HNHacker News
TopNewBestAskShowJobs

roenxi

30,626 karma · joined August 8, 2016

submissionscomments
roenxi··on Computers Cannot Make Decisions
It happens; sometimes the manager's manager needs a scapegoat and the magnitude of the problem is too big to blame someone low level.
roenxi··on Has the Autonomous Trucking Revolution Arrived?
I'm thinking of a time I had to help someone with reverse parking for example. And most people aren't tested in extremes - if they think they can't conditions are harder than they want to handle they just won't go out driving.

Looking at the SAE standards document [0], I can think of a few people I know who only technically reach Level 3. Mostly older women, to be honest. They are only going to drive under certain well defined conditions and will put someone else in the hot seat the trip will be outside a limited comfort zone. They would need someone else to be able to step in at a moments notice to operate a vehicle truly safely. As an aside; the older men are more willing to be adventurous, but also more likely to kill themselves and/or run someone else over based on my experiences with accidents. And in an AI from an engineering perspective that probably bumps them more towards Level 2.

But a truck isn't going to be facing level 5 conditions in the vast majority of circumstances where it is generating economic value. It's probably going to be going in a straight line down a highway or parking in a loading bay.

EDIT And drug/alcohol affected road users probably deserve their own paragraph; but I'm not going to make time to type it up. Just because they might not ask doesn't mean they don't need someone else to drive for them.

[0] https://saemobilus.sae.org/standards/j3016_202609-taxonomy-d...

roenxi··on Has the Autonomous Trucking Revolution Arrived?
I'm not sure humans have reached Level 5 autonomy either; many of them can't handle the full range of conditions humans drive in and occasionally need other humans to step in and help them.

Level 3 or 4 is more than enough to get most of the commercial benefits.

roenxi··on A statement on the Tor Project's relationship with Mullvad
Sure. But the beliefs are easy enough to describe. Eg, the flat earthers believe the earth is flat. And it's obvious why that isn't the case because even with a good telescope I can see stars but not mountains in foreign countries. Case pretty much closed, earth can't be flat.

And that probably isn't exactly correct but it is a big chunk of the way there. It isn't "what they believe is just so far out, and crazy, I don't even know how to describe it". It's easy to describe.

roenxi··on 11 of 23 Core Open Source Projects Run on 1 or 2 People
Reflecting the enormous effectiveness of free markets in the software space. If there was a strong regulatory system around software development it might cost orders of magnitude more to run a modern OS.
roenxi··on A statement on the Tor Project's relationship with Mullvad
That generally suggests a lack of empathy on your part. If there are an organised group of people who you don't agree with it should generally be easy enough to state what ideas they believe, why it is wrong and some plausible explanation for why people believe what they do despite being wrong (that goes beyond "well they're just inexplicably stupid").
roenxi··on A statement on the Tor Project's relationship with Mullvad
> The party itself is closed and self selected.

Isn't that how all parties work? If you accept ideological opponents of a party in then the party will just collapse. There has to be a process to make sure that new members of the party agree with its core principles before they join. Which is to say, the party has to be closed and self-select members.

If a party just lets any person join then there isn't any point to the party.

roenxi··on Iranian campaign planted fake articles in real U.S. publications using ChatGPT
You make a sound argument for not trusting the media (more precisely, the New York Times or whoever the editors were for that article). If an institution has an incident where they support batshit crazy claims from the military in support of getting who-knows-how-many people killed it would make sense to stop listening to them on military topics.
roenxi··on Iranian campaign planted fake articles in real U.S. publications using ChatGPT
> It's not whataboutism, unless somehow Pearl Harbor is whataboutism to Nuking Hiroshima. Sometimes things have direct cause and effect, even if it's inconvenient to your argument.

Wouldn't that technically be a whataboutism? As far as I've seen it used, "whataboutism" usually means that there is a cause and effect that is inconvenient to someone's argument. Whatever a whataboutism is in the popular parlance, it isn't a logical or factual error.

Anyway, the key point here is it'd be really helpful to understand exactly what magic the Israelis used to get the US into this war. It has to be the most bizarre promise-not-kept of any US president in decades. Nobody seriously thought this sort of escapade could end well and although Trump is pretty vague he seemed clear enough in understanding that lying the US public into wars was a legacy killer. It looks like it'll re-cripple the Republicans in the same way the last Bush administration did.

roenxi··on Why were Victorian elites so effective?
> That feels like retrofitting a modern frame onto a very different situation?

What, the British weren't up the challenge of forecasting a major war in Europe might end badly for them? I mean, I'm sure their leadership didn't, but that is why they don't get to run an empire any more.

