interface idea of the day - photo Q&A

I was in a very excellent talk on monday about creating smart 'companions'. One of the passing comments was that old people ( the target market for the companions ) like to sit around and talk about the photos in front of them. The idea was that the companion would start converstions with 'who is that in the picture'.

Naturally the table top people are up to their necks in photo sorting systems.

I wondered what it would be like to have some image processing software looing for areas of skin colour in the image. The software could highlite the person (area of pixels to the machine) and go 'who is this?' the user might then fill in the missing information. The idea is that the machine could then use this as a prompt to get more information. For example if the machine had ment this person before it might also know other information about them. For example it might guess who they might be with or where they might be from.
'Uncle Dan used to live in Manchester didn't he? Where was he when this picture was taken?' - this might extract out the location of the photo.
'This is on holiday in Blackpool'
For the next photo the machine might say.
'Is this also in Blackpool?'

Now what I thought would be quite cool would be to look at the distribution of colour and then try to guess who might be in the next picture(s). So matching colour might indicate the same person.
'Is this also a picture of Uncle Dan?'
'No that's Julia'.
'Sorry. 'This is Dan, thats Julia but who is this in the corner?'

more as it happens.