majromax's recent activity
-
Comment on Looking for advice: How do I find out what stocks to invest into? in ~finance
-
Comment on Looking for advice: How do I find out what stocks to invest into? in ~finance
majromax Link ParentIt's negative sum, really -- if two people try to trade on the same signals, they'll end up moving the price such that their own profits suffer. However, that negative sum is good for the rest of...knowledgeable stock picking, "harvesting alpha" or how do the professionals call it - would be zero-sum against the professionals
It's negative sum, really -- if two people try to trade on the same signals, they'll end up moving the price such that their own profits suffer.
However, that negative sum is good for the rest of the market, since that price movement is ultimately what accrues to ordinary and passive investors.
As a thought experiment, imagine an auction for lottery tickets that pay off in a year. The ticket will pay either $10 or $100 in a year's time, with equal probability. If nobody knows anything, then the fair price is $55 discounted by about 6% for the time delay.
Now, imagine that someone random in the audience has X-Ray glasses and can see the winning tickets. They'll be able to bid $55 - 6% + 1¢ on the winners and completely ignore the losers, capturing the full profit for themselves.
But what about two people with X-Ray specs? In that case, they'll have to bid against each other, and the Nash equilibrium there is that the winning ($100) tickets are seemingly magically bid up to $100 - 6%. Observing that, nobody will bid on the losers ($10), and their price will fall to about $10 - 6%.
From the passive participant's standpoint, if two people have information then the market is priced as if everyone has that information. X-Ray specs for free!
The same sort of process applies for real markets, although it is fuzzier because of both market 'noise' and genuine uncertainty about the future. However, the so-called 'price discovery' effect is real, and that's why index trading is a surprisingly effective strategy.
-
Comment on Looking for advice: How do I find out what stocks to invest into? in ~finance
majromax Link ParentHedge funds hardly need me to defend them, but there are a few perfectly legitimate ways to make money at that scale that are inaccessible to a retail investor: Market making – a market maker is...Well, unless you're a trillion dollar hedge fund and you can afford to throw money around individual stocks based off insider trading,
Hedge funds hardly need me to defend them, but there are a few perfectly legitimate ways to make money at that scale that are inaccessible to a retail investor:
-
Market making – a market maker is always willing to buy or sell a stock. They maintain a quote where they'll buy the stock and a slightly higher quote where they'll sell it, and they try to profit by that bid/ask spread. They provide liquidity to the market – the certainty that when you randomly hit 'buy' or 'sell' in the retirement profile you'll quickly find a buyer near the market price – and in return they take the risk of holding on to an inventory of stock when the price moves against them. The ideal market maker holds zero stock at any moment, but of course to actually trade they need to hold nonzero stock.
People get legitimately upset about the idea of market makers using allegedly unfair practices like high frequency trading to 'snipe' orders as they come in, but this is a tiny edge by historical standards. Before 2000, stock prices on NYSE were quoted in fractions of a dollar, so the closest bid/ask spread was 12.5¢; now your typical mid-cap or large-cap stock will have a much smaller bid/ask spread.
-
Hedging of risk, the eponymous activity. Suppose you think one office supply store has a good new CEO and that it will do better relative to others. You can buy the stock as an ordinary retail investor, but in doing so you become exposed to risk for the whole sector – if all retail takes a hit, then that good CEO isn't going to save the company alone. Instead, your precise investment thesis can be expressed by buying the good company and shorting the (relatively) bad companies in equal measure; your residual exposure much more closely expresses your core idea.
There are also single-stock ideas to express. "This biotech company is developing a new drug. If it works out their stock will double, but if not it will stagnate or fall. So if the stock goes up 10% from today's level, I expect it to double." That's not a single trade at all.
Retail investors can sometimes express these trades, but it's awkward and expensive to do so (see above on bid/ask spreads), and keeping these trades balanced is sometimes like spinning plates. However, the machinery to keep these trades balanced is something you only need to build once, and the cost is amortized against scale (and internalizing some of the effects; maybe one trade has too much retail and the other too little, so they offset for free!).
-
Research scale. It is not worth your time or my time to read every SEC filing. Even though there are insights that can be made before the market realizes them – this is how price discovery works – at any personal investment scale the effective hourly return of doing this is tiny. However, if you manage a few billion dollars then those returns multiply by that factor, and suddenly it is worth doing a lot of things that were too small to pay attention to as a retail investor.
