This was a fun blast from the past. Written in 2001, Johnson was writing his second book about the bubbling, boiling tech scene. He makes a few good predictions, which I prodded him about on Twitter- namely that he predicted Twitter would be a shithole.
I think that Johnson kind of confuses a few concepts here. Ants and cities certainly have emergent behavior. So does software that is designed around the same principle, namely: distributed agents following simple rules create large scale patterns that tend to equilibrium. That's good, and it reasonably predicts that changing the simple rules will change the large scale patterns and that the Internet would (has) changed a lot of those simple rules.
He also seems to mix up a lot of concepts that we would categorize today as 'machine learning' and that really have nothing to do with agents making their own decision. I guess ensemble machine learning is technically "many agents"?... but not really.
Good entertainment book if you like Johnson (and you should like him), and maybe a good book if you're doing historical research into pre-Web 2.o Web.20 concepts... otherwise it is a bit dated.
No comments:
Post a Comment