What does M3 tell us, page-2

I think you make some valid points which are inline to things I have been mentioning for weeks but as you say they only get shot down. From what I can tell the tech isn't anything revolutionary. The unsupervised aspect could be interesting although we may still be a long way away from making that useful / commercial.

I think the remote server API is obviously cheaper and more versatile as it can offer a one size fits all implementation to many different potential clients without having to worry about physical installations / compatibility issues. This allows potential clients to test and play around with things very easily.

I am more interested in M4 at this point than M3 as I don't think the market cares about a video demonstration nor an API and is purely waiting for the hard dollars in the bank.