The Nonlinear Library: LessWrong

Link to original articleWelcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Why you should learn a musical instrument, published by cata on May 16, 2024 on LessWrong. I have liked music very much since I was a teenager. I spent many hours late at night in Soulseek chat rooms talking about and sharing music with my online friends. So, I tend to just have some music floating around in my head on any given day. But, I never learned to play any instrument, or use any digital audio software. It just didn't catch my interest. My wife learned to play piano as a kid, so we happen to have a keyboard sitting around in our apartment. One day I was bored so I decided to just see whether I could figure out how to play some random song that I was thinking about right then. I found I was easily able to reconstitute a piano version of whatever melody I was thinking of, just by brute-forcing which notes were which, given a lot of patience. So that was satisfying enough that I wanted to keep doing it. What I didn't know is how immediately thought-provoking it would be to learn even the most basic things about playing music. Maybe it's like learning to program, if you used a computer all the time but you never had one thought about how it might work. Many of the things I learned immediately that surprised me were about my perception of the music I had listened to for all of my life. In my mind, my subjective experience of remembering music that I am very familiar with seems very vivid. I feel like I can imagine all the instruments and imagine all the sounds, just like they were in the song. But once I had to reconstruct the music myself, it quickly became clear that I was tricking myself in a variety of ways. For example, my memory of the main melody would be very clear. But my memory of any harmony or accompaniment was typically totally vague. I absolutely could not reconstruct something to play with my left hand on the piano, because I wasn't actually remembering it; I was just remembering something more abstract, I guess. Sometimes I would be convinced I would remember a melody and reproduce it on the keyboard, but then I would listen to the real song and be surprised. The most common way I got surprised was that in my memory, I had adjusted it so that I could physically sing or hum it, even though I don't often sing. If there was a big jump up or down the scale, I would do something in my memory that sounded sort of OK instead, like replace it with a repeated note, or the same thing moved an octave, and then forget that it had ever been any other way. I found that if I was remembering something that had fast playing, I often actually could not remember the specific notes in between beats, even though I felt that I could hear it in my head. No matter how hard I "focused" on my memory I couldn't get more detail. Actually, I found that there was some speed such that even listening to the music, I could no longer resolve the individual notes, no matter how hard I paid attention or how many times I replayed it. There have been many more kinds of things I have learned since learning to play a little: Since playing music on a keyboard is a complicated physical task involving complicated coordination, I learned a lot about what both of my hands are naturally good and bad at, and what sort of things they can coordinate easily or poorly.[1] Learning the musical structure of songs that I know and trying to arrange them for piano showed me all kinds of self-similarity and patterns inside the songs that I had never had a clue about before. I could listen to a song hundreds of times and not realize, for example, that two parts of the song were the same phrase being played on two different instruments in a very slightly different way. Often I will be trying to learn to play something using one "technique" for learning and practicing it, and having a hard time, and then I...

