HNHacker News
TopNewBestAskShowJobs

tjradcliffe

2,541 karma · joined June 17, 2014

Poet, physicist, engineer, author: http://www.tjradcliffe.com/?p=1422
submissionscomments
tjradcliffe··on Ten Years of Git: An Interview with Linus Torvalds
I used Git for a couple of years then changed to Mercurial and have never looked back, except now and then to wonder at the size of the crowd that has gathered around Git, driven in no small part by the success of github (which is not a bad thing!)

Git is an entirely appropriate tool for managing as something as large and complex as the Linux kernel, but I'm doing much simpler stuff, and Mercurial is simpler and has a much shallower learning curve. I'd recommend it unequivocally to teams of modest size working on average applications.

The state of git documentation is much better today than it was five or six years ago, so maybe its complexity is less of a big deal, but so far I've not encountered any issues with Mercurial that make me think "If only I was using git this would be easy!"

What I found with git was I had to maintain a fairly complex mental model of the current state of affairs, and because I'm not very intelligent that took quite a bit of effort. With Mercurial the model is much simpler, so I can spend more of my very limited attention on writing code. It's quite possible that I simply never took the time to learn git properly, but with Mercurial I didn't have to.

tjradcliffe··on Burn rate says little about whether a startup is on track
No examples, but the observation that a great deal of success depends on luck, or chance, or whatever you want to call it. Factors beyond your immediate control. The lower your burn rate the longer your runway, all else being equal. The longer your runway, the more opportunities you have for the dice to roll your way.

Admittedly, all else will not be equal as your vary your burn rate, but keeping it down the bare minimum required to do the essential technological development, business development, marketing and sales will always give you better odds of long-term success. That "bare minimum" rate may actually be very high, but it's still something any CEO should be paying careful attention to.

tjradcliffe··on The Case of the 5-Millisecond Cosmic Radio Burst (2014)
> Telling someone there's no such thing as the centrifugal force in physics is a good starting point but you shouldn't stop there.

No, it's a terrible starting point. Maybe the worst possible starting point. "Here is something you know to be true that I am going to tell yo is false, creating confusion and a tendency to reject anything else I say as obvious nonsense."

End with the notion of fictitious forces (which do exist, so the name is terrible), don't start with them, unless you want to be like one of those dreadful people who talks about negative generalized temperature or negative generalized resistance to neophytes without first explaining the generalization, so you deliberately mislead people into thinking you are talking about the familiar, ungeneralized concept.

There is no evidence whatsoever that such introductions make anyone more motivated or able to understand, and a good deal of reason to expect they don't.

tjradcliffe··on So Your Company Has Been Found Using Alex’s Photographs Without Permission
It is well-known and well-documented that our ability to infer tone from electronic communications is poor, so your chances of being correct in judging his tone are not much over 50%: http://www.entrepreneur.com/article/239121

That's better than chance, but not really enough to pass judgment on someone. This may be why you have been downvoted. It may also be that others read his tone (with 56% accuracy) differently.

The tone of this message is meant to be neutral and informative.

tjradcliffe··on Nature Ontologies
The problem with the semantic web is that radical meaning variance is alive and well. As a friend who works in GIS for geology says, "If I send a team of geologists out to map a region, when I put their data together I can tell who mapped where but not what anyone mapped."

There are ways in carefully controlled situations (like geological mapping) to deal with this, but in the uncontrolled wilds of the web we don't have any very good idea of how to handle it. Published ontologies are a hopeful attempt to get people to agree, but how they are interpreted by different individuals will still vary a great deal, because it always involves mapping the individual's ontologies (plural) onto the published document and that will rarely happen cleanly.

To take a likely quite common example: a published ontology may put A and B in the same genus, whereas you or I might give each one a different genus, based on our understanding of the world. You see this a lot when crossing between fields, as a pure physicist will often have a quite different set of abstract categories than an applied physicist, an engineer, or a mathematician, much less a geologist or biologist. And individuals within fields vary a great deal as well. So what you end up with, while better than nothing, is still some kind of weird and disjoint intersection between the published ontology and the personal ontologies of all the people who use it.

