I am not convinced it is that useful to have explicitly targeted audiences for advertisement. It may be better to understand and interpret an ad campaign, but generally, what works best when doing statistical prediction is not what is the most intuitive.
Also, even though your example is obviously not meant to be taken as is, it shows that targeted demographics quickly don't have a lot of data behind them. Successful stories in AI usually involves lots of averaged data with only little "focused" data to adapt your model quickly (e.g. as done in speech recognition where models are estimated on 1000s of hours from many speakers, and the model is then adapted for the one speaker to be recognized).
I think the value of the so called social graph for advertising is overestimated. IMO, what's interesting about facebook is more the amount and diversity of data than its personalized nature. But then, I have little knowledge about algo for advertisement targetting, maybe the situation is different than the domains I am familiar with.
Also, even though your example is obviously not meant to be taken as is, it shows that targeted demographics quickly don't have a lot of data behind them. Successful stories in AI usually involves lots of averaged data with only little "focused" data to adapt your model quickly (e.g. as done in speech recognition where models are estimated on 1000s of hours from many speakers, and the model is then adapted for the one speaker to be recognized).
I think the value of the so called social graph for advertising is overestimated. IMO, what's interesting about facebook is more the amount and diversity of data than its personalized nature. But then, I have little knowledge about algo for advertisement targetting, maybe the situation is different than the domains I am familiar with.