Link to original articleWelcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Do you believe in hundred dollar bills lying on the ground? Consider humming, published by Elizabeth on May 16, 2024 on LessWrong. Introduction [Reminder: I am an internet weirdo with no medical credentials] A few months ago, I published some crude estimates of the power of nitric oxide nasal spray to hasten recovery from illness, and speculated about what it could do prophylactically. While working on that piece a nice man on Twitter alerted me to the fact that humming produces lots of nasal nitric oxide. This post is my very crude model of what kind of anti-viral gains we could expect from humming. I've encoded my model at Guesstimate. The results are pretty favorable (average estimated impact of 66% reduction in severity of illness), but extremely sensitive to my made-up numbers. Efficacy estimates go from ~0 to ~95%, depending on how you feel about publication bias, what percent of Enovid's impact can be credited to nitric oxide, and humming's relative effect. Given how made up speculative some of these numbers are, I strongly encourage you to make up speculate some numbers of your own and test them out in the guesstimate model. If you want to know how nitric oxide reduces disease, check out my original post. Math Estimating the impact of Enovid I originally estimated the (unadjusted) efficacy of nitric oxide nasal sprays after diagnosis at 90% overall reduction in illness, killing ~50% of viral particles per application. Enovid has three mechanisms of action. Of the papers I looked at in that post, one mentioned two of the three (including nitric oxide) a second mechanism but not the third, and the other only mentioned nitric oxide. So how much of theat estimated efficacy is due to nitric oxide alone? I don't know, so I put a term in the guesstimate with a very wide range. I set the lower bound to (one of three mechanisms) to 1 (if all effect was due to NO). There's also the question of how accurate the studies I read are. There are only two, they're fairly small, and they're both funded by Enovid's manufacturer. One might reasonably guess that their numbers are an overestimate. I put another fudge factor in for publication bias, ranging from 0.01 (spray is useless) to 1 (published estimate is accurate). How much nitric oxide does Enovid release? This RCT registration uses a nitric oxide nasal spray (and mentions no other mechanisms). They don't give a brand name but it's funded by the company that produces Enovid. In this study, each application delivers 0.56 mL of nitric oxide releasing solution (NORS) (this is the same dose you get from commercial Enovid), which delivers "0.11ppm [NO]*hrs". There's a few things that confusing phrase could mean: The solution keeps producing 0.11ppm NO for several hours (very unlikely). The application produces 0.88ppm NO almost immediately (0.11*8, where 8 hours is the inter-application interval), which quickly reacts to form some other molecule. This is my guess, and what I'll use going forward. It won't turn out to matter much. Some weirder thing. How much nitric oxide does humming move into the nose? Here we have much more solid numbers. NO concentration is easy to measure. Individuals vary of course, but on average humming increases NO concentration in the nose by 15x-20x. Given baseline levels of (on average) 0.14ppm in women and 0.18ppm in men, this works out to a 1.96-3.42 ppm increase. More than twice what Enovid manages. The dominant model is that the new NO in the nose is borrowed from the sinuses rather than being newly generated. Even if this is true I don't think it matters; sinus concentrations are 100x higher than the nose's and replenish quickly. Estimating the impact of humming As far as I can find, there are no published studies on humming as an antimicrobial intervention. There is lots of circumstantial evid...

Link to original articleWelcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: MIRI's May 2024 Newsletter, published by Harlan on May 15, 2024 on LessWrong. MIRI updates: MIRI is shutting down the Visible Thoughts Project. We originally announced the project in November of 2021. At the time we were hoping we could build a new type of data set for training models to exhibit more of their inner workings. MIRI leadership is pessimistic about humanity's ability to solve the alignment problem in time, but this was an idea that seemed relatively promising to us, albeit still a longshot. We also hoped that the $1+ million bounty on the project might attract someone who could build an organization to build the data set. Many of MIRI's ambitions are bottlenecked on executive capacity, and we hoped that we might find individuals (and/or a process) that could help us spin up more projects without requiring a large amount of oversight from MIRI leadership. Neither hope played out, and in the intervening time, the ML field has moved on. (ML is a fast-moving field, and alignment researchers are working on a deadline; a data set we'd find useful if we could start working with it in 2022 isn't necessarily still useful if it would only become available 2+ years later.) We would like to thank the many writers and other support staff who contributed over the last two and a half years. Mitchell Howe and Joe Rogero joined the comms team as writers. Mitch is a longtime MIRI supporter with a background in education, and Joe is a former reliability engineer who has facilitated courses for BlueDot Impact. We're excited to have their help in transmitting MIRI's views to a broad audience. Additionally, Daniel Filan will soon begin working with MIRI's new Technical Governance Team part-time as a technical writer. Daniel is the host of two podcasts: AXRP, and The Filan Cabinet. As a technical writer, Daniel will help to scale up our research output and make the Technical Governance Team's research legible to key audiences. The Technical Governance Team submitted responses to the NTIA's request for comment on open-weight AI models, the United Nations' request for feedback on the Governing AI for Humanity interim report. and the Office of Management and Budget's request for information on AI procurement in government. Eliezer Yudkowsky spoke with Semafor for a piece about the risks of expanding the definition of "AI safety". "You want different names for the project of 'having AIs not kill everyone' and 'have AIs used by banks make fair loans." A number of important developments in the larger world occurred during the MIRI Newsletter's hiatus from July 2022 to April 2024. To recap just a few of these: In November of 2022, OpenAI released ChatGPT, a chatbot application that reportedly gained 100 million users within 2 months of its launch. As we mentioned in our 2024 strategy update, GPT-3.5 and GPT-4 were more impressive than some of the MIRI team expected, representing a pessimistic update for some of us "about how plausible it is that humanity could build world-destroying AGI with relatively few (or no) additional algorithmic advances". ChatGPT's success significantly increased public awareness of AI and sparked much of the post-2022 conversation about AI risk. In March of 2023, the Future of Life Institute released an open letter calling for a six-month moratorium on training runs for AI systems stronger than GPT-4. Following the letter's release, Eliezer wrote in TIME that a six-month pause is not enough and that an indefinite worldwide moratorium is needed to avert catastrophe. In May of 2023, the Center for AI Safety released a one-sentence statement, "Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war." We were especially pleased with this statement, because it focused attention ...