> but WW1 is definitely not the fault of the any kind of British elite, we can blame it on the Austrians and, of course, the Serbs and ultimately one guy with a gun

It looks a lot like the fault of the British elite. They failed to use their influence to get a positive result in Europe; which descended into a bonfire over 40 years until it ran out of fuel to burn and half the continent was under the thrall of communism. And if they were doing something productive in the mean time overseas it didn't deliver much result. There'd be an argument that their laissez faire attitude was a problem except they kept actively declaring war on Germany over and over until their empire fell apart.

That generation of British leadership demonstrated no capacity to get positive results apart from a remarkably swift decommissioning of the largest empire ever built. I suppose history can remember them in the positive light that as group they could have failed even worse except they at least had the good sense to bring in US management before the situation became unrecoverable.

roenxi··on Why were Victorian elites so effective?
The British Empire controlled something like ~20% of the world and other European powers were adding in 10% here and there. Probably more than that weighted by economic and military strength. It is kinda on the British elite that the world wars even started. Even with the most charitable read we're talking about a situation where the message was "lets not break the power of Europe" and they couldn't sell it to Europe.

The world wars were a crisis that caused change, but that is a bit like going bankrupt. Just because the crisis is fast doesn't stop the process being slow.

roenxi··on Sharing AI progress in mathematics
Developing solutions to mathematical problems generally leads to improvements in quality and quantity of life at roughly the speed they percolate from the ivory tower down to the shop floor. So "how do we like it" is probably going to be "we like it a lot, this is awesome".

Every company is about to have a staff Ops Researcher who has a better grasp of the underlying math and theory than any university professor. That is an unambiguous win.

roenxi··on An algorithmic failure beneath the secret ballot
Extend the question why are rapidity of results and ease of setup/teardown of the voting place design goals at all?

There are elections in India and the US to control billions of people and some very aggressive military people armed with weapons capable of ending modern civilisation. This is one of the most fundamental social technologies we have to keep the whole system stable (up there with limited liability corporations and courts). The downside of one of those elections failing because of voting machine trickery is stupendous.

Frankly, I'm not sure even auditability is that important as a design goal compared to voter security. If the election is close enough that an audit matters the two candidates are pretty similar. In practice, if there is any meaningful gap in preferences then the outcome of the election should be clear enough that an audit isn't that meaningful. It's nice to have and it is cheap enough to include in a paper ballot.

roenxi··on An algorithmic failure beneath the secret ballot
These volunteers will have a far easier time scaling their conspiracy if voting machines are involved. There are fewer points of failure to corrupt and it is easier to produce a plausible sketchy result where their guy wins by 100 votes.
roenxi··on Improper redaction reveals Google Data Center water and electricity usage
If we compare 2025 to 0025, a truly remarkable number of problems have been solved by, effectively, intelligent people doing research. While it is impossible that AI will fix everything it is plausible it will fix most things since it looks like it'll be able to do research faster, better and with a higher quality than humans as well as more effectively communicate the results to the masses.
roenxi··on Improper redaction reveals Google Data Center water and electricity usage
Is there a well known reason that the US has ruled out building more power plants to meet the increasing demand? This seems like a weird complaint; if it is a problem that this industrial activity uses power, why not ban all industrial activity? Traditionally the response to more demand has been to build more supply.
roenxi··on Venice’s failed war against Constantinople led to the first bond market
> I'm glad you've abandoned "FDR was deflationary." That was silly, but maybe if it convinced you that you need a history review it wasn't for nought.

I never said anything about FDR.

> I'm disappointed to see that you aren't engaging with the core observation that Germany's inflationary and deflationary episodes were separated by the better part of a decade.

There isn't much there to engage with. It is difficult for inflationary and deflationary episodes to happen concurrently. It seems quite reasonable to say that politics lags the economy by 10-20 years; that is what we've been watching play out in Europe and the US in the present era - there is a substantial lag between problems developing before it leads to political revolt. It takes a long time for malinvestment to become an obvious downstream problem.

Besides, the Nazis had an economy capable of gut punching the British and French empires then ripping through a chunk of the Soviets before anyone could stop them. If you want to associate that sort of result with deflation then I don't think it hurts the case it is associated with positive economic results. Just a shame that the productive strength was used to build a war machine instead of a better society. Although I will reiterate it probably wasn't the inflation or deflation that mattered.

> Except by funding those businesses, which wouldn't happen in a deflating economy.

It would. The real return is the same as it's always been. It isn't like a deflating economy has less real resources and opportunities. If the businesses aren't making enough returns to justify starting them then they shouldn't be started - it is silly to just start businesses for the sake of it; the unspoken assumption is that they are businesses capable of generating some sort of return.

---

I'm pretty sure people are just making stuff up. There seems to be no argument or evidence given the wild claims that upper class welfare is a good idea and the mathematical evidence is that this is just another tax. I only see "whatever makes sense has been tried before; didn't work. We don't need to provide evidence or an actual explanation beyond that pithy observation, it's just obvious read history". Which is outstandingly weird because usually when economics gets challenged there is a good theoretical argument and a lot of practical case studies up to and including the present.

