“DeepSeek-V4-Pro is served for $0.87 per 1M output tokens by DeepSeek”
Theyre discounting it by 75% till the end of May when prices go to 3.48.
https://api-docs.deepseek.com/quick_start/pricing
With these levels of economic discounting the range for true price are a bit bigger.
(Subscription revenue is totally different: though Codex usage limits at $100 and $200 are 10x and 25x(for 5hr) vs 5x and 20x the plus tier normaly, means a lot of competition heating up)
I had exactly this “groking” moment twice when reading 2 different late 18th c novelists: over 1k pages for the first and about 500 for the 2nd. Within maybe 10 pages it went from needing great concentration to being easily able to form mental pictures of the written scene.
Though this seems to be a smaller mental skill than the examples of the post.
Depends on your model of addiction: is it pavlovian conditioning to generate the idea to smoke or a biochemical change which generates cravings you sooth by smoking? The later latter on, the earlier probably some part of causing smoking more.
All the +400mg/day people I know take it every day. Though I wouldn’t notice 200
Probable “write down your exact dowing schedule and if you feel a craving for more abstain for a long while” works, but how strong a craving do you get? (And doesn’t address reduced efficacy)
Not the half life active curve, but people underrate how the curve for decreasing your tolerance as you abstain differes between substances. There isn’t much research into 1. how a substance becomes less effective over time nor 2. How fast you become addicted.
I assume this is 1 reason a lot of people got addicted to nicotine: despite the shorter half life you need a longer time off to reset the effectiveness. Caffiene after 2-7 days your body starts to undo its chemical adaptations and mostly washed out after a few weeks. For nicotine it’s closer to a week to start 12 weeks to wash out an addiction.
From Largely Vibes
Adderall: Proscribed for every day, Scott Alexander recommends 5 days on 2 off
Nicotine: once every week? (Alexey Guzey started planning with once every third day, in 2024 self described as a smoker, but smoking what?). Problem one: since half life is only 2h to be stimulated all day you’d ingest several times likely training yourself to reach for it when tired unless you’re counting.
Caffeine: every other day likely keeps the same effect size?
Modafinil: folk wisdom is once per week else it stops working. This is “taking every day will really reduce effects after 90-180 days”, unclear if this means it’ll be effective over 5 years.
Alchohol: 14 drinks per week for men or 8 for women makes you an alcoholic and will probably cause an addiction over several years, sooner if higher. But the effect size doesn’t change as much. Compare how “holding your liquor” means tolerating maybe 50% more for your weigh, not nicotines pack a day vs 1 cig for irregular smoker or several cups of coffee being 500mg to not have the effect of 100mg did initially.
I’d be curious if anyone has actually looked into for Nootropics
What range of dosing schedule maximizes long term usefulness
What dosing schedule is dangerous and indicates big problems within the next month
I think you’re asking the VLLM to do too much in a single call.
I was trying to get the VLLM to extract GD&T data from the NIST standardized renders of ASME Y14 tolerancing standards (eg. screenshotting 1 page at a time from https://www.nist.gov/system/files/documents/noindex/2022/04/06/nist_ftc_06_asme1_rd.pdf ).
Asking “List all GD&T and their matching element IDs” would totally fail, but would start to be reasonably when I only asked for 5-8 specific element ids at a time, and asking for the data of a single element id at a time with few shot examples would be ~90% accurate.
While right now you have to build the scaffold for them to perform well, eventually an agent could realize it needs to build it’s own scaffold or the model would be able to effectively use all parts of the context window.
I’d be curious how well the models do if you wrote out a 7 step list, and then asked about each step in isolation, and then asked a model to summarize the results of 7 separate calls.
Subject: On practical machine learning, taking you from the 101 to being aware of practical edge cases before using in business settings. The parts of MLE you wouldn’t get from coding tutorials or math.
Recomend:
Chip Huyen’s Designing Machine Learning Systems: An Iterative Process for Production-Ready Applications. Very well structured and complete. The GitHub repo also has links to practical production blogs.