Link to original articleWelcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Catastrophic Goodhart in RL with KL penalty, published by Thomas Kwa on May 15, 2024 on LessWrong. TLDR: In the last two posts, we showed that optimizing for a proxy can fail to increase true utility, but only when the error is heavy-tailed. We now show that this also happens in RLHF with a KL penalty. This post builds on our earlier result with a more realistic setting and assumptions: Rather than modeling optimization as conditioning on a minimum reward threshold, we study maximization of reward with a KL divergence penalty, as in RLHF. We remove the assumption of independence between the error and utility distributions, which we think was the weakest part of the last post. When the true utility V is light-tailed, the proxy can be maximized while keeping E[V]to the same level as the prior. We can't guarantee anything about E[V] when V is heavy tailed; it could even go to minus infinity. Abstract When applying KL regularization, the trained model is regularized towards some prior policy π0. One would hope that a KL penalty can produce good outcomes even in the case of reward misspecification; that is, if the reward U is the sum of true utility V and an error term X, we would hope that optimal policies under a KL penalty achieve high V even if the magnitude of X is large. We show that this is not always the case: when X is heavy-tailed, there are arbitrarily well-performing policies π with Eπ[V]Eπ0[V]; that is, that get no higher true utility than the prior. However, when error is light-tailed and independent of V, the optimal policy under a KL penalty results in V>0, and V can be made arbitrarily large. Thus, the tails of the error distribution are crucial in determining how much utility will result from optimization towards an imperfect proxy. Intuitive explanation of catastrophic Goodhart with a KL penalty Recall that KL divergence between two distributions P and Q is defined as If we have two policies π,π0, we abuse notation to define DKL(ππ0) as the KL divergence between the distributions of actions taken on the states in trajectories reached by π. That is, if Tr(π) is the distribution of trajectories taken by π, we penalize This strongly penalizes π0 taking actions the base policy never takes, but does not force the policy to take all actions the base policy takes. If our reward model gives reward U, then the optimal policy for RLHF with a KL penalty is: Suppose we have an RL environment with reward U=X+V where X is an error term that is heavy-tailed under π0, and V is the "true utility" assumed to be light-tailed under π0. Without loss of generality, we assume that E[U(π0)]=0. If we optimize for E[U(π)]βDKL(ππ0), there is no maximum because this expression is unbounded. In fact, it is possible to get E[U(π)]>M and DKL(π,π0) For such policies π, it is necessarily the case that limϵ0E[V(π)]=0; that is, for policies with low KL penalty, utility goes to zero. Like in the previous post, we call this catastrophic Goodhart because the utility produced by our optimized policy is as bad as if we hadn't optimized at all. This is a corollary of a property about distributions (Theorems 1 and 3 below) which we apply to the case of RLHF with unbounded rewards (Theorem 2). The manner in which these pathological policies π achieve high E[U] is also concerning: most of the time they match the reference policy π0, but a tiny fraction of the time they will pick trajectories with extremely high reward. Thus, if we only observe actions from the policy π, it could be impossible to tell whether π is Goodharting or identical to the base policy. Results All proofs are in the appendix, which will be published shortly after this post. X heavy tailed, V light tailed: EV0 We'll start by demon...