Automated tools for processing documents and inferring their semantic categories may help enough for the semantic web to be useful, but it'll take a concerted effort to ensure that such tools produce vaguely similar results, and we aren't there yet by a long shot. And then there's the problem of deliberate gaming.

As such, I don't think the semantic web is "under-rated" so much as it is still, after almost 20 years, in the early days of development work on a very hard problem.

tjradcliffe··on We Should Have a Better Condom by Now
In Vancouver they get handed out on the Granville Street strip by some crazy lady with a bucket full of them, as well as being freely available many other places.
tjradcliffe··on We Should Have a Better Condom by Now
The most interesting thing is also depressing: there was a fairly well-known researcher at the University of Manitoba who was working on designing a better condom in the early 1990's, and there is still work going on there today: http://news.nationalpost.com/news/condoms-treated-with-silve...

The work I recall from the '90's (I was a post-doc there, working in a different field) involved changes to both shape and design, with some kind of extra band around the base to prevent leakage and accidental slip-off.

As near as I can tell, none of this research has had any impact on condoms as they are manufactured and sold. The best you see is a non-latex condom being sold now and then, but they never seem to stay on the market very long (they are generally more expensive).

So the most pressing topic for research would seem to be going after the question, "We know all kinds of things about building a better condom, and have for decades, so why haven't the condoms we buy changed appreciably?"

Some of this will be due to the difficulties of getting things past the FDA, particularly for smaller operations as the article suggests, but since we do see non-latex condoms from major manufacturers (http://www.walgreens.com/q/non-latex-condoms) it is clearly not the case that the FDA makes stuff impossible. It may be there are economic issues at work as well, but I don't think the situation is so obvious that we can just apply $IDEOLOGY_OF_YOUR_CHOICE and say we're done. This is particularly true because condom availability isn't great outside the US, either (although what the US approves does have an influence on what gets sold elsewhere, I'm sure.)

tjradcliffe··on Fish Oil Claims Not Supported by Research
This is interesting, but 1) it does not test the effects of fish oil as a mood stabilizer and 2) there is something sketchy about the way the data are reported.

1) It is a study of vitamin, mineral and essential oil supplements. There is no basis for ascribing any measured effect to essential oils as such.

2) In a number-rich report, it's odd that they don't report either the number or distribution of incidents in any detail, and I can't make their averages work out.

For example, they have 172 people in the study for an average of 142 days, and they talk about numbers like 11 and 16 incidents per thousand person-days, which would lead one to predict around 170x140x(11 to 16)/1000 ~ 250 - 300 incidents, but they then go on to talk about studying 754 Governor and minor reports, or 338 Governor reports and 416 minor reports (in all cases the reports appear to cover the full study population, which is the majority of the prison population).

One can conclude from this that most of the incidents in the prison are due to the people who weren't in the study, maybe? I don't know, but there's something odd going on.

Furthermore, they don't given any idea of individual variation within the population. The conclusion you'd draw if most of the incidents are associated with a few individuals vs spread out across the population might not be the same. Were all individuals pretty much similar in their response, or was the effect localized to some? The results as reported are consistent with a few problematic individuals becoming peaceful and everyone else getting slightly worse.

When making an argument that an intervention is good for a population, you really have to dig into the support for the claim across the population, not report averages as if they could only arise one way.

So while this is intriguing an interesting, it isn't as strongly convincing as one would like.

What I'd like to see (and would have insisted on as a reviewer) is a table showing the raw numbers of minor and Governor's reports issued on each population over the study period, and at least histograms showing the number of individuals in each populations with N Governor's and M minor reports against them, which would allow one to judge the distribution of incidents in the population.

tjradcliffe··on Fish Oil Claims Not Supported by Research
The effects seen in the various sub-groups and endpoints are all non-significant at the p = 0.05 level. The lowest I can see is p = 0.07 (cardiac death prevention, as reported in their Table 3).

The multiple-hypothesis correction they apply is reasonably appropriate in the case of simply looking at all results, which is what they do.

Furthermore, because the outcomes are disjoint (cardiac survival vs stroke vs sudden death vs all-cause mortality, etc) there is no simple way to combine the results.

To take a silly example, studying the effect of seatbelts on cancer and cardiac death might well show a bit of an effect on both (because people who wear seatbelts are generally healthier, say) but it would be illegitimate to combine those two studies because the endpoints are (so far as we know) unrelated to each other. Without some kind of causal account the issues become very deep and difficult to say anything very definitive about.

