Weekend Reading CIII [Late]

by uber

This week has been absolutely crazy. I was in Antwerp, then we had visitors from a foreign delegation, and I am at a review at the moment. It’s been so busy that I missed the links altogether. Apologies.

How not to Count Prison Overcrowding

This reminds me of the tricks around waiting lists and cancellations that some hospitals in the UK engaged in.


The Making of Lemmings

I played it on the Amiga, a computer which was a gem of a machine but completely odd. Cartoon violence, strong visuals and perfectly structured flow were the secrets of a real classic.


The Origin of the 10,000 steps per Day Goal

My phone tracks my steps, and sometimes I do look at it. It can be nice to see a day of long walking being clocked up.
I have not succumbed to a 10,000 step target, mainly because I don’t use my phone on the treadmill. It does underestimate steps when I run (I’ve compared it to the GPS) and I do wonder how many ‘steps’ a plane ride or car ride’s bumps clock in at. I have seen some Sedaris-like behaviour in some of my friends. I think when fitbit HR handles swimming I will consider one.


The Real Scars of Korean Gaming

It strikes me that these people are highly dedicated professionals. It’s not a sport as such, in the same way that being a concert pianist isn’t a sport.


Library of Congress Chief Retires Under Fire

Anthony Panizzi was an extraordinarily colourful figure for a Librarian. Libraries need charismatic leadership. They also need considerable luck. There are a lot of situations where missing out on one generation’s fad has saved documents or artefacts for wiser, cooler, later heads. The keynote at the conference in Antwerp showed some of this, where they recovered Archimedes from a scraped religious text. One line from the talk:

We sent this to a researcher at Stanford, he pointed to this and said “This is the earliest known depiction of a circle in the Western World”

sent a shiver down my spine.


The Unreasonable Effectiveness of RNNs

Apologies if this one is a bit technical and a bit dense. The title is a reference to an earlier piece on how unsupervised learning works surprisingly well in things like machine translation and spell checking. Machine Learning (as opposed to Artificial Intelligence) has yielded some extraordinary advances recently. This is the third wave at least of sweeping advances, and the best is yet to come. The key thing is that these are ‘soft’ AI, which is to say they don’t simulate general intelligence, but rather are very (very) clever statistical representations of cognitive problems. The interesting thing is that we might end up climbing the cognitive ladder through this back door, though general artificial intelligence may not be achievable. In any case, it’s not clear we’d recognise a true AI if we made one.