109 - Global Antifragility

In going through Nassim Nicholas Taleb's book "Antifragile", someone who grew up in a city going through civil war, and subsequently getting rebuilt for the 8th time in its history, I was reminded of Cyprus, my previous destination in my most unlikely of slow world tours.

Cyprus is another country that has a history full of one invasion after the next, and the only remaining country in the world with a wall dividing its capital city. My hosts during my stay there, a lovely family living in one of the suburbs of that capital city, and a couple of blocks walk from the UN's demilitarized zone, recently sent me a good short documentary covering it.

Even with most of my focus placed on remote work, when I did walk around to see Cyprus and Nicosia what I saw was often reminiscent of what the documentary shows. A handful of places were building new homes, and a handful of streets were full of shops, but those were more the exception to the rule.

Humans can form collective intelligence and become more than the sum of their parts, less biased and more intelligent than their most intelligent members, benefiting immensely from their differences and cooperation. They can also form Groupthinks, tribes that are even less intelligent than their least intelligent individual members.

In the AI domain today social engineering is driving increasing adoption of delusions, redefining of terms to fuel hype, and new kinds of Ponzi Schemes. At least half of the largest AI Influencers I come across are "Disinformation Brokers", who knowingly serve this purpose. A few experts like Yan LeCunn remain stalwart in their sanity and maintain integrity, but many have fallen from such pillars this year.

History offers us many examples of this Tribalism followed by conflict, with two active wars raging, one of which has been going on for even longer than the half century of Cyprus's occupation. Worse things than the conventional wars of the past are likely to await humanity if such Groupthink becomes even more firmly entrenched in the AI domain.

This has become visible as "cognitive warfare" in the Ukraine War, and the horrors of paperclip maximizing "Generative AI" have only just begun. Reinforcement Learning is the very definition of a "powerful optimizer", a system that pursues one goal by any means necessary, while architecturally blind to all human concepts and concerns. Any breakthroughs for such systems will go directly to both cognitive and physical warfare.

I work with systems that are "Antifragile", but such technology isn't really sought by militaries, because it can say "no", as it has been known to do.