So I'd say their statistical treatment is fair and appropriate. If fish-oil is supposed to have such a large effect as to be worth taking the risk that it increases the risk of prostate cancer, its effect should be unequivocally measureable in population studies. That is not the case.

For what it's worth, I think the prostate cancer studies are at least as flawed, at least the one I've seen, which is a case-control study that shows an extremely modest increase in relative risk of the kind it is very easy to produce from statistically identical populations: http://www.tjradcliffe.com/?p=1745

tjradcliffe··on Dark Energy Tested on a Tabletop
Sometimes non-physicists accuse physicists of ad hockery with regard to dark matter and dark energy, and chameleon models are very useful in that case because they show what ad hockery really looks like:

"Chameleon models are not especially well motivated from the standpoint of fundamental physics, admits Burrage, who began studying them in graduate school, but since dark energy presents such a profound mystery, physicists are willing to consider just about anything."

They are also something to bring up when your New Age [1] friends claim that some weird effect "can't be tested by science" because it only happens when no one is paying attention, or something. Chameleon models are specifically designed to be "not there when there is a detection apparatus present", and yet to be useful as an explanation at all they have to have some interaction with the rest of the world, so it turns out we can test them regardless.

This is a general property of things that are useful in explanation: if they can explain something, they can cause something, and if they can cause something, they can be investigated by systematic observation, controlled experiment, and/or Bayesian inference.

[1] I just realized how dated the term "New Age" sounds. It turns out it went out of fashion in the '90's: https://en.wikipedia.org/wiki/New_Age#Social_and_political_m.... We really need an inclusive term for the whole swath of people from across the political spectrum who reject science.

tjradcliffe··on How a robot pharmacist wound up giving an overdose
> If these people in the course of their work choose to ignore alerts

This is where your mistake is coming from. You think for some reason that people are choosing to ignore the alerts, rather than being habituated to ignore them.

It is theoretically possible for a person with a decade or so of training to maintain a constant level of attention on repetitive trivia. Studies on Buddhist monks have shown this. But the people who were doing this didn't study Buddhist meditation, they studied medicine, so expecting them to behave like Buddhist monks, which you seem to be doing, is the height of irrationality.

Humans have very limited attentional resources and extremely powerful internal non-conscious filters to conserve those resources that can only be overcome by lengthy and rigorous training, which would exclude the kind of training these people actually needed to do the rest of their job. These are just facts, well-established by a wealth of empirical research over the past several decades.

Responding "Give me a break" to a system whose failure was fully predictable based on known facts of human psychology and physiology is precisely the problem. Perhaps someone on the design team pointed out the issues, and were told by someone who knew nothing about human psychology beyond what their gut told them, "Give me a break!"

tjradcliffe··on Quantum entanglement of a single particle has been observed by researchers
Dunno why this is getting downvotes, as it is simply a statement of fact. If we could distinguish a stream of photons that had "collapsed" onto a particular basis from one where individual photons were in a superposition of states on that basis we would be able to communicate faster than light using EPR pairs. It is generally believed this is impossible.
tjradcliffe··on Quantum entanglement of a single particle has been observed by researchers
Current thinking is that "collapse" is a hack. We've always talked about the wavefunction collapsing but never been able to describe the onotology of collapse. Non-locality seemed for a time to explain why that is so: because collapse happened non-locally, it could happily violate the law of non-contradiction from the viewpoint of a Lorentz-invariant observer.

The alternative to collapse is decoherence due to entanglement with a heat bath. There are still unexplained issues with this, in my view, but that may simply be because I don't fully understand how the ontology of the classical world is constructed by this process, or why once that ontology is constructed the beings that are governed by its laws are only capable of being aware of the foundations via the most indirect kind of inference.