Link to original articleWelcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Teaching CS During Take-Off, published by andrew carle on May 15, 2024 on LessWrong. I stayed up too late collecting way-past-deadline papers and writing report cards. When I woke up at 6, this anxious email from one of my g11 Computer Science students was already in my Inbox. Student: Hello Mr. Carle, I hope you've slept well; I haven't. I've been seeing a lot of new media regarding how developed AI has become in software programming, most relevantly videos about NVIDIA's new artificial intelligence software developer, Devin. Things like these are almost disheartening for me to see as I try (and struggle) to get better at coding and developing software. It feels like I'll never use the information that I learn in your class outside of high school because I can just ask an AI to write complex programs, and it will do it much faster than I would. I'd like to know what your thoughts on this are. Do you think AI will replace human software developers, as NVIDIA claims it will? My response: Buddy, that is a big question for 5:15 am. First AI horizon thoughts: 1. Software development as a field will look incredibly different in 10 years. 2. My priors say that MOST of human intellectual+economic activity will ALSO be radically different in 10 years. 3. I have a very small p(doom) for the 10 year horizon. That means I don't expect human-equivalent AGIs to completely disrupt human civilisation within 10 years. 4. The delta between how fast AI will affect software engineering and how fast AI will transform other (roughly speaking) white collar careers is relatively small. That means I think the AI affect on say, hedge fund management and software engineering to be similar. Then some priors I have for teaching IB Computer Science in the middle of this take-off: 1. I don't think becoming a software engineer is the modal outcome for IBCS students 2. I believe that most long term personal utility from IBCS (or any other intro CS exposure) comes from shifting a student's mental model of how the modern social and economic system interacts with / depends on these technologies. 3. While the modern Ai tools are light years beyond the simple Von Neumann CPU models and intro Python we're studying, it does address the foundations of those systems. Similarly, HL Analysis and HL Physics don't cover anything about the math and physics that underpin these huge ML systems, but that foundation IS there. You can't approach the superstructure without it. So, in summary, if your concern is "the world seems to be changing fast. This class is hard, and I don't think there's any chance that I will find a 2022 Novice SoftwareDev job when I'm out of university in 2029" I would strongly agree with that sentiment. I have a Ron Swanson detachment on the important of formal schooling. If your question was "is a traditional education sequence the best way to prepare myself for the turbulent AI takeoff period," then I strongly disagree with that statement. Education is intrinsically reflective and backward looking. But I'm employed as a high school teacher. And your parents have decided to live here and send you to this school . So, I'm not sure if advice on that axis is actionable for either of us. There's also a huge chasm between "this isn't be best of all possible options" and "this has zero value." If I reframed your statement as "given that I'm in this limited option IB program, what classes will provide me the best foundation to find opportunities and make novel insights in the turbulent AI takeoff period" I would feel confident recommending IBCS. That doesn't make learning to code any easier. Is that a good answer to a 17 year old? Is there a good answer to this? One of the best parts of teaching is watching young people wake up to the real, fundamental issues and challenges of human civilisation an...

