faulty redundant asynchronous communication is efficient

I found myself with 60 relevant business emails in my inbox last Wednesday, the morning before I flew out on a business trip, and I jumped over to twitter to find out if anything important was going on. Baroo? Why did my brain tell me to do that?

Four things about twitter communication of note:

  1. Human editors. You follow people who by nature editorialize. You know a human is posting something that they find interesting.
  2. Redundant. Lots of people retweet the same content. So if you miss the message from one person odds are you will see someone else post it if the message is important enough.
  3. Asynchronous. You have a minute to think about it before you hit send. This time delay is key to avoiding mistakes common in real-time communication.
  4. Faulty. There is no guarantee that someone saw your last tweet. And we LIKE this fact. The imperfect delivery is a good thing. We get annoyed when someone asks “did you see my tweet?” as if it were email!

I had an old boss in the early 1990s who used to not watch TV or read the newspaper. When asked how he would know if something big happened he always replied “if it’s that important, the news will come to me.” And I think he was right.

When looking at the four elements above, if you change “human editor” to “biological” you have a very accurate description of evolution. Biological, redundant, asynchronous and faulty are all attributes of all living things and result in an intelligent and evolved communication method.

When these attributes (biological, redundant, asynchronous, faulty) are applied to Neural Networks, these researchers show

Biological systems have a large degree of redundancy, a fact that is usually thought to have little effect beyond providing reliable function despite the death of individual neurons. We have discovered, however, that redundancy can qualitatively change the computations carried out by a network. We prove that for both feedforward and feedback networks the simple duplication of nodes and connections results in more accurate, faster, and more stable computation.

It turns out that “No, I did not see your tweet” creates a more accurate network. A neural network of humans processing whatever it is they are processing on a given day. And somehow that means the news comes to us.

And oh ya, you should follow me on twitter here.