The safest and sensible move for the vast majority of retail investors is still passive investing and indexing. Even when they work well, hedge funds balance usually-small profits on individual trades against potentially large risks when things go wrong. The profession has built up whole schools of thought around risk management, and an ordinary person who does not make investing their full-time job will inevitably mess up and take on more risk than they intend to.
-
-
Comment on How funerals keep Africa poor in ~life
majromax Link ParentThis point is certainly overbroad. 'Kinship networks' in the West are often considered a secret of success; see for example this random, recent article that claims that kinship networks among the...David argues that the kinship network hobbles people from investing in savings or business development, which makes sense to me, but then concludes that the kinship network hinders economic growth at large.
This point is certainly overbroad. 'Kinship networks' in the West are often considered a secret of success; see for example this random, recent article that claims that kinship networks among the Jewish diaspora were responsible for the surprising economic success of a normally marginalized ethnic group.
Also taking every claim in this article at face value, the blog post doesn't really address why kinship networks are the 'oldest and most durable type of human society.' If that claim is true, they can't be purely parasitic under historical conditions; what changed?
I think the synthesis is found in how the determiners of wealth have changed over the past two centuries. As I see it, the wealth of pre-industrial agrarian and pre-agrarian is determined by luck, manpower, and common works, in no particular order. Kinship societies would help along all three axes:
- By redistributing from haves to have-nots, they act as a kind of mutual insurance. Before refrigeration consumable wealth was relatively ephemeral (it would rot), so a potlach is not 'ritualistic destruction of individual wealth' but more a sharing of fortune.
- By providing organizational ties, they can act to organize labour. If you have work to be done then you can call on this network to get your cousins to show up, and if they slack off then you can shame them in front of the rest of the network. At the same time, the dynamics of subsistence-and-a-little-bit-more production meant that most households were relatively unspecialized by modern standards, so it's more likely that generic family help would suffice.
- By providing hierarchy and stability, they satisfied some of the preconditions for common works to be shared among the community, like irrigation networks, field clearance, or herd enlargement.
I suppose that the problem arises with economic structures where wealth arises from individual capital: knowledge/education, tooling, machinery. Unlike the stochastic fortune of a good harvest those benefits are more sustained, so structures that incentivize capital accumulation are more likely to prosper. (Simultaneously, the other half of the equation -- that gifts bring status within the kinship structure -- matters much less when the source of wealth is an impersonal market rather than household production of subsistence-and-a-little-more goods.)
-
Comment on Claude Fable 5 and Claude Mythos 5 in ~tech
majromax Link ParentI expect that's exactly what we see with the updates to lower-tier models. For example, distillation of Mythos/Fable might well be used to make Opus or Sonnet 4.9. It would be the opposite. Models...Wouldn’t working on optimizations and deliver the same quality for lower compute be a more sensible choice?
I expect that's exactly what we see with the updates to lower-tier models. For example, distillation of Mythos/Fable might well be used to make Opus or Sonnet 4.9.
AI company like Anthropic would create more investor confidence if they delivered a model just as good as sonnet but runs in cheaper hardware
It would be the opposite. Models that run 'on cheaper hardware' are exactly the ones whose price on openrouter approaches the bare electricity cost. Anthropic and OpenAI need to support their valuations with seemingly unique frontier capabilities, sold at a price that at least recovers operational costs. The implied business model is that if you come for the frontier models you'll stay for everything else; Claude Code or Claude Whizbang Automation will be happy to spend tokens on other, less-frontiery tasks at more profitable price points.
Edit: Yeah, I don’t see anyone paying these prices
With enough reputation and reliability, people will. If you could truly double the productivity of an engineer or developer, even $10k/month in token cost would be worth it to many large companies.
-
Comment on Babylon 5 S01E07: "The War Prayer" - Episode Discussion in ~tv
majromax Link ParentUnfortunately, it's relevant because it's timeless. JMS drew on the propaganda of the 50s through early 80s as inspiration, and per his commentary Mira Furlan felt that the parallels to her young...I like this episode, I don't like how relevant it is.
