Antifragile: Things That Gain from Disorder

Before Datentreiber, I worked for a company that used real-time text analytics to detect market-moving events in stock market news. The business idea originated from the book “Black Swan Events”: events that are so rare that they cannot be predicted reliably, but so significant that they must be recognized quickly.

Most Black Swan Events such as the recent CloudFlare outage are negative for our society, economy, industry, etc., because our systems are fragile: for example, efficiency-driven centralization leads to single points of failure.

VUCA (Volatility, Uncertainty, Complexity, Ambiguity) shifts system design from efficiency to resilience. According to author Nassim Nicholas Taleb in his latest book “Antifragile”, robustness is not the opposite of fragility. Antifragile is about “Things That Gain from Disorder.”

As a bioinformatician, I was familiar with this concept: our immune system is antifragile. It gains from pathogens. Vaccinations exploit this antifragile property of the system. Some AI systems are also antifragile: they learn from “disorder”.

Taleb cites simulated annealing as an example, an algorithmic process for machine learning (based on the hardening of metals). Random errors at the beginning (in exploration) lead to a more stable model at the end (in exploitation).

“No stability without volatility.”

For data & AI systems, this implies a bimodal or so-called “barbell” strategy: on the one hand, investing in a robust foundation: data & AI architecture, governance, literacy – known in the context of Data & AI Business Design as TOP structures (Technology, Organization, People). On the other hand, taking risks with rapid innovation: prototyping data & AI solutions to respond quickly to challenges and chances.

Taleb’s recommendations for antifragile systems also sound very familiar: agile management (Lean, Scrum, etc.), decentralized organization (data mesh), redundant architecture (ensemble learning), human monitoring (human-in-the-loop), continuous improvement (data-driven), etc.

In short: the antidote to fragility is data & AI.