HNHacker News
TopNewBestAskShowJobs

currymj

3,822 karma · joined March 27, 2016

submissionscomments
currymj··on Legalizing sports gambling was a mistake
financial markets also work this way — the current market price is for a limited quantity and if you trade the price will move.

moreover people go to great lengths to try to avoid trading with winners.

there have been cases where people’s banks refuse to do any more foreign exchange trades with them when it becomes clear they are just arbitraging. it’s exactly analogous to the sports book case.

currymj··on Legalizing sports gambling was a mistake
this pattern is a pretty defensible way to run a sports book. obviously you don’t want to accept large bets from someone who is doing arbitrage or consistently has inside information.

any business with variable prices works this way, if some mysterious person shows up to your car dealership and seems really excited to unload a bunch of used cars on you, you should feel nervous that you’re overpaying or something is wrong with the cars.

in my view the diabolical part is the predatory marketing tactics, and making gambling platforms ubiquitous.

i say this as someone who, like you, thinks legalizing sports betting is an ongoing disaster, but wants the strongest arguments against it.

currymj··on Legalizing sports gambling was a mistake
like a lot of libertarian experiments in the US, we seem to have done it in the worst possible way. implementation matters.

if it had been up to me, sports gambling should have been restricted to physical locations, and marketing prohibited or highly restricted (perhaps only print advertisements in local markets informing people where they can gamble). perhaps also allow an existing customer to place bets via a telephone call.

the way I think about it is, the main reason you want to legalize a vice is to prevent criminals from selling it.

so you want legal operators to have an easier time doing business than the criminals, so they can outcompete them -- but just barely.

app on your phone and unlimited marketing on the internet and primetime television goes way too far.

currymj··on Timeshare owner? The Mexican drug cartels want you
what happens to your estate if you die with a timeshare? does it just have to stay in probate forever while they take out a management fee every month?
currymj··on Scaling up linear programming with PDLP
if you have a license for a good fast solver already for some reason, it can be very effective to reduce some NP-hard problems to integer programming.

also, one interesting application related to data science is to embed a machine learning model (usually regression, maybe decision tree or neural network) inside the integer program, and then maximize over its inputs. this can let you characterize worst-case behavior, or answer questions like "how must a user change their features to flip the prediction?".

solvers are so expensive that MIPs will probably never become part of mainstream data science, but they can be very powerful beyond classic OR problems.

currymj··on In the US opioid-maker Purdue is bankrupt. Its global counterparts make millions
it got started with massive overprescription of opioids in the 2000s, then the government belatedly cracked down on that, and a lot of people who got addicted switched to street drugs, which is when the crisis became obvious to the public.
currymj··on Scaling up linear programming with PDLP
JuMP is comparably good I think. People reasonably don’t want to add a new language to their stack. But if you’re just formulating MPs it’s as nice as anything, free, and you have a well designed modern programming language if you need it.
currymj··on Scaling up linear programming with PDLP
the Mosek Modeling Cookbook is really good for seeing the "tricks" of how to reformulate things as convex programs.

https://docs.mosek.com/modeling-cookbook/index.html

Not for total beginners though but great 201 level resource.

currymj··on Gaining access to anyones Arc browser without them even visiting a website
Arc is a great product, it's the nicest web browser to use, you can tell these people are really good at their jobs in many respects (though apparently not security?!?). probably a lot of investors saw that too and are willing to fund a very strong team with the hope of eventual product-market fit.
currymj··on Gaining access to anyones Arc browser without them even visiting a website
i'm rooting for them to succeed, but if the concern is security, switching your daily driver browser to a brand-new browser that's still in alpha is unfortunately not a good idea.
currymj··on Don't Take Advice from a Habsburg
but nobody has to pick between “the worst manifestation of counterculture” and “traditionalist monarchism”. you can have a whole range of normal political views instead.
currymj··on Visit Bletchley Park
it does seem like working on the Bombe got Alan Turing really deep into thinking about hardware. he'd previously played around with building some logic circuits but seems like he didn't go deep into it before Bletchley, before that he was mostly a very pure mathematician.
currymj··on Micrograd.jl
julia the language is really good. but a lot of core infrastructure julia libraries are maintained by some overworked grad student.