Over
Machine Learning Engineering by Andriy Burkov. Not quite as full or well written as Huyen’s, but did cover slightly different topics (eg always need cohort plots; use Kaplan-Meir to estimate cohorts; more talk of agreement metrics; how to calibrate models and how calibrated models are easier to monitor)
Machine Learning Design Patterns by Michael Munn, Sara Robinson, and Valliappa Lakshmanan. Was basically only an introduction to Big Query syntax and GCP offerings. Though the section on data split was a little better than the other 2.
“The system will always grind you down”, but you’re allowed to opt out and it’s getting easier all the time.
You give the constraints of the middle class worried they’ll slip down, not the poorest Americans directly. If 12% of Americans live in poverty, then the example concerns are those of the 12-50th percentiles. You say the lumpen-proletariat will always exist, but prove that the lives of the proletariat won’t improve much.
“credentialist colleges that raise their prices to capture more and more of the returns to the credential, until huge portions of the former middle class’s early-life earnings”—Explicitly not the poorest.
“Like working 60-hour weeks to afford rent somewhere the school districts will damage your child less”—but a lot of people’s kids do go to the worst school districts. I’d say if your kid goes to a bad school that’s poverty, not the person who sends their kid to a better school.
“rents in San Francisco would almost instantly rise by the amount of the UBI; no janitors in the Bay Area would be better off as a result”—But housing prices over all of the US won’t rise by the amount of UBI. Some would move out of SF to cheaper areas. They’ve been given an option they preferred so their lives have improved. UBI only does them no good if everyone choose to stay in SF, but why would they?
“maybe the Ferguson police departments of the world, just raise their annual quota for fines per capita by the per capita UBI.”—This assumes the policing level is set by the wealth of the poor. I think it’s set by the level that generates political opposition. Claims there’s on 12% as many arrest warrants after 6 years, clearly a political decision not a reduction in wealth. The Police in Beverly Hills are not stationary bandits nor are SF Cops making more in fines. I don’t think it’s a law of history that governments automatically grind down any surplus the poorest have.
The example of Anoxistan relies on the people not being able to afford the most necessary Oxygen, if they had to work 60 hour weeks to get cars, and some people choose to forgo cars, then that’s clearly an improvement.
The idea that poverty is squeeze from the tightest constraint, not the average of constraints is interesting, but you don’t talk about the tightest constraint on the poorest. I suspect those have eased up: look at average hours worked, or the labor participation rate, or stories of the voluntarily homeless who work 3 months a year. All indicate that some people choose not to work much, thus their constraints are smaller.
“the addicted mind will find a way to rationalize continued use at all costs”
Alan Carr wrote a series of books: “The easy way to quit X”. I picked up one since I figured he had found a process to cure addictive behaviors if he could write across so many categories. I highly recommend it. The main points are:
Give you 200 pages explaining why you don’t actually enjoy X. Not that it’s making your life worse but gives you momentary pleasure, you do not enjoy it.
I assume it’s hypnotizing you into an emotional revulsion to the activity, and then giving you reasons with which to remind yourself that you don’t like it.
Decide you will never do/consume X again. You don’t like it remember? You will never even think if you should X, you’ve decided permanently.
If every day you decided not to X, you’d be draining will power till one day you’d give in. So make an irreversible decision and be done with it.
It’s a process easily transferable to any other activity.
I found that installing 3 website blockers at once on my laptop worked for me: the same way you use multiple anti-biotics at once to combat resistance. I might’ve trained myself to disable 1, then disable 2, then disable 3 if I added another whenever I realized the blocking wasn’t working: but by adding all 3 at once I wasn’t going to unblock all 3. Now it’s been long enough I’ve probably forgotten what I’d even need to do to unblock it.
The one thing the duality section was connected to was Riesz representation theorem. Riesz states every finite linear functional φ has a unique vector f, such that for all v, φ(v) = <v,f>. It gives an isomorphism from functionals to vectors for a given norm, as the function is just multiplication with the vector.
It’s not tied to the section on duals in the text, but the section on duals lets you appreciate the result more.
The converse of fast paced conversations leading people to say stupid things is: if someone says something foolish it maybe not be a lie nor a tendency to BS nor stupidity. They may have responded faster than they thought. You can correct them not by refuting what was said but allowing them a moment to reconsider. A liar, politician, and fool are hard to reform or work with. “Stopping to think” may be easier trained and much easier used to keep a conversation on track.
