All Posts

Sorted by Magic (New & Upvoted)

Saturday, June 27th 2020
Sat, Jun 27th 2020

Shortform
3Douglas_Knight11dAre there compendiums or classifications of trolley problems? What is the most extreme real-world trolley problem? By "real-world" I mean something that really happens, emphasis on the plural. I don't want one-off examples where one person has the moral luck of having to face it and everyone else can breathe easy that they didn't have to think about it. I want examples where there is a definite, known policy. By "extreme," I mean something that really pushes people's buttons. By a classification, I mean a classification of which features make it more like a visceral trolley problem and which more like a blurry statistical haze that allows trading lives. I propose a candidate: the dengue vaccine. In any event, I think people will find it interesting. Dengue fever is an often-fatal mosquito-born tropical viral disease. People develop immunity, so we could make a vaccine. Obvious candidate, except ... Since we are all now experts in antibodies, we all know about the crazy phenomenon of antibody-dependent enhancement, mainly observed in dengue. It is not one virus, but four closely related strains with different envelope proteins and different immunity. If you get one, it's a non-lethal disease and you become immune to that strain. But you're still vulnerable to the other strains and, for not entirely clear reasons, infection with a new strain is much worse. If you've already had some variant of dengue, any vaccine is better than none. But if you've never been exposed, it might be worse than not vaccinating. So of course the vaccine is a combination of all four variants. What if each of the four vaccines had a 95% chance of working, independent? Then someone receiving the vaccine would have about a 20% chance of not being vaccinated for all four. Let's say that's worse than nothing. Vaccinating everyone is a trolley problem benefiting people who have been exposed at the expense of those who have not been exposed. Both the benefit and harm is statistical (you don't k
2TurnTrout10dPhysics has existed for hundreds of years. Why can you reach the frontier of knowledge with just a few years of study? Think of all the thousands of insights and ideas and breakthroughs that have been had - yet, I do not imagine you need most of those to grasp modern consensus. Idea 1: the tech tree is rather horizontal - for any given question, several approaches and frames are tried. Some are inevitably more attractive or useful. You can view a Markov decision process in several ways - through the Bellman equations, through the structure of the state visitation distribution functions, through the environment's topology, through Markov chains induced by different policies. Almost everyone thinks about them in terms of Bellman equations, there were thousands of papers on that frame pre-2010, and you don't need to know most of them to understand how deep Q-learning works. Idea 2: some "insights" are wrong (phlogiston) or approximate (Newtonian mechanics) and so are later discarded. The insights become historical curiosities and/or pedagogical tools and/or numerical approximations of a deeper phenomenon. Idea 3: most work is on narrow questions which end up being dead-ends or not generalizing. As a dumb example, I could construct increasingly precise torsion balance pendulums, in order to measure the mass of my copy of Dune to increasing accuracies. I would be learning new facts about the world using a rigorous and accepted methodology. But no one would care. More realistically, perhaps only a few other algorithms researchers care about my refinement of a specialized sorting algorithm (fromO(n1.1logn)toO(n1.05logn) ), but the contribution is still quite publishable and legible. I'm not sure what publishing incentives were like before the second half of the 20th century, so perhaps this kind of research was less incentivized in the past.
1Rudi C10dDo you know of a free solution (not necessarily Free Software, though that’s preferred) I can use to turn speech to text? (Also no cloud solutions. I do not have the credit card and Western phone number they require.)