That's neat, but I think having this with some lightweight, web-optimized static images would make this better without being detrimental to lightness.
Sometimes a picture does say a thousand words. A photograph of a person being interviewed is something that adds a lot to an article. Plus there's the whole photojournalism aspect that's lost in only keeping text.
It's not the images, per se, that make heavy pages heavy, after all. It's tons of JS that pulls in FSM knows what.
Does it? It seems that it ought to add nothing to the content - unless the content specifically references their appearance.
As well as that, unless it's a person you can google independently (like a head of state, or large company CEO) there's very little chance you can know whether the image is actually the person being interviewed, and even less chance you can tell whether it was /this/ interview. What you do know is that the image will have been posed, chosen from a lot of candidate images, and quite likely edited after that.
It might feel good to recognise and see a human face, and therefore "add a lot" from the publisher side - exploiting that sense to make people click and keep people there longer - but that's not quite the same.
"Add"? It's a part of the content. It's been in paper for more than a century.
Consider e.g. https://web.archive.org/web/19990208015258/http://www.nytime...
Oh my God. I forgot how nice it is when websites Just Work without any lag or stuttering or ads
http://lite.cnn.com/en
On the other hand, I think there's definitely a case to be made for rich, application-like presentation of content, when it enhances the delivery of information. For instance:
http://www.r2d3.us/visual-intro-to-machine-learning-part-1/
The problem is that everyone dreams of making something cool like the latter, when their actual content is much closer to the former.