I was watching this BBC drama and it said something about a coded message in a public digital image. I thought this would be a nice coding exercise.

Imagine we have Alice and Bob. They both own a private image X. Alice wants to send Bob a secret meassage. She alters the pixels in image X and send the Y version to Bob, via social media for instance. He can now decode it: message = Y - X.

An image has a RGB channel for each pixel. (255, 255, 255) is white and (0, 0, 0) is black. We can…


Clichés are the word blocks that printers used instead of spelling the word letter for letter, they are a shortcut, a frequently used word or phrase. Our language is full of clichés. Digital makes it easy to copy words and insert into a “new” text. A meme is also a block that can be easily copied. When language becomes clichéd then thinking slows down. We live in a postmodern world of over quoting. We are building the hypertext that is the web of our cultural reality.

Clichés and memes are contagious, they transfer quickly across our network.

Originality is rare…


Wissensdurst und Wissenslust sind die Dränge immer mehr zu wissen, Wissen anzuhäufen. Das Wissen als Besitz. Die Neugier, die Gier Neues zu erlernen, und zum Wissensstand zu hinzuzufügen. Das besessene Sammeln des Wissens. Die Wissensgier, das Verlangen, mehr zu erfahren. Die Bibliothek in unserem Kopf. Die Verbindung der verschiedenen Wissenschaften, das Einsehen, dass Vieles Eines ist. Das exponentielle Wachstum der Wissensverflechtungen. Das Wandelbare des Wissens, Wissen ist immer zeitbezogen. Die Unfassbarkeit des Wissbaren, jedoch die Wissenslust. Das Unwesentliche der kleinen Tatsachen. Das Eureka des neuen Wissens. Der Rausch der Wissensflut.


The physicist Maxwell famously devised a thought experiment where he postulated that a demon could sort the atoms, hence violating the second law of thermodynamics (2TD). The 2TD states that entropy must always increase as warm and cold particles will mix. It is one of the few laws, if not the only one, that has an arrow of time.

The Lamarckian (anti-Darwinian) view in biology is that an ancestor can pass on selected acquired genes/traits to its descendants, hence leading to an improvement in traits.

Both of these concepts are similar in that they have an invisible hand of information that can sort and select atoms or genes. The truth is though that there is no hand but that randomness (entropy) rules. In the case of evolution, the fit survive and thrive because their random mutations bring surprise improvements.


I am not a medical expert but approach this from a statistical point of view.

I was wondering what a positive test result would tell you about the probability that you actually have the disease. I think for the UK swab test we have a true positive rate of 0.99 and the true negative rate is 0.80. I think this is because you administer it yourself and it can be unreliable.

In order to derive the probability you have the disease given a positive test P(D|T) we need a prior for P(D) and then apply Bayes theorem. If we assume…


Will AI ever be able to model and predict the complexity of the world? The answer is no. AI systems are becoming increasingly complex themselves, with complex architectures and parameter spaces. This means that AI doesn’t make our world simpler but more complex. We need to handle AI to human interactions and maintain AI systems.

This can be visualised in this graph

Or in this equation where AI models the world and adds to total complexity

Total = world + AI(world)


AI research has made great advances in the last decade. But at the same time we are still far away from strong AI or AGI. I think there are several leaps of faiths, assumptions or generalisations when people compare recent inventions to reaching full AI.

  1. Assumption an AI good at one task can be good at other tasks. All the celebrated successes have been in single tasks (win Go, drive a car etc).
  2. Assumption that AI has a body that can perform intelligent tasks. …

I was wondering whether we can estimate under some assumptions whether the gain in life years due to Covid lockdown (distancing and shutdown) outweighs the economic cost.

In the UK there are some estimates what a human life year is worth, which is used when the NHS makes decisions to extend a life. This is called QALY — quality adjusted life years. I assumed the average age of a dead Covid patient is 60 (I ignore the gender bias here). I estimate that the death toll would be 47x in case of no lockdown (60% of world population with 1%…


Watching the Bulls documentary on Netflix, I was curious how NBA teams performed over many seasons. I could not quite find this in ready form so I had to create it myself.

I scraped Basketball Reference to get regular season stats (1980–2019). It should be a good proxy for overall season performance.

The code can be found on Github. The visualisation script is in written in R.

I calculated a few metrics: total number of wins, win percentage and the mean rank (across seasons). I was surprised to find that the Spurs actually come first in terms of games won…


There have been stories of never seen unemployment claimant figures in the US due to the Covid lockdown. The outlook for the US and global economy looks bad. US GDP for 2020 hasn’t been released yet, but I wanted to test whether I could predict it using weekly unemployment claims as a leading indicator.

I got the data from FRED and the code is in R (on Github). I got 3 series from the FRED site: quarterly real GDP, weekly claimants and monthly working age population. I wanted to use the latter to normalise claimants but that didn’t actually improve…

Dirk Nachbar

Googler, Data Scientst, views my own, twitter.com/dirknbr

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store