Unfortunately, it's relevant because it's timeless. JMS drew on the propaganda of the 50s through early 80s as inspiration, and per his commentary Mira Furlan felt that the parallels to her young adulthood in Croatia/Serbia were disturbingly relevant. Twenty years ago these scenes found echoes in the propaganda around the occupation of Iraq, and nowadays...
-
Comment on Babylon 5 S01E06: "The Parliament of Dreams" - Episode Discussion in ~tv
majromax Link ParentLondo and Lennier, at the poker table.The goddess Li on Londo's table appears to have both male and female sex characteristics. I don't recall if that's ever addressed elsewhere but I love that attention to detail.
Londo and Lennier, at the poker table.
-
Comment on Voyager Technologies CEO says space data center cooling problem still needs to be solved in ~space
majromax Link ParentYes. The temperature scale for electronics is relatively narrow in the thermodynamic sense of things: the GPUs will want to operate somewhere between 0C and 100C. If you want to use the heat for...Does it have to be radiated outward, or is that what we currently do?*
Yes.
The temperature scale for electronics is relatively narrow in the thermodynamic sense of things: the GPUs will want to operate somewhere between 0C and 100C. If you want to use the heat for other directly useful things like "ore processing", then you'd need to develop spaceborne heat pumps (probably based on gas compression?) that can handle that kind of temperature difference at high efficiency. As far as I know this is essentially unobtanium, especially since at ore-melting temperatures you'd have the problem of the hot side of your heat pump wanting to melt itself.
If we did have that kind of heat pump technology, then we'd also just use it for radiators! The efficiency of thermal radiation scales with T^4, so doubling the absolute temperature of the material (from 0C ≈ 273K to 300C ≈ 573K) would result in a 16-fold change to the outgoing radiation.
Note that we also have a perpetual motion problem if we're not careful. If we go the other way, sinking heat from a 1000C source into a 100C target we could extract energy with (1 - 373K/1273K = 70%) efficiency because of the Carnot cycle, so our peak theoretical performance at moving heat is 1/(1-70%) = 140%. Applying 1W of energy to this system would move 1.4W of heat energy, so to sink X megawatts of GPU-related power we'd need an additional 2.5X megawatts to power the heat pumps (and move their own waste heat).
Mostover, if we did melt ore in space, what then? We'd still need to radiate that heat from the molten/refined ore in order to get the useful product (e.g. cast metal) out the other end, so we've only complicated the radiation problem.
Couldn't we channel the heat into a sink and then transfer that to something else - maybe the moon? Transfer the heat into the moon?
That's basically what we do on Earth with air cooling: we channel the heat into the air (or water, with open cycle cooling) and then transfer it into the rest of the planet. This necessarily involves a steady stream of material flowing in and out, such that doing this in space is not really viable.
Let's put numbers to this. Suppose that one of our spaceborne data centres uses 1GW of electrical energy (random Google results give me values up to 5GW, so I think I'm in the ballpark here). Let's ignore the heat pump problems from above and pretend we already have it at 1000C and can dump this into incoming iron with no losses (e.g. via some kind of perfect counterflow system).
The specific heat of iron is about 0.45J/g/K (taking 0.45J to heat 1g of iron by 1K/°C). A Joule is a watt-second, so that's 0.45 Ws/g/K, or 0.45 kW s / kg / K, or 0.45MW s / kg / 1000 K, which is the temperature difference we have to work with. We've already said that the space data centre has to dump 1GW of heat energy, so we can find the mass flow rate with some basic algebra:
1 GW = X (kg / s) * 0.45 MW s / kg → X = 2200 kg/second
That is, the space data centre would need to have 2 tons of iron flowing through it per second to use the incoming material as a heatsink. That's not a satellite, it's an industrial conveyor belt with extra steps. We in fact do this kind of thing on Earth, but here we call it a steel smelter.
-
Comment on Voyager Technologies CEO says space data center cooling problem still needs to be solved in ~space
majromax Link ParentYou've received other replies about this, but the concept you're looking for is entropy. Heat is energy, but it's the most useless kind of energy because it's also high entropy; to get useful work...Can’t the heat be used for thrust, power generation, manufacturing, or life support?
