davidmr

> Mr. Zuckerberg, the founder of Facebook and Twitter, retaliated by blocking Mr. Trump from all sites.

I stopped reading there and I realized that the 20 seconds I spent reading up to that point were wasted.

And a photographer! MRO took what might be my very favorite picture of all time: https://www.space.com/16946-mars-rover-landing-seen-from-spa...

For the Top500 lists in particular, the list always comes out during the international supercomputing and supercomputing conferences in June and November respectively, so I’m sure there was no coordination. No idea about the TSMC thing, but I doubt it.

That’s an incredibly useful table! Do you know where the data in that table came from?

I mean, they regularly build robots, strap them on huge rockets, shoot them 300,000,000 miles away to mars, land them without a scratch, and drive them around for years. That’s awesome.

> I was just about to write something similar - the reason nasa projects fail is because they refuse to allow for iteration.

If that's why they fail, do they regularly succeed in spite of it or also because of it?

Just for my own curiosity, since I don’t recognize the city from the photo, where is it? I’m wondering how it compares in size (both area and population) to Manhattan.

You can do both. Job postings from at least a couple well-known HFT firms shows that they’re getting into sports betting.

It's genuinely shocking how far behind the USA is in terms of consumer banking. I moved from Chicago to London about six months ago, and I can't believe how much more convenient it is. You can wire money to anyone with a bank account without an intermediary like Venmo or Zelle from your phone, either as a one-off or recurring. I can revoke permission to draw regularly on my account from my phone, so I don't have to go 10 rounds with the gym or something to quit. You can pay for anything you like with a tap of your phone, and it takes maybe a second to clear. My credit cards never leave my possession.

I still have the envelope of cash I pulled out when I got here in February, minus maybe £40. I haven't needed it!

The IBM licenses are so obtuse that it can only be intentional. This may be outdated, but they used to use these stupid “processor units” that had some loose correlation to CPU performance. If a perfectly well-intentioned systems person moved licensed software to some new server that had half the cores, but was on a newer generation CPU, and didn’t think to ask the lawyers about it, you could get tagged for huge bucks. I will never buy IBM software ever again if I can go to any reasonable length to avoid it. RHATs licensing seems not to be moving to the IBM model yet, but I’m frightened.

That man has an absolutely unmatched gift for it. It’s genuinely impressive.

https://onehundredpages.wordpress.com/2013/06/12/dont-make-f...

What evidence to you have to suggest this extremely uncharitable interpretation of the Fedora project’s aims?

“On top of this the Intel modems consumed vastly more energy to do their slower work than Qualcomm, a trend that SemiAccurate has personally measured in the labs across multiple generations of Intel modems but is not at liberty to disclose exact figures on, sorry.”

Why not? Presumably they would be opening themselves up to some liability?

That seems like an uncharitable conclusion. I think most people worth hiring (in an interview where that question is asked) should be able to expound at some length on at least some part of the process. What part of the process they spend the most time talking about is an interesting window into their experience.

I also have a sneaking suspicion that the intersection of the set of people who are at risk of dying of malaria and the set of people who make that second argument is very small. It’s a pretty easy argument to make from your armchair in London or New York, and a harder one to make in Bangui or Kinshasa.

This genuinely surprised me, but I’m less sure than everyone else that this spells the end of Cray. They’re still cranking out monster systems: they have 3 in the top 10 now (and almost certainly more than that since the DoD stopped listing their systems). As everyone has noted, they also signed contracts for new monsters at Argonne and Oak Ridge, which will likely debut at #2 and #1 respectively.

I think it’s very unlikely that the US government lets them go out of business after HPE closes the sale. They’re the best competitor to IBM, and the DoE and DoD are always careful to spread their procurements around to keep more than one company capable of supplying the big defense/weapons supers. This is just totally a guess from having worked in HPC for so long, but I’d be very surprised if this purchase didn’t include some sort of back channel wink and nod by the feds at a promise by HPE to keep building the big computers.

I’ll miss them though. While their systems weren’t always the best, when you got your problems escalated to their R&D group, you got to work with some cool people. I imagine those people will get sucked into HPE and/or get fed up and defect to Intel pretty quickly.

There’s definitely audio from all those meetings. You can hear the mission manager plow right through the issues with bureaucratic ease. They’ve seen something similar before and even though it was unexplained, because they’ve seen it, they think it’s a known issue.

You want to blame her for it and shake her out of it as you hear her say it, but it wasn’t malicious. I believe the mission manager even had a spouse who was an astronaut, so it’s obviously not like they don’t care. I’ve always found it fascinating how organizational structure and pressure can take really brilliant and motivated people and beat them into making such poor decisions.

It’s really important to remember that these people are not idiots. It’s literally a room of rocket scientists and space shuttle mission managers. It’s so easy for us all to say here that it was so easy to see it happening and that anyone who didn’t ask the right question was a moron, but these structures take on a life of their own. If I had it to start all over again, studying that would make a fascinating and potentially rewarding career.

As much as I appreciate the sentiment and I never like to miss a chance to pile on PowerPoint, this is really, really missing the point.

As the CAIB report makes clear, the PowerPoint slide was a small symptom of the actual problem of a complex organization gradually accepting more and more risk as “in family” simply because unexplained phenomena hadn’t caused serious issues before (while remaining unexplained). The CAIB report really is a masterpiece (as is Feynman’s appendix to the Challenger report) of understanding how the understanding of risk can be subjugated to organizational pressures over time.

I’m absolutely not defending academia as an industry. I did my time there, and I would never go back.

