Some of these “ML” methods have applications outside of what you’d think of as ML. My background is in control theory, which relies on guarantees which you just can’t get from neural nets. Skimming through the outline, there are tons of methods here which are used in controls and estimation — certainly they’re still useful
I think your more charitable interpretation is also more-correct, but it is still an indictment of sorts: we’ve all read this poem, but very very few were critically paying attention to what we read. We mostly trusted our instincts, and didn’t care to double check.
To be fair, I don’t think we can expect more of elementary school kids. But I think I might have read it in high school, and should have done better.
>Foster was basically the rallying point for people opposed to the grammarian methods of teaching languages that started in Classics but ended up taking over how foreign language is taught in most schools and contexts
Humans naturally learn languages when they are immersed in the language. It sounds like Latin instruction was more focused on rules, and didn't provide that immersion before Foster. I can attest that many other foreign language classes also don't provide enough immersion to really learn the language, although being limited to ~10 hours a week makes that virtually impossible.
To be fair, I suspect the difference between “auto brake” and “assisted braking device” is mostly legal liability. In practical use I would understand both terms to mean the same thing. I think very few people believe a grigri will _always_ catch them. They just (accurately!) believe that in most circumstances it will. The 5% where it won’t catch you is of course deadly.
The military version is auto-brake for rappelling and you can go fully hands free per petzl but you have to use specific rope. My guess is since they cant control the rope in civilian version they can’t vouch for it working in 100% of cases
I was a defender of Google
Maps last week. Then I had to drive some roads I was unfamiliar with at the weekend to get to a familiar funeral.
Google Maps gave me instructions to turn off a junction right at the turn off, resulting in me missing it. Worst still, it did this not once but twice. And the detour was via road works.
Google Maps added a considerable amount of time onto what was already a 3 hour journey.
Thankfully I had planned for problems so there wasn’t any negative outcomes aside time lost and a considerable amount of additional stress when I was already in an emotionally bad place.
On the journey home I used Apple Maps and the difference in the directions were night and day.
Apple Maps gave me ample warning before each junction. It described the junctions (eg “right turn 100 yards after the traffic lights”) so I knew exactly where I was going. Even told me what lane I needed to be in.
Google Maps did do this on occasions to be fair. But it was highly inconsistent. Hence why I got stung so badly when its instructions were last minute. Whereas Apple Maps was 100% reliable 100% of the time.
It will be a long time before I rely on Google Maps for satnav again.
I've had a similar experience as well. I had sort of hand waved it as being better due to tighter CarPlay integration (i.e. that there is usually, IME, higher-detail UX - which ostensibly leaves less room for interpretation).
> It will be a long time before I rely on Apple Maps for satnav again.
I used to trust Google Maps, but in the last year it has given me some very bad directions that not only were wrong, but were sometimes dangerous. I've started to very much not trust its directions, which is unfortunate, because it used to be so great.
I've been using Apple Maps exclusively for a few years now, as I like the turn by turn directions much better, and the UI is more readable to me, as others have noted.
My wife seems to like Google Maps better, so we tried it on a recent trip. I ended up going around in circles in an unfamiliar area, and it kept telling me to turn where no road existed. I was flabbergasted it was so bad, as the last time I used it, I didn't remember it being so unreliable.
Engineering excellence has been a part of the Google founding culture, but I think that has been eroding in the past few years ... MBA types are taking over and engineering talent are slowly draining away. I'm just a nobody opining here -- they needed better product leadership than better business leadership. Now that their cash cow (ads on search) is eroding and they don't have a way to make solid products, they're kinda floundering.
My wife doesn't have that problem with Google Maps on her iPhone, so I'm aware of that counter-point. It might have something to do with that I use offline maps. Things started breaking -- points of interest doesn't load, or maps freezes, etc. I have not been able to unbork it. And it had been noticeably getting worse.
But even when it had been working, I had started noticing navigation errors started creeping in. It was nice that I hear things like, "at the next McDonald's, turn left" ... but I also get other strange navigational errors. Even just last year, my wife and I hear the directions, and it is having us go in a confused way. I don't trust it as much. (And there is a lot I can say about this using Promise Theory to analyze all of this).
Last night, at a town I do not reside in, I'm making an errand to got to the grocery store, and Google Maps breaks again. So I use Apple Maps. Apple Maps navigation tells me to turn in, and that I am in the parking lot for my destination. I have never heard Google Maps doing that. Another time just last night, it told me that I am crossing this light, and to turn at the next light. I also never heard Google Maps tell me that. Those were literally the kinds of things I would ask my human navigator to tell me in addition to the standard Google Maps navigation (on my wife's phone).
