xyience

I just ask the interviewers how much freedom they have in the interviewing process (usually it's pretty clear) and whether they would do something different or not if they could. This kind of follows naturally from questions about non-dev responsibilities that devs have. A lot of fellow engineers at my current company believe their little quizzes actually have a lot of signal value when they really don't, I avoid working with them since they tend to have other weird beliefs. If my interviewer is giving me a whiteboard problem somewhat reluctantly because it's what he's supposed to do rather than what he'd like to do, he's probably OK to work with. The followup question is then whether I'd be working with him or not, since so many of these companies just throw you in an interview loop with whoever is available that day...

Surface Pro / Surface Book. I presume Linux support on them is good enough now, but at least if you're stuck with Windows you can get a proper bash environment now without cygwin that I hear works well.

Maybe some fingers of the government are demanding a backdoor, but other fingers of the government want tor to be as secure as possible. Tor is useful for the operations of national intelligence agents, too, and can only maintain that use when they know it doesn't have a backdoor which say the Chinese could then discover and exploit.

Panzer Dragoon Orta was on the x-box. :)

Sega should really make an HD remastering of all of them though. Or even a new one. Especially after the sadness of Star Fox Zero.

How do you know the stuff at the gas station is actually 91 and not 87, how do you know the regulators do their jobs properly and aren't bought out by Big Oil?

The answer to all questions is that random private individuals or groups do the tests. If it's a big enough concern among their customers, some of them will do it. Some customers at regular gas stations already do these tests on their own because not all gas stations are the same. (A few people even make their own gas.)

And it works great with NoScript. What a breath of fresh air from the SPA stuff I work around every day.

> (you can use GCC but the code quality wouldn't be that good)

Maybe.. on the other hand you can actually use the full program space available, and gdb. (Last time I used Keil was 2011, I quickly moved to a GCC toolchain rather than pay for a license. They're really not selling me on their tool when they arbitrarily restrict my max program size such that I can't actually evaluate it properly.)

Tests don't magically make changes safe or give the confidence to make changes.

Skimmed the PDF, it seems like the point of this paper is just to show that yes, classical systems can (with memory, e.g. the reference https://arxiv.org/abs/1007.3650) simulate quantum ones, and thus finding "characteristic quantum numbers" shouldn't make one immediately suspect something quantum is going on. In the discussion it's like the ultimate nitpick: "The characteristic trait of QT rely on the fact that the quantum bounds are achieved without employing extra resources such as memory. Therefore, the principles needed to fully derive QT (in the spirit of Refs. [33–37]) should account for that."

There are some people who find the concept of quantum physics philosophically displeasing and try however they can to ignore all the experimental evidence and say we're really in a classical universe. This isn't a case of one of those.

You can do that with some proprietary software too -- proprietary doesn't have to mean no source code and no source code modification, unfortunately it has come to mean that for too many businesses and users willing to pay money to those businesses, so yeah, it's a point in favor of open source. I also appreciate that contributing to and encouraging use of open source has a societal benefit of training my eventual replacements for free, and open source of a particular license guarantees improvements to that software from other parties than the primary maintainer remain open.

There's a point in this talk https://www.youtube.com/watch?v=5Vt8zqhHe_c where I realized he is a brilliant comedian, but also pretty far left, and a lot of his funny criticisms just boil down to criticizing for not being as far left as he is or wants you to be. So I don't think his mind is being poisoned -- it's already poisoned, from the point of view of someone not as far left. ;)

I think a lot of tech people lean libertarian, which is a strange double-think ground of left and right views, because the role of a technologist has inherently anti-left and anti-right components. Technology is upstream of culture, upstream of society, to work on it implies defying authority whether in the form of a single sovereign or public opinion. While much of the world is still the same, much has been remade according to man's desire, through technology, and there is still much more to be remade. And often not a collective mankind, but an individual, often male, or a small group of individuals, often male. That kind of influence is inherently anti-democratic, anti-egalitarian, but anti-authoritarian as it is highly individualistic. Yet its benefits can be made private (until someone else figures out how to do the same, anyway) or open to all according to the whim of the creator, because it's fundamentally just knowledge that once derived can be shared endlessly.