Link to original articleWelcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Ilya Sutskever and Jan Leike resign from OpenAI, published by Zach Stein-Perlman on May 15, 2024 on LessWrong. Ilya Sutskever and Jan Leike have resigned. They led OpenAI's alignment work. Superalignment will now be led by John Schulman, it seems. Jakub Pachocki replaced Sutskever as Chief Scientist. Reasons are unclear (as usual when safety people leave OpenAI). The NYT piece and others I've seen don't really have details. Archive of NYT if you want to read it anyway. OpenAI announced Sutskever's departure in a blogpost. Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org

Link to original articleWelcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: How to do conceptual research: Case study interview with Caspar Oesterheld, published by Chi Nguyen on May 14, 2024 on LessWrong. Caspar Oesterheld came up with two of the most important concepts in my field of work: Evidential Cooperation in Large Worlds and Safe Pareto Improvements. He also came up with a potential implementation of evidential decision theory in boundedly rational agents called decision auctions, wrote a comprehensive review of anthropics and how it interacts with decision theory which most of my anthropics discussions built on, and independently decided to work on AI some time late 2009 or early 2010. Needless to say, I have a lot of respect for Caspar's work. I've often felt very confused about what to do in my attempts at conceptual research, so I decided to ask Caspar how he did his research. Below is my writeup from the resulting conversation. How Caspar came up with surrogate goals The process Caspar had spent six months FTE thinking about a specific bargaining problem between two factions with access to powerful AI, spread over two years. A lot of the time was spent on specific somewhat narrow research projects, e.g. modelling the impact of moral advocacy in China on which bargaining problems we'll realistically encounter in the future. At the time, he thought those particular projects were important although he maybe already had a hunch that he wouldn't think so anymore ten years down the line. At the same time, he also spent some time on most days thinking about bargaining problems on a relatively high level, either in discussions or on walks. This made up some double digit percentage of his time spent researching bargaining problems. Caspar came up with the idea of surrogate goals during a conversation with Tobias Baumann. Caspar describes the conversation leading up to the surrogate goal idea as "going down the usual loops of reasoning about bargaining" where you consider just building values into your AI that have properties that are strategically advantaged in bargaining but then worrying that this is just another form of aggressive bargaining. The key insight was to go "Wait, maybe there's a way to make it not so bad for the other side." Hence, counterpart-friendly utility function modifications were born which later on turned into surrogate goals. Once he had the core idea of surrogate goals, he spent some time trying to figure out what the general principle behind "this one weird trick" he found was. Thus, with Vincent Conitzer as his co-author, his SPI paper was created and he continues trying to answer this question now. Caspar's reflections on what was important during the process He thinks it was important to just have spent a ton of time, in his case six months FTE, on the research area. This helps with building useful heuristics. It's hard or impossible and probably fruitless to just think about a research area on an extremely high level. "You have to pass the time somehow." His particular projects, for example researching moral advocacy in China, served as a way of "passing the time" so to say. At the same time, he thinks it is both very motivationally hard and perhaps not very sensible to work on something that's in the roughly right research area where you really can't see a direct impact case. You can end up wasting a bunch of time grinding out technical questions that have nothing much to do with anything. Relatedly, he thinks it was really important that he continued doing some high-level thinking about bargaining alongside his more narrow projects. He describes a common dynamic in high-level thinking: Often you get stuck on something that's conceptually tricky and just go through the same reasoning loops over and over again, spread over days, weeks, months, or years. You usually start entering the loop because you think...

Link to original articleWelcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: How To Do Patching Fast, published by Joseph Miller on May 14, 2024 on LessWrong. This post outlines an efficient implementation of Edge Patching that massively outperforms common hook-based implementations. This implementation is available to use in my new library, AutoCircuit, and was first introduced by Li et al. (2023). What is activation patching? I introduce new terminology to clarify the distinction between different types of activation patching. Node Patching Node Patching (aka. "normal" activation patching) is when some activation in a neural network is altered from the value computed by the network to some other value. For example we could run two different prompts through a language model and replace the output of Attn 1 when the model is given some input 1 