In this demonstration we teach the iCub robot the names of new objects and highlight the capacity to learn the names of objects that are not in sight. Replicating the results of psychology experiments by Linda Smith and Larissa Samuelson (Smith & Samuelson, 2010), we use neural network models to associate names and objects via body posture (i.e. where the robot is looking). Details of the underlying model can be found in Morse et al 2010.
"'I' want to push the buttons... 'I'! want to push the buttons... now!"
doceigen 4 months ago
but will it blend?
htsk91 4 months ago 2
Just give him his twuck.
arronhunt 5 months ago 2
the iCub is Cool :) QC
Quaaludedude714 6 months ago
This is the first A.I. YouTube video I see without any stupid Skynet comments. There is hope.
rcparts 7 months ago
and as it happens the generalization you mention in this paper...
Morse, A.F., Baxter, P., Belpaeme, T., Smith, L.B., Cangelosi, A., (2011 – in press) The Power of Words and Space. Proceedings of the joint International Conference on Developmental Learning (ICDL) & Epigenetic Robotics 2011
anthonyfmorse 7 months ago
Hi CodeJeffo, you can find details of the model in the following paper...
Morse, A. F., De Greeff, J., Belpaeme, T., & Cangelosi, A., (2010) Epigenetic Robotics Architecture (ERA). IEEE Transactions on Autonomous Mental Development. 2(4), p. 325-339
It is still early days for this model but I don't think we are cheating ;-)
anthonyfmorse 7 months ago
quite interesting. But to be frankly you are probably cheating. That's too easy to understand the difference like ORANGE ball and BLUE car. Why don't you show how that lovely robot GENERALIZE knowledge ... show them BLUE ball and ORANGE car and we'll see. At this point I am not convinced at all that it is learning based on rgb based segmentation and using this information. I wish that I am not true :) Good luck!
CodeJeffo 7 months ago
very cool, its needs more intense training.
spider853 9 months ago
They could just give him a dictionary....and let him learn.
quangluu96 10 months ago