FWIW I tried apple maps today, motivated by this thread. I doubt my observations are one-of-a-kind but:
- Apple Maps has a different UI (obviously), which I think was the most annoying aspect of the switch
- Google Maps shows the speed limit, and the speed you're going at. Apple Maps just shows the limit. I prefer the google maps version here
- Apple Maps gave instructions like "go through this light, and turn left at the next one" or "turn right after the Starbucks". This is very close to how a human would give instructions, and I like that -- I'm already 'familiar' with it. I find Google's spoken instructions less natural.
I didn't notice anything else very different. Are there killer features that I missed? If not, I'll probably stay with google maps for just because of the UI familiarity.
I've been an iPhone user for about ten years and have largely had the same relationship with Apple Maps for that time. Indeed, for the first time in a long time, I'm seriously considering switching back to Android later this year over a handful of annoying interop issues— chief among them the stuff recently surfaced by Pebble around APIs for wearables.
Not sure if that means I should try Apple Maps again at the 11th hour or not— maybe if it's finally gotten good it would be better to not know and just remain in blissful ignorance.
As a naïve fool with no understanding of quantum physics, I want to take a stab at this! Here’s my hypothesis:
Consider a world in which everything is “very quantum”, and there are no easy approximations which can generally be relied on. In such a world, our human pattern-matching behavior would be really useless, and “human intelligence” in the form we’re familiar with will have no evolutionary advantage. So the only setting in which we evolve to be confused by this phenomena is one where simple approximations do work for the scales we occupy.
Sincerely, I don’t think this argument is super good. But it’s fun to propose, and maybe slightly valid.
The main objection is: if there wasn't a classical limit, our brains would have evolved differently.
So yes, we can use the antrophic argument as evidence for the existence of the classical limit, but it doesn't have explanatory power for why there is a classical limit.
This is called the anthropic principle. I personally have objections to it, specifically that due to emergence it is hard to make definitive statements about what complex phenomena may emerge in alternate universes. However, it's taken seriously by many philosophers of physics and certainly has merit.
My point is that it isn't possible to determine the emergent behaviour of a complex system from first principles. So arguments of the type "these physics don't result in atoms being produced, so life can't emerge" doesn't imply that other complex structures _like_ life don't emerge.
Technology is made iteratively by repeated trial and then observed error in the physical structures we've created (i.e. we build machines and then watch them fail to work properly in a particular way).
Technology that works in a different universe without atoms, would require us to be able to experiment within that universe if we wanted to produce technology that works there with our current innovation techniques.
I'm a fool too but two things I remember. One was a paper discussing the thermodynamics of groups of particles. When they have strong interactions with nearby particles classic behavior emerges very quickly as the number of particles increases. And not n equals 1 million, or 1000, but more like two dozen.
And then there was Feynman asked to explain in layman's terms how magnets work. And he said I can't. Because if I taught you enough to understand you wouldn't be a layman. But he said it's just stuff you're familiar with but at a larger than usual scale. And he hinted even then one level down and you run out of why's again.
Always remember that without the dotcom bubble, eastdakota would be counting in 6 minute increments :P
Sincerely, can you say more about the 8 years of pain? I’m curious how you navigated that, especially with/without relationships, family obligations, “runway” restrictions, etc
Edit: looking at the profile, eastdakota is CEO and cofounder of CloudFlare. There are probably interviews and Wikipedia pages that address my questions.
Those 8 years were painful. To make money, I worked as a bartender, an LSAT test prep instructor, as an adjunct law professor at a law school that was so bad it doesn’t exist anymore. I remember 4am at the bar in Chicago where I worked, cleaning up some patron’s puke off the floor, and thinking: I need to figure something else out.
All the time I was trying to find an idea for a startup. I still had the lawyer bit flipped on so lots of things I tried had a legal/regulatory bent. That was definitely a blind spot that held me back for a while.
The fun YC-related story on the founding of Cloudflare is that, before YC, Paul Graham used to host a conference called the “MIT Anti-Spam Conference.” He invited me the second year of the conference (2003, I think) to give a talk on how to write effective anti-spam laws. The very technical crowd was polite to the lawyer. I met a ton of interesting people, many of whom played outsized roles in machine learning over the next few years, including John Graham-Cumming, now Cloudflare’s CTO. Paul invited me back the following year saying I should do something similar.
I was pretty sure the audience wouldn’t tolerate the lawyer giving another talk about regulation, so I went to a young engineer on the team of the (bad) startup I was working on and suggested we build a system to track how spammers scrape your email addresses. He agreed to build the backend if I built the front end (which I largely stole from the hot startup of the time: LinkedIn). That turned into Project Honey Pot, which I gave a talk on at Paul’s conference. Project Honey Pot gave the initial seed of an idea that turned into Cloudflare. And the young engineer was Lee Holloway who cofounded Cloudflare with me and Michelle Zatlyn.