The comments about techies not going out of their domain to try and solve things is just nonsense, as cross domain work happens all the time. Applied technology is often the catalyst (and money source) that brings multiple disciplines together who otherwise would be perfectly content sitting around in academia pursuing their narrow focuses for all time (this isn't to say that such narrow pursuing can't bring great fruits). Ok, it's not the only catalyst, and maybe a lot of technologists have that habit and viewpoint of being able to do better than narrow field experts, but either they succeed and are shown to be right all along that they didn't need outside help, or they are wrong, and they fail, or they change their minds as they realize they're failing and study up and get consultants. I don't know anything about Google's anti-aging prospects but if they aren't already in contact with Aubrey DeGrey and the other biologists and doctors he's in contact with they soon will be once they realize how hard the problem is.

Is he rich from his superior models?

It depends on the mutual fund and the ETF. For Vanguard 500, the VOO ETF and the VFIAX index fund are effectively the same, they even have the same expense ratio. The latter you need $10k upfront though. (VFINX, if you only have between $3k and $10k, has a higher expense ratio.)

These things aren't inevitable, scaring people into exercise won't work. But even if they were, it wouldn't be terrible to die '"on my back, lying on my many rolls of fat, scarcely uttering a word, taking labored breaths, and eating my fill," for of all the ways a man might die, an excess of luxury was the only truly happy death'. (From http://www.dimensionsmagazine.com/dimtext/kjn/people/heavies...)

Reminds me of this: https://www.youtube.com/watch?v=9JYkMhQ9gf8

Kids need a motive. With a motive, they'll figure things out, you don't have to handhold them every step or create broken abstractions for them to play around in then get bored.

We already have "visual/physical programming" for kids, in the form of Minecraft. And for the kids who want to go the extra mile, well, they learn Java. Not enough Java they could work at BigCo, because they learn Java with the motive to do stuff in Minecraft, not to actually understand the semantics of Java -- but if they got bored with Minecraft, their retained Java knowledge would be enough that they could then teach themselves the more formal aspects, or even another language.

Kids don't want to know the difference between '=' and '==' and '==='.

I think you need an additional component: no other desires except work. Many of us don't have jobs that fully satisfy (or let's be honest, even barely satisfy, since even 50% satisfaction would be nice and tolerable at some places) our minds, and once the thought patterns of "I can't wait for the day to be over so that I can X" start, they're hard to stop. Then you kind of need monk-like discipline to continually deny yourself your greater desires over the timespan of years if you don't want to go crazy / suffer burnout.

What's your burn rate? That is, if you stop doing any income-producing activities, how long can you live off of savings? Considering burning through them all for a long period of no work. Or if you think you can manage for longer, try to get to the magic number where 4% of your investments (assuming at least 5% annual ROI) covers your expenses, thus making you 'retired'.

Recovery (at least partial) is still possible, but you might be at the point where the first year of recovery looks like a lot of doing nothing but sleeping: http://jacquesmattheij.com/dealing-with-burn-out

Haven't watched the video yet, but then Conway and others formalized a bunch of theory for 2-player combinatorial games like Go, using surreal numbers for game states. (e.g. the number * = {0|0}, or, whoever moves first wins.)

You will never have the technology to decrypt aes256.

So is <insert other modern OS>. That doesn't mean it should be auto-installed.

There are historical examples of discontinuities, though I don't think the FOOM debate will be settled soon. It may be quite a while before we even get to "dumb AGI", but the hardest part of that is the G part. Right now "they don't work" is indistinguishable from "they don't exist", but if we get that G, I don't see how you could claim either. From there, even if we suppose it's another huge leap to get to true super-intelligence instead of a FOOM, the time to get to merely smart AGI, and indeed smarter-than-human AGI, would be short, if only for the basic advantages of a silicon machine substrate. If all we had were human brains running on silicon even, that would be enough to quickly reach superhuman general intelligence, even if not true super- (or perhaps ultra- as I.G. Good originally put it) intelligence that we expect for a Singularity event.