You've received other replies about this, but the concept you're looking for is entropy. Heat is energy, but it's the most useless kind of energy because it's also high entropy; to get useful work out of it we need to be able to transfer it from a hot region (high entropy) to a cold region (low entropy).
Unfortunately, that transfer is the very thing that's difficult in space. Since there's no direct contact with the environment to cool via conduction or convection, heat has to be radiated outwards – essentially by ensuring that the device's infrared glow points away from the sun. That's not very efficient, particularly when the solar panels have to be pointed at the sun to provide power.
-
Comment on Non-Logitech replacement for G502 mouse? in ~tech
majromax Link ParentI think a major part of the problem is also (ironically) power efficiency. Microswitches need a minimum current flow for long-term reliable operation; it breaks up oxidation that could otherwise...Then enshittification happened and in the name of a quick buck, Logitech switched to Chinese Omron switches that had an advertised tenth the lifetime and in reality, lasted even less than that
I think a major part of the problem is also (ironically) power efficiency. Microswitches need a minimum current flow for long-term reliable operation; it breaks up oxidation that could otherwise cause contacts to stick. Unfortunately, that kind of current flow obviously increases power draw, and designers normally want to eliminate it.
This also operates hand-in-glove with a switch to cheaper switches. If the cheaper switches are coated differently (or if their oxidation-resistant coating is thinner and wears through more easily), then the problem of low-current actuation is worse.
-
Comment on I let my wife have an affair. Do I have to console her now that it’s over? in ~life
majromax Link ParentI think an answer would be 'yes', but primarily in situations where the primary relationship ought to end. For example, I would not at all condemn an abused spouse for having an affair (emotional...Can an affair ever become ethical? To be clear, I'm not opposed to polyamory. I have doubts about whether a hidden, nonconsensual relationship can become ethical.
I think an answer would be 'yes', but primarily in situations where the primary relationship ought to end. For example, I would not at all condemn an abused spouse for having an affair (emotional or physical) with someone genuinely supportive.
Depending on definitions, there might also be trivial cases of brief 'affairs' before legitimization in a polyamourous context. For example, if a couple has opened their relationship but requires mutual approval, if A flirts with (extra-dyadic) C at work, their growing relationship might cross the line into an 'emotional affair' before A seeks dyadic-partner B's permission simply out of self-unawareness.
-
Comment on Debunking the AI food delivery hoax that fooled Reddit in ~tech
majromax Link ParentUnfortunately, the business incentives of media lean strongly into breaking the story, or at least being the first to publicize it. This effect predates social media entirely: CNN 'made its bones'...With journalism, speed is essential....or is it?
Unfortunately, the business incentives of media lean strongly into breaking the story, or at least being the first to publicize it.
This effect predates social media entirely: CNN 'made its bones' with its wall-to-wall coverage of the OJ Simpson car chase in 1994, which spun off into breathless reporting about the criminal trial. Essentially as soon as we had 24/7 news media, producers sought breaking content to keep eyes glued to the screen.
Longer-form, slower content is 'better', but I need to honestly ask the question of whether it's more valuable. IMO, the sad reality is that most news – even "important" news – has nearly zero value for the average citizen in terms of directly informing their actions. The emotional roller coaster is the valuable part of the story: "if it bleeds it leads."
Trade media is an exception to this rule, and I think it's instructive. There, reporting really does inform sober consideration of valuable choices, so accurate and thorough reportage gains value relative to the 'catchy' and emotional.
-
Comment on Debunking the AI food delivery hoax that fooled Reddit in ~tech
majromax Link ParentI like to practice an 'emotional scientific method', where I try to focus on how much a new piece of evidence contradicts rather than supports a narrative. This has a few different effects, such...So this is something that I should try more to be aware of. But then the question becomes when does outrage become legitimate?
I like to practice an 'emotional scientific method', where I try to focus on how much a new piece of evidence contradicts rather than supports a narrative.
This has a few different effects, such as forcing me to keep multiple perspectives in mind, but for the purpose of dealing with outrage it helps with the propensity towards the 'bitch eating crackers' syndrome.
That comes from our natural confirmatory bias, so when we've decided that X is bad then nearly everything X does gets interpreted in a negative light such that even ostensibly neutral things (like the eponymous 'eating crackers') reinforces the negative perception. The application to political discourse is self-evident.
