From a profile of Robert Cottrell, who reads over 1000 articles a day for his newsletter:
Robert has built an AI to help him read even more articles.
As I said before, I’m currently constrained by my physical capacity to read so many articles in a day.
Last year, I began working with a computer scientist named Jeremy Davies to build a machine learning algorithm to read articles for me.
Basically, what we’ve done is trained the algorithm on all of my past issues of The Browser. And we can set it to read thousands of posts that I don’t have time to read, and judge them.
This ML is essentially learning to be me. It’s basically an output from the virtual Robert Cottrell.
It’s still very early, but if we set it to read 1,000 articles it will return roughly 50 articles as being high quality. Of those, roughly half are going to be false positives and I’m going to look at them and think, “What’s going on here?” The other half of them are going to be real. I’m going to say,“Yes — exactly, I can see why it has chosen this.”
When it becomes better trained and a little bit better overall (when it’s finding three or four slam dunk pieces a day), I won’t have it fully substitute for me, but I’m going to use it to read more things. So, if I’m able to cope with a thousand feeds already, I can send the ML to go read 10,000 feeds.
While Cottrell is doing this for extreme productivity, the idea of a personal AI reminds me of the human editors that Neal Stephenson describes in his book Fall.
In the rural parts of the country, there’s “Ameristan,” where the poor and uneducated live steeped in misinformation, hoarding ammunition for a war that will never come and crucifying those they deem to be heathens.
But the coasts and the cities are full of wealthy elites who pay to have professional editors weed out fake news from their digital feeds, and regard the rest of the country with a kind of disgusted anthropological fascination.
We already live in an age of misinformation, and an AI to help readers sift through data to glean credible knowledge is already necessary and would be hugely valuable. Whether it’s made broadly, cheaply available versus commercially distributed will determine whether we end up with a schismatic future of the sort Stephenson sketched, or we understand and collaborate with each other. At this moment, information discovery services like Twitter, Facebook (.com/Whatsapp) and Google (search/Now/Youtube) are doing a pretty pedestrian job.