This seems to be the only topic where the debate was settled a century ago and there isn't any need to justify it, despite that fact that exponential money printing is obviously crazy in theory and practice. And the practical evidence is that deflation is more likely than not associated with better economic outcomes than inflation.

roenxi··on French Bond Risk Hits Euro-Crisis Levels [video]
Depends what you're looking for. If you want fun factoids I like https://marginalrevolution.com/. For introductory topics and macroeconomic ideas, YouTube search topic + pdh economist and videos >=20 minutes is often good although I will say I don't like Varoufakis' opinions when he turns up. That'll all usually end up in the world of podcasts because they host long-form content and leads naturally into more current headline topics.

For trading I do my own research based on whatever statistics come to hand. I have contempt for short term trading news on the theory that I'm too small to be tactical and will just going to ride along with big obvious trends. No recommendations there.

roenxi··on Venice’s failed war against Constantinople led to the first bond market
> ...by confiscating the gold, devaluing the USD from $20/oz to $35/oz, and taking on a lot of debt.

These things are only useful if interacting with a strong industrial economy. If the economy is in a bad place then it won't help - there are a lot of countries that have confiscated, devalued, taken on lots of debt then sunk because their economy isn't any good. Pretty much any country that runs in to economic trouble tries some combination of them at some point.

What actually matters is formation of new businesses and capital. Monetary inflation doesn't help with that in any meaningful way; it's basically just another tax because it reallocates resources from the productive economy to whoever is getting first dibs on the free money.

> The deflationary shock of the late 20s, on the other hand, propelled the Nazis into power...

Yeah, that's unconvincing. It's well established that the people with power in the German economy couldn't bring about general prosperity (arguably because they'd lost a war, arguably because they just weren't very competent). They let a hyperinflation happen, that's quite compelling to show they failed to inspire any confidence. Decades of pain and mismanagement and then the clear lesson we can learn from all that is if they'd printed even more money they'd get a good result? The evidence is too thin to support the conclusion, and contrary to the fact that they tried exactly that tactic with terrible results.

> The root problem is too much debt. Inflation and deflation are just how to deal with it. They both suck, but an inflating economy has jobs while a deflating economy doesn't.

So don't do either? Just forgive the debts directly.

Inflation and deflation don't theoretically have any effect on debt, the people lending the money can charge a real interest rate and account for inflation. Unless the government intervenes which makes the argument for the indirect inflation solution a bit moot because the regulators directly controlling the debts anyway. And if it's supposed to be an overly complex debt forgiveness scheme it doesn't work. The US has the largest debts in history under an inflationary monetary scheme.

roenxi··on French Bond Risk Hits Euro-Crisis Levels [video]
A 6 minute video is not an appropriate way to assess interesting things happening in the economy. It might as well just be a "French Bond Risk Hits Euro-Crisis Levels" tweet.

Economies are complicated beasts and the bond market represents the sum of all the thinking around what opportunities exist and how likely they are to pay off. It is a big, complicated beast. There isn't a lot of meaningful information about it that exists between the 7 word headline and 5 minutes of talking heads.

YouTube has some pretty good economic content but it takes at least a half-hour and more reasonably an hour+ to get even introductory coverage of a topic.

roenxi··on Things that apparently cause cancer
There is a quote in the article suggesting that they don't understand why the method is failing [0] because if they completely understood the mathematics they wouldn't say it. But their argument is that the results are obviously stupid for 2 reasons. Which is sufficient to stop looking. If they are correct on those two points then the method is faulty.

1) If applied to any landmark, that landmark causes significant cancer deaths.

2) The result is impossible because these people aren't exposed to anything from the nuclear power plants. The average nuclear plant is a remarkably environmentally-friendly thing.

[0] "There is, in fact, a real chance that you simply cannot get a negative result from this method."

roenxi··on Venice’s failed war against Constantinople led to the first bond market
> because the events of 100 years ago showed us what deflationary shocks look like (even worse than the inflationary shocks)

This comes up a lot, but it seems to be a just-so story. People seem not to have a justification for why deflation is the major factor instead of any other economic lever and every time one of these terrible deflationary events happen the country involved tends to end up a really nice place to live. The big example of the US in the 1930s led to an economy that conquered most of the known world in short order and reigned pre-eminent for 60 years. It doesn't appear to have held them back.

The inflationary shocks have a bad track record and be associated [0] with impoverished backwaters and collapsed states. Even then it isn't as obvious that the inflation is the cause as much as that the government of the time didn't have any better ideas than printing money to try and solve their problems, which obviously isn't going to work.

> Challenge: propose something better.

This seems easy to do; the money could be handed out per-capital instead of disappearing into asset markets.