Lesson to me has always been even in times where you don’t feel like you’re making forward progress in your life and career, find ways to stay involved with interesting people and projects and chances are they’ll pay dividends in ways you don’t expect later in life.
I clearly remember walking back to Paul’s house in Cambridge after the 2004 conference where I’d presented Project Honey Pot. I believe he and Jessica had relatively recently started dating. They were talking about startups and how people didn’t understand how they worked. Paul suggested they should teach a class at MIT. And that, of course, is what later turned into YC.
There were other dramatic events that evening in Cambridge that I think sharpened all our minds and made us appreciate there’s no time like the present, but I’ll leave that story for another day.
I met a ton of interesting people, many of whom played outsized roles in machine learning over the next few years, including John Graham-Cumming, now Cloudflare’s CTO.
And the other way around. I met eastdakota which would later lead to me being at Cloudflare. Turns out networking (human and computer) is important.
And then, 20 years later, I meet a kid on the local playground who claims to be the son of one of the Cloudflare co-founders or CTO. I was thinking, wait, isn't JGC in Britain, not Boise? Even though the kid was visibly disappointed that I'm bearish on nVidia and LLMs, he strained my understanding of how puts work in stock trading. So that amusing conversation, sporadically interrupted by loading and unloading my toddler from the swingset, encouraged me to get more familiar with the broader investment market, beyond the simple kid invest products I've been working on the past four years. So thanks for that proverbial push on the swings, Cloudflare kid!
There's not a high chance that this piece of history will be lost "forever", it will just remain as a HN comment.
I wonder what would be the best way to harvest these and add them to, say, a wikipedia page. I'm pretty sure it's something that you could decently do with an LLM.
Damn, what a turn of opportunities from just saying yes and showing up (and obviously a ton of hardwork and sacrifices). Thanks for sharing!
I can't resist ...
> There were other dramatic events that evening in Cambridge that I think sharpened all our minds and made us appreciate there’s no time like the present, but I’ll leave that story for another day.
> ... appreciate there’s no time like the present ...
The present is now! Some of us are dying to hear the story.
Definitely. I think this is true of the internet in general, we have an amazing ability to communicate with almost anyone, even busy experts in niche fields, if we just make/ask something that interests them. I don't think I appreciate that enough.
I’m sure I’m missing context, and presumably there are other benefits, but 5-15% improvement is such a small step to justify rewriting codebases.
I also wonder how much of an improvement you’d get by just asking for a “simple rewrite” in the existing language. I suspect there are often performance improvements to be had with simple changes in the existing language
Far better justification for a rewrite like this is if it eases maintenance, or simplifies building/testing/distribution. Taking an experienced and committed team of C developers with a mature code base, and retraining them to rewrite their project in Rust for its own sake is pretty absurd. But if you have a team that’s more comfortable in Rust, then doing so could make a lot of sense - and, yes, make it easier to ensure the product is secure and memory-safe.
Disagree - a rewrite for “maintainability” is an engineer saying they want to rewrite in their preferred language. I wouldn’t allow someone on my team to rewrite a core dependency for “maintainability”, but I absolutely would if they suggested it would be faster and safer.
We’re talking about a rust rewrite of a fairly core level library. I don’t think C is inherently unsuitable or difficult to hire for. If the library was in Fortran then maybe.
But yes you are technically correct, congratulations.
I was responding to a general claim. In any case, I certainly disagree that C is suitable in 2025 for the vast majority of possible use-cases. For fun? Sure, but not for shipping code you want to rely on.
Obviously the code isn't going anywhere, and obviously we DO have reliable code we've built with C. But acting like C and Rust deliver equivalent value is simply farcical: you choose C for rapid development and cheap devs (or some other niche concern, like using an obscure embedded arch), and you choose rust to solve the problems that C introduced.
I agree that simple rewriting could have given some if not all perf benefits, but can it be the case that rust forces us to structure code in a way that is for some reason more performant in some cases?
5-15% is a big deal for a low-level foundational code, especially if you get it along with some other guarantees, which may be of greater importance.
There are hopefully very few things that can be done to low level building blocks. A 15% improvement is absolutely worth it for a library as widely used as a compression library.
Even 5% on a hot path are quite the big gains, actually.
Furthermore, they said that they did not expect any performance gains. They did the rewrite for other reasons and got the unexpected bonus of extra performance.