In any case, this experiment shows some of the power of "entanglement" vs "collapse" as an account of measurement. If "collapse" is a short-hand for measurements that involve "entanglement with a heat bath" then it is possible perform a measurement that involves entangling part of a wavefunction with a coherent signal (the local oscillator in a homodyne detector) without doing to it what we would call "collapse". That appears (to me at least) to be what is happening in the present case. No magic involved.

tjradcliffe··on Quantum entanglement of a single particle has been observed by researchers
I'm competent in QM but not a quantum optics expert. In particular homodyne detection is new-ish to me. There is a bit of a description of it here that might be useful: http://relativity.livingreviews.org/Articles/lrr-2012-5/arti... That said, this is my take, which is mostly me trying to wrap my head around the problem, so take it all with a grain of salt.

The idea is that Alice mixes the (weak) signal photon stream in her lab with a (strong) "local oscillator" of the same frequency (that is the "homo" in "homodyne") and uses the interference between them to perform measurements on the signal without doing photon counting on it, which when combined with Bob's measurements on the other part of the signal photon wavefunction can demonstrate non-local effects. It is important, as always in "spooky-actions-at-a-distance" experiments to emphasize that nothing Bob sees can be used to infer what Alice measures or vice versa: there is no possibility of faster-than-light communication, and it is only when the measurements are combined after the fact that the non-locality becomes manifest.

Homodyne measurement seems to be the key thing that makes measurements on single photons possible, and this may be one of those cases where the notion of "collapse" breaks down in favour of "entanglement": the part of the signal wavefunction in Alice's detector doesn't collapse, it just gets entangled with the local oscillator, and because everything is still coherent her results can still be combined with results from the wave function components in Bob's lab. Entanglement with a heat bath emulates collapse; entanglement with a coherent local oscillator does not. [I'm still agnostic on the claim "entanglement solves the measurement problem" because I don't think it properly answers the question "why is there a classical world at all?", but that may be just me.]

There are a number of loopholes in previous experiments that this closes. I'm pretty sure it closes all detection efficiency loopholes, and there is a subtle critique of Aspect's experiments regarding the timing of the two-photon cascade that this makes irrelevant. There is a small (and in my view fairly implausible) literature on timing and photon-pair-identification issues that goes after two-photon experiments, and this work is not subject to any of these criticisms. I'm not sure how Joy Christian's work on Clifford algebras would be applied to this experiment either, although I expect they will have something to say about it.

tjradcliffe··on China's Man-On-the-Side Attack on GitHub
> Of course the Chinese government knows this and was likely not responsible for the attack.

This is an example of the logical fallacy of "argumentum ad stultum", or "appeal to stupidty".

It goes like this:

- X would be stupid. - No one would ever do anything stupid. : Therefore no one would ever do X.

There are so many counter-examples to this argument that they hardly bear mentioning. People do stupid things every day of the week and twice on Sundays. Organizations multiply stupidity as often as they moderate it.

It may be that this wasn't the Chinese government, but pointing out that it would be stupid for them to do so is not an argument against it at all.

tjradcliffe··on Ask HN: Who are your favorite poets?
Thomas Grey ("Elegy Written in a Country Churchyard," with its lyric celebration of ordinary life, is arguably the greatest English poem).

TS Elliot, on his good days (Prufrock counts as a good day, as do the Landscape poems. He oversteps in "The Waste Land".)

Tennyson, when in the mood for melancholy.

Robert Herrick, when in the mood for insouciance, contemplation and the poet's struggle.

tjradcliffe··on No bullshit guide to math and physics
What "math" means to people outside of math--including physicists and engineers--is completely different from what "math" means to people inside math. We really ought to have a different word for the two fields. Even "applied math" doesn't cut it, because most applied mathematicians still think like mathematicians, and it ain't like the rest of us.

What mathematicians don't see is that when they rail against the kind of concrete, heuristic definitions that non-mathematicians use for parts of mathematics, is that it isn't just that we don't care that there is a better, more general, more abstract, more justifiable definition out there, it's that we can't understand them so they aren't useful to us, even a little bit. You may as well try explaining Shakespeare to your dog: it'll frustrate you and annoy the dog, who is perfectly happy to use Shakespeare for his intended purpose, which is to say, chewing on the edges of the "Complete Shakespeare" for entertainment.