To say that the overhead goes to flashy buildings and fat admin paychecks is a bit of a misrepresentation. Those buildings almost always have a major donor and a substantial endowment component. The academic (especially physical and biological science) divisions do actually need to pay for office space, lab space, heat, staff to run it, clean it, make the computers go, etc. If you think my salary was ballooning while building and running everyone’s HPC clusters, I can assure you it wasn’t in any sense other than perhaps slowly deflating over time. Compared to whatever the football coach and president are being paid, there are thousands of people on staff who aren’t getting those fat paychecks.

I just moved from the USA to the UK. It was shocking how much better the consumer banking system is here.

With 10 seconds of typing on my phone, I set up my standing rent payment to my landlord (which would have been too much money to use Chase's QuickPay at home). I gave my gym and utilities my account number and sort code and they can (with my permission, which I can revoke any time without talking to them) bill me regularly. I recorded a little video of me holding my passport and a scan of it, and that was sufficient ID for them to give me a new account. For everything under £30, I just tap my wallet on the little terminal. For everything over, they bring a terminal to me and I type my PIN. I never have to give up my card to anyone for any reason.

It's a pretty low bar to get over, but Monzo is definitely the best bank I've ever had. The US consumer banking system sucks by comparison.

Serious question: are there any large (say >2,000,000) cities in western-style democracies (for example, Singapore is quite safe, but I think their system of government makes it a little easier to fight petty crime) in which the police really do take that stuff seriously? It makes me think of that scene in The Big Lebowski where he asks the cop if they have any leads on his car breakin.

I completely sympathize with this guy, and hope I don’t wind up in his situation, but how seriously does any big city resident truly expect the police to take relatively small-value nonviolent property crimes?

I’m genuinely curious about who the target customers are for this platform. In terms of simple strategies (see signals A and B, fire order, wait for signal C, exit market, profit), there’s just no chance of a startup beating the established HFT players on execution speed. In terms of sophisticated model generation, training, optimization, etc., I don’t see being able to compete with them either. Their data and computational resources are just too good. You interviewed 100s of traders; I wonder how many truly successful ones you talked to. Did you give more weight to their input than the less successful ones?

I’m certainly biased here, as I’ve spent many years building infrastructure for large HFT firms. If nothing else, it’s made it crystal clear that there’s smart money in the market and dumb money in the market. On my own, I’m absolutely the dumb money when I compare my resources to infrastructures costing $300M or more and hundreds of the brightest people around. Unsurprisingly, I keep my money in mutual funds.

I wish you the best of luck, though. I hope you prove me wrong.

I spent a year at a “Great Books” college whose program Adler helped found. It was a really amazing luxury to be able to spend so much time on it (I left after freshman year, so I only got through the Greeks).

Ultimately, St John’s generates a lot of lawyers and humanities grad students, but not a whole lot else without extra pre-grad school to school to get some of the STEM prereqs out of the way. I love that the school generates humanities scholars, but it’s not my bag, so I moved on.

If I had the money, I’d love to go back and finish when I retire, but at 40 I’m already not as smart as I was in college, so I’m pretty sure I’ll be too dumb to be a useful classmate when I retire!

If you haven’t gotten to Euclid yet, do yourself a favor and get the Heath translations with his notes. A few of them are really, really hard for people trained in late second-millennia mathematics to be able to get through without some hints.

The article is a little light on details, but seems cool nonetheless. I’d like to see more details on the tensorflow modifications they made for parallel computing. I’ve never had awesome results using distributed tf. The parameter server model just doesn’t seem to take good advantage of a really high performance interconnect.

I'm not 100% sure that this buys you a ton since the multinode training is still done with tensorflow clusters or horovod, so there's nothing new I can see (after all of 10 minutes of looking, so take that with a grain of salt) with the actual distributed pieces. That said, workflow management is still an important piece of the training puzzle. I've seen a lot of good quantitative work put to poor use for lack of a good workflow tool.

I just finished Virus Hunters of the CDC, a 1996 book written by a couple of CDC epidemiologists, one of which had led a couple of Ebola responses in (what was then) Zaire and CAR. (He was even on the team that named the disease for a river near the town the outbreak was in.)

He certainly didn’t mention cannibalism. They were most concerned about the ritual washing before burial. Apparently the bodies are cleansed, bowels washed, etc. Given how Ebola spreads, this was virtually a guarantee that anyone handling the dead would contract the disease.

Their biggest successes were in teaching families how to use gloves, masks, and bleach. Any danger the CDC teams were in seemed to be mostly limited to political instability.

I wondered why the feed wasn’t as good these days; that’s really sad. He was really funny, and it was so unusual to associate humor with (in my experience) the most humorless branch of the federal government.

I’m not sure you understood. Regardless of whether they send out the helicopter, if they don’t take a report, the crime isn’t counted. You can’t say anything meaningful about whether crimes are getting more or less prevalent if your stats are bad.

I agree, but I think I’ve put less thought into it than you have. Just because I tend to focus on research and low-latency infrastructure, I’ve been really happy with the container-first approach. I really like being able to deal with these processes at the Linux level instead of the VM level and being able to tweak that stuff (cpu placement and isolation, accelerators and rdma network devices, etc.) There’s a reason VMs never really took off in HPC, but I think CRI-O is really poised to change the HPC paradigm, and k8s can be really beneficial in some business applications of HPC.

I definitely understand and agree that multitenancy is super important, but it would be a shame to agree that the bare metal performance is an okay sacrifice.

More than anything, I’m astonished that an article about k8s in Forbes of all places is both cogent and well-informed. There’s nothing in it that anyone who payed attention to Kubecon or k8s in the last couple years wouldn’t have known, but I’m so used to seeing Forbes push out least-common-denominator clickbaity crap that I’m actually impressed they published this article.

The title is still kind of dumb and clickbaity, but there’s no accounting for one’s editor’s taste, I suppose.