Ya, I’m very surprised by the argument “you’re some of the best at numerical analysis and high frequency trading, and you’d be bad at anything else”, lol. That said, I think there are better reasons to work at such a place. Providing liquidity to the market is a good thing, and has real world value, it’s just hard for us to connect it to concrete outcomes. But as a simplified “toy” example, when they orchestrate a trade for/with a retirement fund, they help the fund improve its holdings at very low cost. That benefits everyone impacted by the retirement fund
1. Incentivizing convergence of the price towards fundamental value, to support proper asset allocation decisions.
2. Supporting buying and selling (ie "liquidity") to shift consumption in time (and enable productive investments with the delayed consumption).
Suppose a retirement fund holds their investments over a 20 year period on average, growing at a modest 4%. A 1% wider spread would reduce the return from 119% to 118%. I'm not sure avoiding that is worth the financial sector constituting 30% of GDP.
What would happen if equity markets were only open a very short period a day? Say you have one auction a day, or maybe two, and no continuous trading?
Everyone whose advantage is speed would lose out (HFT, some prop traders). Their current gain would instead accrue to those on the other side of the trades.
This comment is weird to me. Are you saying that the "financial sector" (incredibly vague term) makes 30% of US GDP? Not even close. A trivial Google search proves that it is way off base.
> finance, insurance, real estate, rental, and leasing ... is 20.7% of US GDP.
Also, most of the GDP in the "financial sector" is in commercial banks and insurance companies. Yes, they take risk, but not the kind being discussed here.
Since the original article is about Jane Street's financial market making business, let's focus on investment banks. What percent of US GDP do you think that investment banks and trading hedge funds represent? It is tiny. I would be shocked if it is more than 5%.
> What would happen if equity markets were only open a very short period a day? Say you have one auction a day, or maybe two, and no continuous trading?
This seems like a question from Econ 101. Let's expand that to all free markets in the US. What if homes could only be bought or sold once per month, instead of daily? How about agricultural products? Quickly this argument falls apart. Wholesale and financial markets with continuous trading have existed for centuries. The purpose of continuous trading (or very frequent auctions, like the agro auctions in the Netherlands) is price discovery. If you do it less frequently, then you have weaker price discovery and worse (less accurate) prices.
Finally, professional financial market makers have an important role to play in reducing the size of bid-ask spread. I recently bought some 1Y US Treasury bills using Interactive Brokers. I was stunned by how tight are the spreads, and I am a "Retail Normie/Nobody". Absolutely, this was not available to people like me 30 years ago. Who do you think is providing this liquidity that keeps bid-ask spreads so tight?
The finance sector contributes only 7% of GDP, correct.
One of the sources I had in mind [0] cites 31%, but that was revenue as a proportion of GDP, which doesn't really make sense. However, the financial sector takes about 30% of US profits [1].
With respect to trading only once a day, I don't think your ad absurdum counterarguments hold water.
The HK stock exchange used to trade 4.5 hours a day, from 10 to 12, and then after a generous 2 hour lunch break from 2 to 4:30. How is trading 8 or 12 hours a day better for pension funds?
Price discovery with a limit order book that's cleared once a day might work just as fine as when it's spread out over hours, maybe even better.
Think about some major event affecting a company happening on the weekend. Then everyone can put in buy/sell orders with adjusted prices, and they'll be cleared during the morning auction. How is that worse than if the event happens intraday, and only the most switched on automatic HFT will pick off people still quoting at the old price, then the professional traders in hedge funds will pick off people?
As I said, the difference would be that the gains of the fast movers would instead be distributed among those on the "wrong side" of the news. Why should price discovery be worse?
Finally, what makes you think that liquidity and bid-ask spread concentrated in a minute a day would be worse than spread out over many hours a day?
I think the amount of money that is creamed off for this service is disproportionate, and the amount of benefit to wider society a case of diminishing returns.
Allocation of capital, market liquidity etc are useful, but the size of the financial sector and the rewards it gives out for this shuffling of money are insane.
I.e., they work to increase the return on capital. Since I believe one of the biggest problems in the world today is economic inequality stemming from an increasing gulf in the remuneration of labour and the remuneration of capital, I don't think this is a good thing. Let alone a good use of stupendous amounts of talent and money.
The only reason software engineers are able to be paid so much at all is because of “capitalism” and “free trade” I.e. vacuuming up money from the rest of the world, we sell them ads and “SaaS” and “intellectual property” while they sell us food and clothes. Would be a bad day for me personally and our profession in the US when that stops
We’ve seen what happens when the price of eggs goes up a little bit, never before seen levels of prosperity seems to be the only way to convince Americans to treat each other kindly, pick your poison I guess
