Briar Patch Observatory Podcast

• https://briarpod.net/. https://twitter.com/PLanghaus
• Pedro Domingues - seminal book on machine learning.
• George Hotz - open source people. Open source self-driving cars.
• Centralized vs decentralization - some things work better than not in a decentralized. We must observe - what emerges? Does an army work better centralized? Decent. Nimbler. Centralized command can be slow. Decentralized army does not end up winning a war. You want centralized decision making, but communication about lessons learned.
• Where we organize centralized or decent - https://en.wikipedia.org/wiki/George_Hotz. Post-modernist - don't believe in grand theories.
• Individual sovereignity - a technical problem. If I have a box that makes food, water, shelter - why do you need the state?
• Replace the state - better technical not ideological. State loses, freedom wins. Focus as a technical question not political one.
• What about protection? Defense wins over offense.
• Whether offense or defense wins - depends on time. In military history - it shifted. Defense may work.
• Why we need to defend ourselves.
• George dissociates from power. If you have
• You may not be interested in politics, but politics are interested in you - Trotsky.
• Algorithms. China does AI for oppression.
• Desire slaves!
• Ask everybody if they would want to be the king of the world? Oppression is very costly.
• Democracy -
• Algorithms.
• Pedro - constitution was designed so idiots could run it.
• Ex. Separtion of power vs checks and balances. Sharing power so no branch could dominate others.
• Pedro - not good enough if you have small fiefdoms of freedom in a centralized world. Dynamics that made it happen - are positive feedback loop for centralization.
• Centralized platform also provides freedom for decentralization - small businesses, etc. Amazon sellers.
• Platforms are centralized. But opens up opportunity for decentralization.
• Where does the platform end, and user begin?
• Platforms as monopolies - they are not that big. Ex. Vampire vs Twitter - $1B to switch over. George is not worrried about platform dominance.
• Network effects - can transcend by incentivizing other platforms?
• To take over a platform - is very difficult. A better search engine will not take down Google. It is data that backs up the old one.
• Ex. Technically - Bing is technically as good, but Google has more data.
• Twitter survived many attempts by management to take it down. You have huge inertia.
• Bing - tried an NLP engine. Tech wasn't there yet.
• There's a limit to how shitty google can become.
• Ex. Page Rank - made Google successful. This was a decentralized mechanism.
• Google will lose dominance at some point. Very vulnerable in ways. 1. They tried machine learning - cornering the market. Didn't succeed. Someone will come out with a 100x better machine learning algorithm.
• 2 - another mortal threat. Ad-blockers vs Ad-sellers. Ad blockers will get an upper hand at one point, collapsing the system.