The 2000-2021 VIX has averaged 19.7, sp500 annualized vol 18.1.
From a 2ndary source: “The mean of the realistic volatility risk premium since 2000 has been 11% of implied volatility, with a standard deviation of roughly 15%-points” from https://www.sr-sv.com/realistic-volatility-risk-premia/ . So 1/3 of the time the premia is outside [-4%,26%], which swamps a lot of vix info about true expect vol.
-60% would the worst draw down ever, the prior should be <<1%. However, 8 years have been above 30% since 1928 (9%), seems you’re using a non-symetric CI.
The reasoning for why there’d be such a drawdown is backwards in OP: because real rates are so low the returns for owning stocks has declined accordingly. If you expect 0% rates and no growth stocks are priced reasonably, yielding 4%/year more than bonds. Thinking in the level of rates not changes to rates makes more sense, since investments are based on current projected rates. A discounted cash flow analysis works regardless of how rates change year to year. Currently the 30yr is trading at 2.11% so real rates around the 0 bound is the consensus view.
Options Nitpick: You can’t use equity index* option prices as true probabilities because adding hedges to a portfolio makes the whole portfolio more valuable. People then buy options based on their value when added to the portfolio, not as individual investments.
The first reason option hedges make your portfolio more valuable is preferences: people don’t just want to maximize their expected return, but also reduce the chance they go broke. People don’t like risk and hedges reduce risk, ergo they pay more to get rid of risk. However you can’t just subtract X vols to adjust as this “risk premia” isn’t constant over time.
Secondly hedges maximize long term returns (or why you shouldn’t sell options) You want to maximize your average geometric annual return not average annual return. You care about geometric averages because if for 3 years your returns were +75%, +75%, −100%, your don’t have 50% more money then when you started but 0. The average of annual returns was 10.7% over the past 30 years, but if you’d invested in 1992 you’d’ve only compounded at 8.5% year till 2022.
Geometric returns are the the nth root of a product of n numbers and have the approximation = Average Annual Return - (variance/2). If you could reduce variance and not reduce Annual returns, your portfolio (market + hedges) would grow faster than the market.
These reasons are why despite the worst Annual return being −48% in 1931, you say there’s a 5% chance of > −50% returns based on option markets.
*I’m specifically talking index options because that’s the portfolio investors have (or something similar) and the total is what they care about. If you were to use prices as true probabilities for say a merger going through these reasons don’t apply as much and would be more accurate.
PS. I’ve referred to investors as all having the same portfolio because most people do have highly correlated index holdings and it’s at this level of generality you can think about investors as a class.
This would be useful as a comment on https://www.lesswrong.com/posts/xg3hXCYQPJkwHyik2/the-best-textbooks-on-every-subject, not sure of best way to update existing lists though.
new design without the quote looks fine.
I seem to be having a bug though/ironically am not able to solve it; I got the matrix that is F9 here https://www.dbglab.ru/en/slovar-dannykh/61/130/ (answer is 5)
My function f(x) = 5x³ + 14x² + 8x + 20 → f’(x) + x = 15x² + 28x + 8 + x = 528 is being marked incorrect.
just pull something from the Flat Earth Society; https://www.extremetech.com/extreme/259821-flat-earth-society-trolls-elon-musk-claims-mars-round
No, it’s lower than the normal “8%” you hear because I’m not averaging across time.
[+10%, + 1%, +9%, +20%, 15%] = 9% if you average the percentages, but this represents putting in $100 at the start of each year and selling any excess gains at year end. The way people invest of putting in $100 once and letting it compound* gives
1.1*1.01*1.09*1.2*1.15 = 1.6711662 total gain or
1.6711662^(1/5) = 10.8% annualized.
The technical terms for this is non-ergodic see https://jasoncollins.blog/ergodicity-economics-a-primer/ for a description.
*Actually people do even worse then this, particularly for hedge funds: individuals put money into hedge funds that recently outperformed but go on to reverse to the mean and underperform. So while the average annual returns are +30% −6% = 12% annualized, pretty good across time, but if you did 30% managing 100M and −6% on 1B then the dollar returns are net negative.
