The Reading Machine Revisited v2
There is, however, another way of looking at the phenomenon. It can be observed from the outside, as a behavioural phenomenon. Although accomplished readers have stopped moving their lips since at least Ambrose, the practice of reading has always left demonstrable traces: bookmarks, highlighting, marginalia, recommendation. These all have their equivalent in the digital world.
My own daily reading is conducted mostly via digital media. I go through my twitter feed and mark anything that looks interesting. Similarly, I save any articles on my feedly that look interesting. Both these bookmarks are picked up by my Instapaper. Throughout the following few days I scan or read with a variable level of thoroughness these articles 1. Sometimes, I mark them as favourites on Instapaper 2. Sometimes, I share them back on Twitter. Sometimes, I highlight passages; and sometimes I tweet these passages, or share them on Facebook. This, in one way, is what reading means now. This is how we perform the public act of reading.
Now, the idea of building a machine to automate the cognitive act of reading fails even as a thought experiment but the entire process described above, of scanning, bookmarking, highlighting and sharing, is sufficiently digitised and interconnected that we can more easily imagine it being automated. An artificial intelligence trained on our own current reading could periodically check the same sources and bookmark anything that looks as though we may find it interesting. Depending on the level of interest, it could then favourite it on Instapaper or share it on Twitter. It reads for us in the same way that Douglas Adams’s electric monk believes for us.