From: "jeffrey E." <[email protected]> To: Noam Chomsky Subject: Re: Date: Mon, 07 Sep 2015 19:21:46 +0000 i want the machine to be put in an environment , with 5 different languages and learn them do your think watching movies. might help On Mon, Sep 7, 2015 at 8:06 PM, Noam Chomsky < > wrote: Don't know if or when this letter will get sent. Lots of internet problems the last few days. One of the problems of living in a third world country. Paris is better, I'm sure. These are interesting suggestions, but I'm not sure they are the right ones. They're very different from the Hilbert questions. These were deeply embedded in rich theoretical understanding, and were at the borders of such understanding. Furthermore, solving the problems would have a major impact on enriching mathematics. They were very different from the imitation game, which, as Turing explained, had no particular intellectual interest but could be useful in encouraging the development of better hardware/software. Incidentally, as a matter of historical accuracy, I'm keeping to his famous paper. Elsewhere he seems to have taken it more seriously. In formulating suggestions, it is important to avoid the fashionable (and moronic) approach of google-style statistical analysis of Big Data. Thus one could perhaps get a better prediction of leaves blowing in the wind by that approach than by invoking the laws of motion and interfering factors, but the better prediction would be scientifically worthless, whatever utility it might have. Now consider these ideas. Take the first. The Big Data approach might provide a tolerable solution that wouldn't have the slightest scientific or other interest — that is, giving some understanding of how infants do it. Take the others. A serious and meaningful answer would require an understanding of the phenotype — the state attained by the acquisition system. Here there are very significant questions, but like the Hilbert set, theory-internal on