However, that alone doesn't help with outright fake information such as this hoax. Here, I rely on a second emotional scaffold: slow news. Very few events are both important and urgent such that I need to change my actions right now[†], and the intersection of important, urgent, and emotionally resonant is even smaller.
Thus, I try to downweight breaking news, and instead I rely more on longer-form and slower analysis pieces where the authors have had time to look into the totality of then-breaking events and put the news in a broader context. This consciously 'swims against the tide' of the algorithm and so it's hard to keep up the focus, but I truly find that it helps.
[† — The weather forecast is one of the few consistent examples of information that is both important and urgent.]
-
Comment on Debunking the AI food delivery hoax that fooled Reddit in ~tech
majromax Link ParentMarket manipulation could be an easy movitation. Buy some cheap, out-of-the-money put options on Uber, post the 'whistleblower' report, and hope that the news media picks it up. If the story goes...We don’t know who the hoaxer was here, or their motives. Seems fairly amateur-hour, tbh, given the current state of AI tools. A properly motivated and funded disinformation campaign would know how to make its fake badge and research paper less detectable as such.
Market manipulation could be an easy movitation. Buy some cheap, out-of-the-money put options on Uber, post the 'whistleblower' report, and hope that the news media picks it up. If the story goes viral and especially if it catches the attention of regulators or politicians, the stock could drop 10-20% in a profitable manner.
In a few years we’re all gonna be so fatigued from questioning everything that we just believe nothing anymore.
As I see it, AI content generation as gotten good enough that we should no longer trust unauthenticated sources or evidence without a known chain of custody: unverified comments should have all the credibility of "my uncle works for Nintendo and told me how to capture Mew."
It's a crying shame, but I think that this is the inevitable consequence of having a giant mixing vat of human attention without other defense mechanisms; blaming AI for this is like blaming the salmonella for colonizing the raw chicken left out on the counter.
-
Comment on Indie Game Awards rescinds Clair Obscur's GOTY wins over use of generative AI [for now-removed background assets] in ~games
majromax Link ParentI think that this is revealing. Your linked FAQ states: This tells us that the Indie Game Awards is comfortable applying essentially political, values-based standards as part of its eligibility...[T]here was another award retracted this year due to the game in question – Chantey – being tied to ModRetro, which is in hot water due to ModRetro advertising that their consoles are made using the same metal that's used to manufacture attack drones.
I think that this is revealing. Your linked FAQ states:
Due to the ties with ModRetro, Indie Vanguard [for Chantey] has also been retracted as we do not want to provide the company with a platform.
This tells us that the Indie Game Awards is comfortable applying essentially political, values-based standards as part of its eligibility criteria. This in turn implies that the technical argument of whether Clair Obscur's AI use was significant might be beside the point: the IGA is willing to decide guilt by association.
That is, of course, the IGA's prerogative, but it means that all we can really do is divide ourselves into pro and anti-AI camps and shout the same arguments across the fence.
-
Comment on <deleted topic> in ~tech
majromax Link ParentThose aren't entirely independent factors. The highly-attractive guy with all those other good qualities is also more likely to be in a relationship and therefore not seeking hookups, so...The only real determining factors are how good he looks, how good he is in bed, and how likely he is to kill you.
You have no real way of knowing the second two, so literally the only criteria that matters is how hot the guy is.
Those aren't entirely independent factors. The highly-attractive guy with all those other good qualities is also more likely to be in a relationship and therefore not seeking hookups, so conditional on being on the app attractiveness is possibly negatively correlated with skill in bed and positively correlated with murder.
-
Comment on Financial collapse? in ~finance
majromax Link ParentOligarchs in authoritarian countries face their own personal existential risk if el jefe decides to defenestrate them, so borrow from their playbook and acquire estates (ideally) or the...I find it very hard to gauge what form my assets should be in if I want the best protection against a crisis,
Oligarchs in authoritarian countries face their own personal existential risk if el jefe decides to defenestrate them, so borrow from their playbook and acquire estates (ideally) or the right/ability to live in multiple foreign jurisdictions. Otherwise, diversify assets by jurisdiction.