That's an overstatement, but the gulf between the mathematician's mind-set and the mind-set of physicists and engineers is still far wider than most mathematicians appreciate, and it really would be valuable to have a different word for what physicists and engineers do when we "do math", because it is not "math" the way you mean it.

tjradcliffe··on Beijing to Shut All Major Coal Power Plants to Cut Pollution
Coal has an energy density of about (depending on type) 24 MJ/kg, which is a shade under 7 kWh per kg, so your final number is off by a factor of 7 or so [EDIT: I think? Or you're saying that coal creates 1 kg CO2 per kWh? Ok... Yeah, I think that's right!]
tjradcliffe··on Ellen Pao Loses Silicon Valley Bias Case Against Kleiner Perkins
Could someone who downvoted this explain why?
tjradcliffe··on Ellen Pao Loses Silicon Valley Bias Case Against Kleiner Perkins
Sure, like there's overwhelming institutional bias against women in medicine... oh, wait. Err... Like there's overwhelming institutional bias against women in law... Um... Well, like there's overwhelming institutional bias against women in polit... Hmm...

The credibility of the claim that there's overwhelming institutional bias against women is dependent on understanding why in cases where there unquestionably was considerable bias--like medicine, law and politics--the bias has either dropped dramatically or been reversed in the past half-century to the point where no one anywhere is the least bit surprised to see a female doctor, lawyer, or politician.

The absence of women in tech may be due to many factors. Given we have seen how easily bias has been overcome in other areas, we should at least keep our minds open to alternative models. One such model is the observation that women remain under-represented in areas where the working conditions are terrible, the career prospects are nil, and the tangible effect of their efforts on others are zero, all of which conditions pertain to tech, but not to medicine, law or politics.

This suggests that maybe we ought to be paying attention to the conditions men work under in tech, rather than making this out to be a man-woman thing.

I'm not claiming this is necessarily true, but nor do I think it is entirely implausible, and hope people concerned with these things will at least be open to alternative explanations.

tjradcliffe··on How I taught my dog to text me selfies
This reminded me of a weird little web-comic-ish thing about a guy who taught his cat to text: http://louisvsrick.com/
tjradcliffe··on What Would You Recommend to a Bright 14 Year Old?
Terry Pratchett's Discworld books, particularly:

"Mort" and "Reaper Man"; "Guards! Guards!", "Men at Arms", and "The Night Watch"; "Wyrd Sisters", "Witches Abroad" and "Lords and Ladies"; and "Small Gods" and "Pyramids".

The first two are about Death, the second three about Samuel Vimes, the third group about Granny Weatherwax, and the last two are stand-alones. All the books are worth reading, but those are some of the most engaging and compelling, and they teach skepticism, humanity, kindness, courage and absurdity in equal measures.

tjradcliffe··on A Strong Welfare State Produces More Entrepreneurs
There are at least two ways of looking at this: one is to compare countries, the other to compare what the same individual would do in different countries.

There are quite a few non-Americans working in the American technology sector. I've worked on teams with 30% American content, the rest coming from around the world. Talking to those people, it's clear that they are taking advantage of the best of both worlds: the large capital accumulations that are available for investment in the US, but with the backup plan of going back home to a sensible social welfare structure if things don't work out.

I've been in the same boat: a Canadian who has lived and worked in both the US and Canada. I've taken risks--particularly as a husband and father--that my American colleagues could only take because they were childless and had spouses who worked. Knowing that no matter what happened, my kids would have access to a health-care system that produces significantly better outcomes at significantly lower cost to the public purse than the American mess, was without question an enabler for me. I was very conscious of this in making business decisions, including the decision to go into business (while broke and unemployed in the wake of the dot-com crash, as it happens.)

Had I been in the US at the time I would have been scrambling to find a job so my kids would be covered by decent health insurance.

Maybe my case is atypical, but some kind of comparison between foreigners living/have lived in the US vs natives might shed some light on this question.

tjradcliffe··on A significant amount of programming is done by superstition
Nonsense. We test stuff all the time we don't understand "completely" (which is some kind of magical standard, as we obviously don't understand the universe completely, so the very atoms and molecules that the machine is made of are in some sense not well understood.)

Certainty is the alchemist's stone of philosophy. People have done all kinds of interesting stuff while attempting the impossible: turning knowledge into certainty. They can't, but that doesn't mean knowledge--which is inherently uncertain--is inferior. It means certainty is inferior, because it can't actually be achieved, regardless of what any pre-Bayesian thinker might have imagined.

