Main Battle Tanks can simultaneously be on the way to battlefield obsolescence and be important for Ukraine's current military operations. The vulnerability of battleships to air power was identified by 1914 yet they were central to WW1 and 2 navies... doctrine lags technology.

A good read on an important development (is this the proliferation ripple?)

We thought about learning from this, but platforms are just so easy to count...

Even though the selectorate in Russia is quite small, this is still important scholarship. If Russian elites think like the Russian public (as evidence suggests is true of US elites/public) then it sure looks like the Russians love their children too.

The fetishization of expensive platforms is so deeply ingrained that we're unlikely to adapt to the dynamics identified here until forced by the bitter taste of defeat. But how much agility can we really expect from an org that only promotes from within? foreignpolicy.com/2023/02/02/us-…

Excellent thread. If you want to resolve really fine detail on the ground and have a way to steer them, balloons are way more capable than expensive & diffraction-limited satellites.

Unraveling the intended and unintended signals sent by incidents like this is often the most interesting bit.

I agree with the premise, but where's the second half of this article? We need more than just the vague suggestion that the US "should complement complex, high-cost technology with cheaper autonomous sensors, communications relays, munitions, and decoys" foreignaffairs.com/ukraine/does-t…

Yeah. Mindreading. And this is early early alpha. But emerging tech is just incremental change, right?

Everything is fine. telegraph.co.uk/technology/202…

Russia will place tactical nuclear weapons in Belarus ria.ru/20230325/khran…

...the most likely result of building a superhumanly smart AI is that literally everyone on Earth will die. Not as in “maybe possibly some remote chance,” but as in “that is the obvious thing that would happen.” time.com/6266923/ai-eli…

For the price of one F-35 a year you could hire 200 absolutely top-tier CS experts to work on cyber. Which option would have a larger marginal impact on state security?

Edward Teller was wrong about nuclear risk in 1942, so hundreds of AI experts are wrong about AI risks today. This is the sort of inane argument from anecdote that feeds the death of expertise.

Excellent and helpful — actual answers to a lot of "well, trust could go either way..." debates

Very interesting thread; not surprising but still concerning. How secure is Russia's second strike? Guess that depends if someone substituted cheap parts there too...

Just came across this cool study confirming a long-held assumption of mine that humans trust AIs over other humans when weighing the answers to objective questions under conditions of uncertainty. sciencedirect.com/science/articl…

Useful paper, though it handwaves 'structural risks' (admittedly w/ a cite to Zwetsloot & Dafoe), further demonstrating the need for better CS/IR integration.