[0] https://en.wikipedia.org/wiki/Hyperinflation#Notable_hyperin...

roenxi··on With most information hidden, the game Stratego had stumped AI until now
> How about the fact that a child can beat the best rock paper scissors player in the world in a game, but no human can beat the best chess engine? Same thing with poker, a novice could get lucky and win a hand against the best poker player.

How about that indeed. What are you trying to say here? The probability of a novice beating an expert doesn't tell us anything except how decisive skill is in a game. We can come up with skilful games where novices sometimes beat experts - in fact, that includes basically all games.

The point of something like an Elo rating is that on very rare occasions a novice will beat a grandmaster at chess. A human child may well beat the top chess AI on a long enough timescale, we really don't know how that will go. No reason it can't happen. They just have to get really lucky enough times in a row.

This has nothing to do with perfect or imperfect information. We can come up with a hidden information game where the odds of a novice beating an expert are approximately zero. And we can come up with a perfect information game where experts lose easily (for example, Go has a handicap system to equalise any skill gap).

roenxi··on With most information hidden, the game Stratego had stumped AI until now
In most good full information games, the best move also depends on knowledge you don't have - a full tree of all possible game states from a position. AI playing Go or Chess can't make the best move because they don't know what it is, they're technically just guessing. There isn't any reason to think AI have more or less trouble with hidden information games.

The practical difference is how many rules a game has and how easy it is to implement the engine. Implementing a chess bot is relatively easy because the amount of state tracking required to set up a simulation is basically nothing (I think just whether the king has made a move yet or not). That makes it easier to implement than something with a lot of signals that need to be recorded. Something like DoTA or Starcraft takes serious engineering effort.

roenxi··on Alarm Grows over 'Clearly Illegal' US Election Meddling in Brazil
One of the interesting things about the Chinese rise has been their paranoid approach to keeping western media (and, I expect, NGOs) out of China. And the relative secrecy that their political processes operates under [0]. Particularly their stiff resistance to the likes of Google and social media. It is fun to question whether a more open approach would have resulted in earlier or more effective US intervention in their decision making than the difficulty that the trade war with China seems to have had.

[0] I suppose I can't read Chinese; maybe if I did it is all well explained. I doubt it.

roenxi··on To grieve, or not to grieve?
Well sure, but labour isn't going to get much done by pretending that it isn't happening or making moral arguments. I'm put in mind of all the various powers that collapsed in the face of European colonialism - feel, argue and say whatever at some point only the massive power gap matters.

These AI don't get tired, they are still improving at unbelievable rates that can't really be matched by a human and they're already good enough to do most intellectual tasks to some sort of standard (often quite high). Attempting to assert the primacy of meat-based thinking is not going to get very far at the rate things are going.

roenxi··on Singapore govt dating app uses Gale-Shapley stable marriage algorithm
What aspect of that do you want them to be concerned about? The bigger concern was population growth, it is nearly impossible for living standards to keep rising if the population is growing exponentially - at some point the number of new mouths overwhelmed the ability to produce food and some sort of underclass has to form. We need at least stabilisation at some point and it seems the easiest way to achieve that is to wobble between growth and decay.

We have all the tools for population decline to work out fine. The economy is run mainly by capital and it seems like we're decoupling military power from number of soldiers.

roenxi··on September 2026: The world today, as seen by one Polish guy
> Every generation believes it is living through the end of something.

Many generations are. Last century saw the collapse of the British and Soviet empires - titans by any measure - through mismanagement. Europe as an international power was also broken (compared to how they used to be). I can easily imagine 1900s me fleeing to America in fear of the end times and feeling quite justified in his choices with hindsight. Less obvious where to run to these days as the US not-an-empire looks like it may crumble.

Most of the people who died didn't get an opportunity to complain, so we live with some serious survivorship bias on how bad things got.

roenxi··on When oil prices spike, where does the money go?
> US has $40T federal debt (+state, +municipal, +household) and probably 100T+ debt worldwide which the future generations have to pay.

Future US citizens are clearly not going to repay the debt, simply because they can't feasibly pay that much back in real terms. They cannot achieve the impossible. The magnitude of the debt is far beyond what it makes sense for the US to make good on.

Future generations of the US won't be able to borrow on such favourable terms (ie, people will stop donating to the US cause) and all the capital investment that would have happened in the US to power their consumption happened in China instead - but the effects of the debt won't look like repaying anything.

roenxi··on Unsealed Briefs in Authors’ Case v. Microsoft/OpenAI
Sounds like a gross misapplication of the law enforcement system. The free exchange of information should be legalised.

As we can see in the good article, the copyright system might almost have cost us a lot of AI capabilities. The damage it has done in cases where the lawyers got ahead of the builders is incalculable.

← PreviousPage 4 of 34Next →