sometimes that grad student is a brilliantly productive programmer + the libraries reach escape velocity and build a community, and then you get areas where Julia is state of the art like in differential equation solving, or generally other areas of "classical" scientific computing.

in other cases the grad student is merely a very good programmer, and they just sort of float along being "almost but not quite there" for a long time, maybe abandoned depending on the maintainer's career path.

the latter case is pretty common in the machine learning ecosystem. a lot of people get excited about using a fast language for ML, see that Julia can do what they want in a really cool way, and then run into some breaking problem or missing feature ("will be fixed eventually") after investing some time in a project.

currymj··on Love them or hate them, this couple reign in Russian literature
it's good to hear that my impressions weren't based on nothing. personally i don't like when a translation works like this, I think it's better to try to replicate the literary effect in the target language. (For example I really enjoyed Emily Wilson's choice to do her Homer translations in blank verse, as the analogue in the English poetry tradition to dactylic hexameter, rather than trying to write English in Greek meter.)

but I guess it's a valid choice and might actually be better in certain contexts (plausibly this is true for a university class about Russia and Russian literature).

currymj··on Love them or hate them, this couple reign in Russian literature
the problem is for any given Russian novel, there are often other translations from the 1960s-1990s which use sufficiently modern language. maybe these are better than P&V or maybe not (in some cases yes to my taste), but good luck finding one in a bookstore.
currymj··on Love them or hate them, this couple reign in Russian literature
the P&V translations always give me the sensation of American actors speaking in mock Russian accents, like in the Hunt for Red October.

I don't really like them for this reason. I imagine they have merits which I am not equipped to evaluate.

I find it sort of frustrating that they have a near monopoly. It can be pretty tough to find a non-P&V translation in a bookstore these days.

I think this near monopoly, and therefore the financial/career/publishing industry implications, might be why some of the critiques and takedowns seem oddly vicious. It's not just about literary taste.

currymj··on DOJ sues realpage for algorithmic pricing scheme that harms renters
there is concern about tacit collusion if all the airlines start using the same pricing algorithms based on the same historical data. it doesn't even take a very smart pricing strategy for them to naturally learn to collude with each other.

However I would guess you will never have an email from an airline CEO that says "I really like this product...That's classic price fixing"

currymj··on DOJ sues realpage for algorithmic pricing scheme that harms renters
Algorithmic pricing that learns to tacitly collude is a hot area of study in computer science and economics. For example, if you train simple online learning algorithms to adjust prices, sometimes they can learn to keep prices high or to take turns winning customers, rather than just competing. People have found some empirical evidence of this on platforms like Amazon where a lot of small sellers use pricing bots.

However, it seems this is a more of a hybrid situation. A big part of the complaint is just all these incriminating emails and documents where RealPage appears to be coaching landlords to avoid lowering rents or giving concessions, independent of the software.

At the same time, there was an algorithmic component, which the customers appreciated: “I always liked this product because your algorithm uses proprietary data from other subscribers to suggest rents and term. That’s classic price fixing...”

currymj··on Seven basic rules for causal inference
it's a tough problem.

there are various schemes for estimating mutual information from samples. if you do that and mutual information is very close to zero, then I guess you can claim the two rvs are independent. But these estimators are pretty noisy and also often computationally frustrating (the ones I'm familiar with require doing a bunch of nearest-neighbor search between all the points).

I agree with the OP that it's better to say "non-independence" and avoid confusion, at the same time, I disagree that linear correlation is actually the standard definition. In many fields, especially those where nobody ever expects linear relationships, it is not and everybody uses "correlated" to mean "not independent".

currymj··on Seven basic rules for causal inference
Rigorous causal inference methods are just now starting to diffuse into the undergraduate curriculum, after gradually becoming part of the mainstream in a lot of social science fields. But this is just happening.

