On 04/02/2008, Joseph Gentle <[EMAIL PROTECTED]> wrote: > I haven't read any of Steels stuff lately, either. I'm not sure if any > of the language he's generating is higher order, but I wouldn't be so > quick to dismiss emergent language generation as a trick for just 5 > minute demos.
Well if you take something like the "talking heads" experiment (http://www.isrl.uiuc.edu/~amag/langev/cited2/steelsthetalkingheadsexperiment.html) and ask what it would take to scale this up to human-like language abilities inevitably you're always drawn back to the fact that the images used are of a trivial nature. If the images which the cameras were observing were natural scenes I doubt that talking heads (as it existed in 1999/2000) would have been able to deliver meaningful results. There needs to be some kind of reliable pattern which you can correlate your linguistics with. Uncertainties can be dealt with, but if the pattern is completely unreliable from one observation to the next you're lost. Simulation doesn't really deal with the problem, or rather it deals with the problem by ignoring it. In a simulation you can take shortcuts which would never be possible in real life. In the real world objects do not come pre-labeled, and instead have to be learned from experience which is ultimately delivered to us through our senses. ----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=93343682-8c6b7c
