Links for 2024-06-23
Quote
To contemn the world and to enjoy the world are things contrary to each other. How, then can we contemn the world, which we are born to enjoy? Truly there are two worlds. One was made by God, the other by men. That made by God was great and beautiful. Before the Fall it was Adam’s joy and the Temple of his Glory. That made by men is a Babel of Confusions: Invented Riches, Pomps and Vanities, brought in by Sin: Give all (saith Thomas à Kempis) for all. Leave the one that you may enjoy the other.
-Thomas Traherne, Centuries of Meditation I.1
Links
Situational Awareness: The Decade Ahead by Leopold Aschenbrenner - I’m undecided about how imminent AGI (Artificial General Intelligence - defined here as a drop in remote worker capable of taking on any intellectual task as well as the median human) is, or if such a thing is even technically possible (it all depends on how you define it), but … I use a piece software every single day that I would not have believed was possible two years ago if you had told me. Aschenbrenner, a former member of OpenAI’s safety team, carefully and persuasively makes the case that AGI is a sooner rather than later thing. I hope to write more about this in the future, but you owe it to yourself to carefully consider Aschenbrenner’s arguments: even if you disagree, you’ll understand where and why you disagree, rather than just having a reflexive “impossible” judgment like I used to.
Of all my links in this post, this is my one “must read.” I don’t know that I will always have them, but this is one of them.Real Engineering: The Insane Engineering of the F-117 Nighthawk (video) - In depth treatment of the engineering of the F-117. Lots of background on the increasingly threatening surface-to-air radar and missile environment that made stealth necessary to preserve US air superiority. Of particular interest: the initial physical theory about stealth materials was made by physicists in the Soviet Union (the country making the air defenses that threatened US air superiority) and the Kremlin somehow allowed to be published in the open-source literature.
(For the curious: Method of Edge Waves in the Physical Theory of Diffraction by Pyotr Ufimtsev, translated by the U.S. Air Forces Foreign Technology Division)Anders Puck Nielsen: Escalation management and Biden’s strategy for Ukraine (video) - Nielsen is one of the most informative commentators on Russia’s war in Ukraine and the collective West’s response to it. In this video he describes the thinking by the US administration that keeps the United States policy stuck at “Russia Must Lose…” but not quite willing to commit to “… And Ukraine Must Win”
Jeremy Parrish: In-console-able rage: Teenage Mutant Ninja Turtles | NES Works 131 - another installment in Parrish long-running chronicling of every … single … game … released for the Nintendo Entertainment System during its lifetime. These videos are more than nostalgia trips or retro game reviews. They are thoughtful reflections on how mediums are shaped by creativity, technical limitations, and commercial and cultural factors, and how that translates into the cultural artifacts that defined so many people’s childhoods. He also has a good deadpan sense of humor.
alexander the ok: The Space Shuttle: A $200 Billion Lesson in Risk Management (video, 1 hour) - Well done retrospective on the Space Shuttle, a complement to videos released by Real Engineering and Scott Manley. This particular video focuses on what its title says - risk management - and is a good overview of Diane Vaughan’s conclusions from her seminal book The Challenger Launch Decision.
Inside the Aquarium by Victor Suvorov - I am currently reading this, a riveting account of life inside the GRU, the Soviet Union’s (and now the Russian Federation’s) military intelligence directorate, written by a defector.
A sample:Connections and Tangents
Situational Awareness >< The Insane Engineering of the F-117 Nighthawk
Looping back to the Situational Awareness book, Real Engineering draws a lot of attention to the fact the F-117 was built because of a vulnerability: the Soviets had created lethal surface-to-air systems that could have nullified the US air power advantage. Aschenbrenner similarly draws out a potential race for AI technology: one lab makes a breakthrough that is, for a given definition, an Artificial General Intelligence. This AGI is able to function as an automated researcher, and not just one: potentially thousands of them running in parallel and passing ideas, code, experiments back and forth. This would be, for the major power that doesn’t have such a capability, a new vulnerability - imagine automating research into weapons and defensive systems. In the last chapter of the book, Aschenbrenner describes a potential race for AI supremacy between the United States and China.
The interesting difference, of course, is that the US Air Force knew that the stealth bombers they built would have a limited-service lifetime: once the capability was revealed, an enemy would be able to create a check to it. People thinking about AGI and the potential for Superintelligence beyond it believe that the Superintelligence would be a checkmate for any opponent: once you had it, you could dictate your terms. Past experience should lead us to doubt this would be the case. History never ends.