Judea Pearl is in some respects a little grandiose, but I think he is right to be express shock that it took almost a century to develop to this point, given how long the basic tools of probability and statistics have been fairly mature.

currymj··on Seven basic rules for causal inference
This is a separate issue and also a good point. Correlation sometimes means “Pearson’s correlation coefficient” and sometimes means “anything but completely independent” and it’s often unclear. In this context I mean the latter.
currymj··on Seven basic rules for causal inference
Rule 2 (“causation creates correlation”) would be strongly disputed by a lot of people. It relies on the assumption of “faithfulness” which is not discussed until the bottom of the article.

This is a very innocent sounding assumption but it’s actually quite strong. In particular it may be violated when there are control systems or strategic agents as part of the system you want to study — which is often the case for causal inference. In such scenarios (eg the famous thermostat example) you could have strong causal links which are invisible in the data.

currymj··on Can kids master the video games their parents loved?
2D games were a relatively mature medium by the 90s, so the games people were making by then were just good.

It took a long time before people developed a set of conventions after the general switch to 3D games. So a lot of the games from the late 90s - early 2000s period might have been really innovative and sold well, but often aren't very fun to play today unless you are a connoisseur of weird old games.

it's almost like the slow shift away from silent film -- studios invested in it hugely and there were blockbuster talking pictures, but it took a while before people really figured out how to handle it artistically.

currymj··on First USGS supercomputer decommissioning makes way for successors
the corporate descendants of Cray, now owned by HPE, still win the contracts to build a lot of supercomputers. but it's not like the days of Seymour Cray himself, hand-designing every aspect of the system. They use off-the-shelf parts mainly, it's essentially building out a highly-specialized datacenter. the main technology that Cray contributes is the specialized interconnect.
currymj··on Coinbase awarded a $500k bug bounty
generally yes to both. it's imperfect but the legal system already has longstanding ways of valuing illiquid assets that don't have straightforward market prices (art, houses, used cars, etc.).

in this case who knows what kind of tokens they stole, but if it was Ethereum or Bitcoin, those markets are liquid enough that you don't even have this problem.

currymj··on Julia for Economists (2022)
certainly yes in scientific computing, less so in ML/data science. there's much of the culture of scientific computing in economics -- lot of heavy numerical stuff in addition to the statistical modeling you might expect.
currymj··on Professional poker players know the optimal strategy but don't always use it
Two player poker is a zero sum game, where GTO play is very well-defined as just playing a Nash equilibrium strategy. The solvers try to get as close as they can to that.

Life is a lot more complicated in multiplayer poker. There are Nash equilibria, but potentially many with different payoffs, and you can't force your opponents to choose the one you're aiming for. So in that case, it's not so obvious what "optimal" means.

As for CFR adapting to opponent play: CFR could bias its compute resources towards really finely optimizing strategies for the most likely scenarios facing certain players, and it seems like this has been done during poker tournaments.

But within those situations, it would still be trying to more perfectly approximate the Nash strategy, vs. more experimental approaches which actually choose a different strategy to exploit opponent weaknesses.

currymj··on Professional poker players know the optimal strategy but don't always use it
GT Sophy is now a permanent feature in Gran Turismo and basically does this!
currymj··on FCC votes to limit prison telecom charges
even if one is concerned with remarkable similarity of prison labor to slavery, implications of 13th amendment clause, etc., most of the prison labor is for wholly-government-owned enterprises, frequently manufacturing things for other government departments.

As another example, Louisiana is phasing out private prisons. That should be great! Meanwhile Angola (state-run) continues to have prisoners picking cotton.

I know people know about this because they always bring up prisoners picking cotton in these conversations, but then the talking point remains "private prisons" somehow. So I still don't get why this idea is so sticky.

currymj··on Professional poker players know the optimal strategy but don't always use it
no, CFR is mainly just a way of computing Nash equilibria and (although in some sense it is an online, iterative algorithm) would typically be used to precompute Nash strategies, not update them in real time. real poker playing systems augment the CFR strategies with some real-time solving, but just to get even closer to Nash at the end of a hand.

on top of this, you could think about augmenting these systems to exploit weaknesses in opponent strategies. there is some work on this, but I don't think it's done much. The famous systems that played against professionals don't use it, they just try to get as close to GTO as possible and wait for opponents to screw up.

← PreviousPage 7 of 34Next →