With higher level languages it's even less important to know theoretic complexities, just represent the problem in the most natural representation and optimize later. (You're already giving up orders of magnitude in exchange for dev time by using a higher level language.) Good engineering practices like being able to swap out structures and algorithms later matter a lot more.

Operations/techniques/algorithms are great to know when performance is critical, the subset knowledge of worst-case time and space complexities not really. (And in interviews do you ask for best case or average-under-certain-probability-distributions cases? How expansive is the trivia matrix you get candidates to fill in?) They can sometimes help guide, but they'll just as often mislead. Bubble sort and insertion sort are both n^2 algorithms, but you should never use bubble sort, and if you never profile you'll be confused why my quicker quicksort uses insertion sort inside when n is small since quicksort is n * log(n) so it should always be better right? (Conveniently ignoring worst case for quicksort is n^2 because people usually just memorize the n * log(n) factoid.) Or you'll be confused why I don't use quicksort at all if I have multiple CPU cores to create a parallel algorithm on. Throw in other aspects of modern hardware like caches (try swapping the loop order in code that needs to loop over all cells in an NxM matrix and note how much performance can suffer if you do it wrong) and branch prediction and special instructions your CPU architecture supports, those are very good to know too if you care a lot about performance. For complexity analysis in the real world, "constants matter", but that doesn't really capture all of the caveats.

Maybe the classic case of my overall point is the existence of https://en.wikipedia.org/wiki/Simplex_algorithm#Efficiency

Edit: one interview problem I like to give (given I have to give one and can't do things my preferred way) is the jump game: https://leetcode.com/problems/jump-game/ Rot13 commentary: Bar fbyhgvba vf whfg qrcgu-svefg frnepu. Va Clguba zl pbqr sbe gung vf 12 yvarf, vg hfrf n fgnpx (jvgu n abezny clguba neenl; vg pna or n dhrhr gbb vs lbh whfg punatr gur nethzrag gb .cbc()) naq n frg, ohg lbh pbhyq whfg hfr n cynva neenl sbe rirelguvat. V yvxr guvf ceboyrz fvapr vg yrgf zr frr vs gur crefba xabjf ubj gb nccyl bgure fgehpgherf orfvqrf cynva neenlf. N ybg bs crbcyr pna erpnyy gur rkvfgrapr bs frgf naq dhrhrf naq znlor svyy va gurve pbzcyrkvgvrf bs PEHQ bcrengvbaf ba n zngevk ohg pna'g npghnyyl nccyl gurz be nal nytbevguzf orlbaq ovanel frnepu naq gur onfvp PEHQ vzcyrzragngvbaf sbe rnpu onfvp fgehpgher.

Can you make an unhackable person? If people can get around your safeguards, even only people with authority rather than in general, the AGI can too. Trying to keep a potentially malicious (actively or passively so, I mean only bad for humans) AGI in a box cannot work, we either get the internal parts right so we can be very confident it's not malicious to begin with (and so no need for a box attempt) or we fail.

The Golden Age trilogy is great. A Fire Upon the Deep is also good.

I don't think there's much reason to be confident that the first AGI will be like a low-IQ human, at least for very long, even if it starts off as an emulated human brain. Machines have the huge advantage of faster materials (neurons are slow), perfect memory, perfect calculation, ability to scale horizontally, backups to restore in the event of bad self-modification experiments, and no need for things like sleep and food to take up its time from learning, improving itself, and acting upon the world.

Is your point that there are unaddressed safety concerns with existing tech? While true, none of them are really existential threats, whereas something with greater than human intelligence yet none of the limitations of a single biological body to upkeep is such a threat.

It'd be nice if they made an area available for random mall-goers who are also software engineers to come voluntarily hang out and instruct / help out random kids or teens that stop by, with potential for networking or further tutoring services for those really interested or who want to do something more interesting. Unfortunately I bet if a lot of us just camped out unofficially for a couple hours we'd be branded as child predators or something, and escorted out by mall security in any case...

I also don't see why one would necessarily need to reason from complexity arguments, rather than more intuitive "I'm using a set because I want to use set operations like union and membership". If performance is crucial, reasoning by complexity analysis is not sufficient and often not even necessary.