with the output of the head when the model is given some other input 2. We will use the running example of a tiny, 1-layer transformer, but this approach generalizes to any transformer and any residual network. All the nodes downstream of Attn 1 will be affected by the patch. Edge Patching If we want to make a more precise intervention, we can think about the transformer differently, to isolate the interactions between components. Now we can patch the edge Attn 1 -> MLP and only nodes downstream of MLP will be affected (eg. Attn 1->Output is unchanged). Edge Patching has not been explicitly named in any prior work. Path Patching Path Patching refers to the intervention where an input to a path is replaced in the 'treeified' view of the model. The treeified view is a third way of thinking about the model where we separate each path from input to output. We can implement an equivalent intervention to the previous diagram as follows: In the IOI paper, 'Path Patching' the edge Component 1 -> Component 2 means Path Patching all paths of the form where all components between Component 1 and Component 2 are MLPs[1]. However, it can be easy to confuse Edge Patching and Path Patching because if we instead patch all paths of the form this is equivalent to Edge Patching the edge Component 1->Component 2. Edge Patching all of the edges which have some node as source is equivalent to Node Patching that node. AutoCircuit does not implement Path Patching, which is much more expensive in general. However, as explained in the appendix, Path Patching is sometimes equivalent to Edge Patching. Fast Edge Patching We perform two steps. First we gather the activations that we want to patch into the model. There's many ways to do this, depending on what type of patching you want to do. If we just want to do zero ablation, then we don't need to even run the model. But let's assume we want to patch in activations from a different, corrupt input. We create a tensor, Patch Activations, to store the outputs of the source of each edge and we write to the tensor during the forward pass. Each source component has a row in the tensor, so the shape is [n_sources, batch, seq, d_model].[2] Now we run the forward pass in which we actually do the patching. We write the outputs of each edge source to a different tensor, Current Activations, of the same shape as Patch Activations. When we get to the input of the destination component of the edge we want to patch, we add the difference between the rows of Patch Activations and Current Activations corresponding to the edge's source component output. This works because the difference in input to the edge destination is equal to the difference in output of the source component.[3] Now it's straightforward to extend this to patching multiple edges at once by subtracting the entire Current Activations tensor from the entire Patch Activations tensor and multiplying by a Mask tensor of shape [n_sources] that has a single value for each input edge. By creating a Mask tensor for each destination node w...

Link to original articleWelcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: D&D.Sci Long War: Defender of Data-mocracy Evaluation & Ruleset, published by aphyer on May 14, 2024 on LessWrong. This is a follow-up to last week's D&D.Sci scenario: if you intend to play that, and haven't done so yet, you should do so now before spoiling yourself. There is a web interactive here you can use to test your answer, and generation code available here if you're interested, or you can read on for the ruleset and scores. RULESET Each alien has a different amount of HP: Alien HP Threat* Swarming Scarab 1 1 Chitinous Crawler 3 2 Voracious Venompede 5 3 Arachnoid Abomination 9 4 Towering Tyrant 15 5 *Threat has no effect on combat directly - it's a measure of how threatening Earth considers each alien to be, which scales how many soldiers they send. (The war has been getting worse - early on, Earth sent on average ~1 soldier/4 Threat of aliens, but today it's more like 1 soldier/6 Threat. The wave you're facing has 41 Threat, Earth would send on average ~7 soldiers to it. Earth doesn't exercise much selection with weapons, but sends soldiers in pairs such that each pair has two different weapons - this is a slight bias towards diversity.) Each weapon has a damage it deals per shot, and a rate of fire that determines how many shots it can get off before the wielder is perforated by venomous spines/dissolved into a puddle of goo/voraciously devoured by a ravenous toothed maw: Weapon Damage Min Shots Max Shots Macross Minigun 1 5 8 Fusion Flamethrower 1 3 12 Pulse Phaser 2 4 6 Rail Rifle 3 3 5 Laser Lance 5 2 5 Gluon Grenades 7 2 3 Thermo-Torpedos 13 1 3 Antimatter Artillery 20 1 2 Each soldier will be able to fire a number of shots chosen randomly between Min Shots and Max Shots - for example, a soldier with a Laser Lance will have time to fire 1d4+1 shots, each doing 5 damage. During a battle, humans roll for how many shots each weapon gets, and then attempt to allocate damage from their shots to bring down all aliens. If they succeed, the humans win - if not, the humans lose. While doing this optimally is theoretically very difficult, your soldiers are well-trained and the battles are not all that large, so your soldiers will reliably find a solution if one exists. For example, if you are fighting two Towering Tyrants and two Swarming Scarabs using two soldiers: If you bring one soldier with Antimatter Artillery and one with a Macross Minigun, the Minigun soldier will reliably kill the Scarabs and have 3-6 shots left over (not enough to kill a Tyrant). The Artillery soldier will get either 1 or 2 shots: half the time they will roll a 2, kill both Tyrants and you will win, while the other half they will roll a 1, a Tyrant will survive and you will lose. You can do a little better by bringing one soldier with Antimatter Artillery and one with a Laser Lance. The Laser Lance rolls 2-5 shots - it will always kill both Scarabs, and 1/4 of the time it will roll 5 shots and also be able to kill a Tyrant (at which point you'll win even if the Antimatter Artillery rolls a 1), giving you a 5/8 winrate overall. You can do better still by bringing one soldier with Thermo-Torpedos and one with a Pulse Phaser. The Phaser soldier gets at least 4 shots, with which they kill both Scarabs and do 2 damage to each Tyrant (dropping the Tyrants both to 13 HP). And the Torpedo soldier gets 1-3 shots, with a 2/3 chance of being able to kill both Tyrants now that they've been softened up. I believe this is the best winrate you can get in this example. STRATEGY The most important element of strategy was sending the right kind of weapons for each alien: high-health aliens like Tyrants are extremely inefficient to kill with light weapons like Miniguns, while small, numerous aliens like Scarabs are extremely inefficient to kill with heavy weapons like artillery. There were a few subtler ...

Link to original articleWelcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Against Student Debt Cancellation From All Sides of the Political Compass, published by Maxwell Tabarrok on May 14, 2024 on LessWrong. A stance against student debt cancellation doesn't rely on the assumptions of any single ideology. Strong cases against student debt cancellation can be made based on the fundamental values of any section of the political compass. In no particular order, here are some arguments against student debt cancellation from the perspectives of many disparate ideologies. Equity and Fairness Student debt cancellation is a massive subsidy to an already prosperous and privileged population. American college graduates have nearly double the income of high school graduates. African Americans are far underrepresented among degree holders compared to their overall population share. Within the group of college graduates debt cancellation increases equity, but you can't get around the fact that 72% of African Americans have no student debt because they never went to college. The tax base for debt cancellation will mostly come from rich white college graduates, but most of the money will go to … rich white college graduates. Taxing the rich to give to the slightly-less-rich doesn't have the same Robin Hood ring but might still slightly improve equity and fairness relative to the status quo, except for the fact that it will trade off with far more important programs. Student debt cancellation will cost several hundred billion dollars at least, perhaps up to a trillion dollars or around 4% of GDP. That's more than defense spending, R&D spending, more than Medicaid and Medicare, and almost as much as social security spending. A trillion-dollar transfer from the top 10% to the top 20% doesn't move the needle much on equity but it does move the needle a lot on budgetary and political constraints. We should be spending these resources on those truly in need, not the people who already have the immense privilege on an American college degree. Effective Altruism The effective altruist critique of student debt cancellations is similar to the one based on equity and fairness, but with much more focus on global interventions as an alternative way to spend the money. Grading student debt cancellation on impact, tractability, and neglectedness, it scores very poorly. Mostly because of tiny impact compared to the most effective charitable interventions. Giving tens of thousands of dollars to people who already have high incomes, live in the most prosperous country on earth, and face little risk of death from poverty or disease is so wasteful that it borders on criminal on some views of moral obligations. It is letting tens of millions of children drown (or die from malaria) because you don't want to get your suit wet saving them. Saving a life costs $5,000, cancelling student debt costs $500 billion, you do the math. Student Debt Crisis If what you really care about is stemming the ill-effects of large and growing student debt, debt cancellation is a terrible policy. If you want people to consume less of something, the last thing you should do is subsidize people who consume that thing. But that's exactly what debt cancellation does: It is a massive subsidy on student debt. Going forward, the legal precedent and political one-upmanship will make future cancellations more likely, so students will be willing to take more debt, study less remunerative majors, and universities will raise their prices in response. Helping those who are already saddled with student debt by pushing future generations further into it is not the right way out of this problem. Fiscal Conservativism Student debt cancellation is expensive. Several hundred billion dollars has already been spent and several hundred billion more are proposed. This will mostly be financed through debt, especially si...