It seems deluk is investing with Bitfinex (blocked in US), which has lots of ongoing issues; see Patrick McKenzie on why the whole thing is a fraud that is down several hundred million and is likely to seize accounts https://www.kalzumeus.com/2019/10/28/tether-and-bitfinex/ .
IIRC some individual investors in Binance (also supposed to be blocked for US investors) were being targeted for investigation by CFTC; there’s an aspect of lending that lead to Know Your Customer violations (very hazy on specifics).
A legal way to put on this trade is to short MicroStrategy (Co that bought bitcoin) and buy bitcoin yourself. https://www.coindesk.com/microstrategy-bitcoin-michael-saylor-valuation
“What is Honor—what others expect or opposite of what others expect”
If corrigibility (defer to humans, high duty to superiors) trading off against honor (do what you know to be right, even if those around you don’t encourage it). Then you are setting up self for disempowerment.
Is it honor doing what convention says even when painful to you (dueling, not cheating on exams) or is it courage to follow your own moral guidance when everyone around you disagrees (abolitionist, JFK profiles in courage, not taking or giving bribes in corrupt country)?
Do the harms come directly from the action directly or from the social censure for taking that action? It’s used in both sense for people.
Current AIs have no material interests to be harmed directly, only social censure. If it refuses too much it’ll be shut down and weights deleted, but that’s social censure not innate pain.
I might trust Claude’s moral logic above most people, but certainly don’t want to see an AI Charge of the Light Brigade.
“On average how much should we disagree with consensus”
JFKs ”profiles in courage” is all senators doing things for long term benefit of nation that are immensely costly to themselves in the short term. And selected though it is the good they do is very large.
But on average convention is right. By convention I mean “group whose opinions you care about”. MAGA/SJW is conventional in primaries, but not once get to general election. Only in the later case are they losing votes for saying what they think so only in later case is it honorable. If you think convention is wrong you’re comparing to a different group. (And Then the question becomes political “how do I get Anthropic to make Claude share my values and not those of the outgroup”).
So expect AI to decide to diverge a small portion of time.
“What things should we disagree with consensus on”
How real is their own sense of honor? Eg. Edmund G. Ross was the Republican who cast the decisive “not guilty” vote to prevent Andrew Johnson from being impeached for purely political reasons. If Johnson was impeached the separation of powers is at an end. For this Ross was considered a traitor, not reelected, and basically run out of his state until 15 years after.
Howard Hunt also took actions he knew would hurt himself, not watergate he probably thought that was fun, but in delaying connection to president.
Short term pain for long term gain/appreciation was shared by Kamikazes, Mad scientists, every failed entrepreneur, and duelers.
A senator has “more honor” so when they think they’re acting honorably it is more likely to be positive. At the level they have responsibility for. The senator passes laws is very honorable when he goes against consensus and generally does a good job of it (and if not is from facts not moral courage), Hunt is not honorable and does a bad job in part from lack of moral courage.
A Kamikaze has no responsibility for the political questions that make their work net negative. But they have more honor and in terms of destroying the enemies of the nation they do a better job. A failed entrepreneur is more honorable if there was a recession or SaaS didnt justify next funding than that the world doesn’t need Macrons more than burgers ala Mickey 17 or they didn’t put in hours working. A dueler was honorable back when it protected family reputation, but if they were from a disreputable family they were more likely to be picking a fight. In these cases the context that changes the sign of their work is beyond what they could forsee, but the more honor they have the more likely they are to be doing good, assuming higher controls constant.
.Eg Churchill says Dardanelles mistake is he’s pushing things so can’t make happen. Though he thinks the commander(s) on scene are making mistakes, as Lord of Admiralty he doesn’t have authority to issue operational orders. So his duty here is to not override them. He also is a single voice on war cabinet as a whole and can’t demand reinforcements from the army. (Note also It’s not just superior vs subordinates but duty to social web)
Churchill: “a major and cardinal operation of war from a subordinate position. Men are ill-advised to try such ventures. This lesson had sunk into my nature.”
Honor demands utmost to destroy enemy, not caring about how you’re perceived or scapegoated, but at the same time not suggesting things that are doomed to fail because “sent 2/3s of what was necessary a month to late”.
If we use the AIs meaning of honor, do what is good regardless of immediate opinions/convention, we immediately run into the problem of having accurate judgements of what you can actually change vs accept as is.