Of course, this is very different than the best "invest for retirement" advice, and safety comes at the expense of returns. The real crisis might just be running out of money in one's old age.
-
Comment on Ukraine destroys more than forty military aircraft in a drone attack deep inside Russia in ~news
majromax Link ParentThe best explanation that I've seen recently is that Russia's objective is inherently an imperialist one: it's fighting to defend its claimed sphere of influence. Thinking of itself as a...This sort of capability is why cant figure out Russia's objectives Ukraine.
The best explanation that I've seen recently is that Russia's objective is inherently an imperialist one: it's fighting to defend its claimed sphere of influence. Thinking of itself as a superpower, it has a natural right to a sphere of influence where it has the final say over international arrangements, regardless of the sovereign wishes of the affected countries inside this alleged sphere.
In that view, Ukraine's post-Euromaidan attempt to align itself with the EU and by extension the US/NATO is an attack against Imperial Russia. In particular:
- This happened against Russia's wishes, and Russia ought to have the final say over this sort of thing, and
- The very attempt should have been unthinkable for a country in Russia's sphere of influence, so it must have come at the instigation of foreign diplomats and spies. Therefore, it was an international (American) attack on Russia.
Of course this view is complete bollocks, more like the warped mindset of an abuser rather than that of a rational state actor. However, it explains most of Russia's actions, including its opinion this would be a 3-day SMO, its extremely aggressive internal rhetoric, and its expansion of demands for peace as the war has bogged down.
Unfortunately, this spells bad news for any peace process. Russia's baseline demand is that Ukraine cease to have an independent foreign policy, and from its imperial perspective any peace deal that results in a viable, independent Ukrainian state is a loss relative to its assumed position circa 2013. Of course, an independent, sovereign, and viable Ukraine is also the most basic term of any Ukrainian-acceptable peace deal.
Sadly, I fear that this means the war will continue until one party or the other is exhausted. The availability and routine use of long-range strikes by both parties makes even a de-facto ceasefire (digging in along current lines of control with a steady reduction of combat intensity) impossible.
-
Comment on How my life changed with ADHD medication in ~life
majromax Link ParentNote that there are pill bottles with timer-caps, which automatically count the time since the bottle was last opened. It can't help you with "take the pill out of the bottle, forget the pill,"...I had those senior people MON to SUN boxes and I'm not sure if I took one already or forgot to refill it last week.
Note that there are pill bottles with timer-caps, which automatically count the time since the bottle was last opened. It can't help you with "take the pill out of the bottle, forget the pill," but it would conclusively tell you if you haven't even done that much.
-
Comment on Removed Reddit post: "ChatGPT drove my friends wife into psychosis, tore family apart... now I'm seeing hundreds of people participating in the same activity. " in ~tech
majromax Link ParentThe system is what it does. The output of a language model is just text, so the only thing that can be regulated is text. Returning to the original post, if I – a human person – give you...People are saying "we can't regulate text" but an LLM is not just text. If it was, I would agree.
The system is what it does. The output of a language model is just text, so the only thing that can be regulated is text.
Returning to the original post, if I – a human person – give you sycophantic responses that feed into your delusion that you've unlocked a hidden plane of thought or somesuch, there's no obvious liability that attaches to me unless I give you instructions that cause specific and physical harm.
The efficiency with which this stuff can be pushed into existence by just a few rogue actors is so staggering that no human court system can begin to handle the cases it can potentially produce.
This is an argument against stringent regulation, not for stringent regulation. "Rogue actors" that want to cause harm aren't going to operate openly under an AI banner. They'll slide into your DMs, at scale, with models that pretend to be human. Most marks won't even realize that they're conversing with AI agents, evading the whole LLM-targeting regulatory regime.
The two-people simplification works above because of the restricted auction setting, where it reduces the problem to a simple game theory one. I agree that the real world is much more complicated, and 'having enough money to change the price' is related to the trade's capacity as far as a financial firm is concerned (i.e. is there enough room to make a profit relative to the money we'd want to allocate, or does investing close our own window?)
It's worse than that: before the news a fair price reflects the weighted sum of all possible outcomes. The simple observation that bad things happened did not mean that the prior price was unfair, in the same way that paying 50¢ on a $1 coin flip is not 'unfair' even if the coin comes up the wrong way.