Knowledge is always uncertain.

Tests are always incomplete.

This does not mean "knowledge is impossible" or "you cannot say you tested the code" but rather "certainty is a chimera" and "compete test coverage is a futile goal."

The whole point if good engineering is to adopt standards that are both achievable and useful, and the kind of test coverage that people run on code they are unfamiliar with to ensure it will do what they want it to is generally adequate for that. Evidence: most shipping code actually works.

tjradcliffe··on A significant amount of programming is done by superstition
Except that it's not blind and it's not faith.

We generally have some idea of what's in the black boxes we use, and we are always open to change our beliefs based on new evidence ("faith" is a belief that will stay the same no matter what evidence is presented: in Bayesian terms it is a belief with a prior of 0 or 1, neither of which can be changed by any evidence.)

"Hopeful" or "optimistic" programming might be better.

tjradcliffe··on A significant amount of programming is done by superstition
The first principles of computation are not the first principles of software development. A computer is not a Turing Machine: it is an electro-mechanical wrapper around the limited imitation of a Turing-like core, and that wrapper and the physical properties of the system put extreme limitations on what can or cannot be proven.

Software development is a perfectly ordinary engineering activity, or should be (although it is mostly done by non-engineers), and the same standards should apply. No mechanical engineer or electrical engineer has ever shipped a "provably correct" machine. They have shipped machines that conform to best practices, including heuristic analyses of MTBF and so on.

tjradcliffe··on A significant amount of programming is done by superstition
So an even better alternative title would be "The Author of This Essay Does Not Understand Engineering, Software, Technology, Medicine, Or Any of the Other Practical Arts Which Have Always Involved Shipping Imperfectly Understood Machines".
tjradcliffe··on Sam Altman: Why Hardware Could Yield the Next $10B Startups
Also: EMC testing, RFI, etc. Certified by a test lab that will cover both North America and Europe.

On shipping: these days it's pretty common to ship everything with mechanical accelerometers in the packaging that trigger on high loads so you can tell when the crate has been dropped. Crazy how often it happens.

tjradcliffe··on ‘I Would Prefer Not To’: The Origins of the White Collar Worker
> Why hasn't "peak office" been reached yet?

Because most offices are staffed by people working for arrogant old people who can't imagine getting through life without staff, because they themselves are not actually capable of doing anything. [Disclaimer: I am both arrogant and an old person, but do not fulfill the requirements for "arrogant old person" because I actually know how to do stuff.]

This is a generational shift. I'm old enough to remember when a certain class of older executive would become enraged when they saw a scientist or engineer in their business unit typing at a keyboard. In their world, secretaries typed up the long-hand notes that scientists and engineers wrote. It wasn't so long ago.

Likewise, the notion that an executive today might book their own travel, or pay their own bills (because really, it's all a couple of mouse-clicks and hardly a significant distraction from thinking the Deep Thoughts[tm] that execs are paid to think) is still anathema to many executives older than their late 40's.

As they retire we will hit "peak office", but not before. My bet is somewhere in the 2025-2030 range.

tjradcliffe··on The Future Bubble
The claim the author leads off with is not plausible: capitalism has seen enormous changes in social relations since its earliest days.

The author is a sociologist, not an economic historian, but they should still be aware of that economic historians (even ones on the Left) tend to place the emergence of capitalism sometime in the 1600's, and break it up into multiple phases, each of which is embodied by significantly different social relations.

Once upon various times anti-capitalists proclaimed that unionization, emancipation, universal suffrage (giving the vote to all men rather than just men of property), women's suffrage, labour parties and social democracy would all overthrow the capitalist order precisely because these were seen as radical departures from existing social relations upon which capitalism was presumed to depend.

Instead, capitalism has proven an enormously resilient mode of economic organization that can be instantiated in societies with extremely different social relations.

So the claim that the modern worker in a mature capitalist society like Sweden or Canada or Germany (to pick a few at random) stands in the same social relationship to their employer as did workers in 17th century Amsterdam or 18th century London or 19th century New York or 20th century Shanghai is extremely implausible.

To further claim that social relations in the future must remain as they are today for the preservation of the capitalism mode of economic organization is equally implausible.

← PreviousPage 8 of 25Next →