognitive processes, it still seems like one of the "low-level" NN libraries (evidently, it is listed alongside FANN and Matlab NNT). I have very superficial exposure to this stuff, but they seem quite a bit removed from creating complex simulations, unless you already have a detailed, theoretical model of particular cognitive processes, which would probably also be isolated models. On the other hand there's ACT-R and SOAR, which seem to fit the description better. Or am I mistaken? By the way, for those interested, ACT-R is aimed at providing a comprehensive whole-brain simulator, and runs on common lisp-based NN.
- robg 18y ago
  You're right on track (though I've barely done anything in this area). The common complaint against ACT-R was that it could never be reducible to a biological level. Since it started to introduce those NN components it's much further along in the last few years. That said, I think there's a way to come at multiple higher-level functions in one complex NN model if you respected anatomical localization of function (and perhaps quasi-modularity). Randy O'Reilly (http://psych.colorado.edu/~oreilly/ http://psych.colorado.edu/~oreilly/), part of the lead group that put this package together, has done some great work along these lines.
wastedbrains 18y ago
Nice I enjoyed doing some AI at CU,good to see there is still some interesting AI work being done there. I worked a lot on text categorization with SVMs and LSA/i while there. I did some NN stuff, but I always wished I got to do more of it.
chadgeidel 18y ago
I clicked the link hoping to see "emergent behavior" from a NN. Instead, the thing is called "Emergent." cries

16 points by robg 18y ago

I hope it doesn't attain AI, bootstrap itself onto the internet and take over the world.

Hey robg, are you familiar with the general landscape of cognitive modeling/simulation implementations? Although Emergent claims to help creating models of cognitive processes, it still seems like one of the "low-level" NN libraries (evidently, it is listed alongside FANN and Matlab NNT). I have very superficial exposure to this stuff, but they seem quite a bit removed from creating complex simulations, unless you already have a detailed, theoretical model of particular cognitive processes, which would probably also be isolated models. On the other hand there's ACT-R and SOAR, which seem to fit the description better. Or am I mistaken? By the way, for those interested, ACT-R is aimed at providing a comprehensive whole-brain simulator, and runs on common lisp-based NN.

robg 18y ago
You're right on track (though I've barely done anything in this area). The common complaint against ACT-R was that it could never be reducible to a biological level. Since it started to introduce those NN components it's much further along in the last few years. That said, I think there's a way to come at multiple higher-level functions in one complex NN model if you respected anatomical localization of function (and perhaps quasi-modularity). Randy O'Reilly (http://psych.colorado.edu/~oreilly/ http://psych.colorado.edu/~oreilly/), part of the lead group that put this package together, has done some great work along these lines.

wastedbrains 18y ago

Nice I enjoyed doing some AI at CU,good to see there is still some interesting AI work being done there. I worked a lot on text categorization with SVMs and LSA/i while there. I did some NN stuff, but I always wished I got to do more of it.

chadgeidel 18y ago

I clicked the link hoping to see "emergent behavior" from a NN. Instead, the thing is called "Emergent." cries