My email is eritead@dashborg.com. I'm based in London and interested in discussions related to web development. I wrote a book about delay insensitive circuit design whose landing page is at https://www.delayinsensitive.com. I've been teaching myself Go and AWS lately. My github is https://github.com/gueststar.
It's tempting to go down the rabbit hole of getting creative with
memory management as the author does, and maybe worthwhile for
educational purposes, but has anyone here observed any noticeable
performance gain from it? My impression is that libraries like
mimalloc and tcmalloc are painstakingly optimized by specialists and
unlikely to benefit from any casual embellishments beyond using the
supported malloc and free functions for everything. What does the HN
braintrust recommend?
I use a distinct key for each remote account. Although it's meant to
be infeasible to infer a private key from a public key, distinct keys
give me another layer of defence in depth in case an attacker ever
finds a way to do it. Another advantage is that I can use passphrases
selectively. For example, if I have an unpriviledged remote account
with shell access disabled because it's used only for proxy tunnelling,
it's less important to have a passphrase protecting its key than it
would be for a remote root account, and not having one could be more
convenient.
A great science fiction premise could be that a technologically
advanced adversary genetically engineers mirror life people to work as
spies and saboteurs at wiping out the rest of the population so they
can have the world for themselves, but they get rumbled by dogs
specially trained to sniff them out or various other means that
might surprise and delight the reader. I vaguely remember reading
somewhere that the difference between the smell of an orange and a
lemon is down to the chirality of a single molecule.
I'm assuming a finite alphabet and a finitely axiomatizable proof system per convention. I can't think of an uncountable set of propositions in which each can be written as a finite string of symbols, so that's what I was missing. Thank you for clearing this up for me.
I'm aware that very smart people have thought carefully about all
this, but I still can't help thinking that this argument is
unnecessarily complicated. It seems to me that a proof is something
that can be written down as a finite string of symbols, so any proof
system admits only countably many proofs. On the other hand, it's easy
to make up an example of an uncountable set of propositions. That's
too many for each of them to have a proof, so some of them must be
unprovable. What am I missing?
Being obtuse comes effortlessly to me, I'm afraid. My comment pertains
specifically to the kind mathematics that has no applications as a
rebuttal to what I read as the author's main claim it will inevitably
decline as a human endeavor. Saying that, I believe that AI has great
potential to advance engineering.
I don't move in these circles but the way I picture it is that higher
mathematics in its purest form has no applications but is rather a
medium of creative expression among a select few. How is AI generating
voluminous opaque proofs of esoteric results so different from
generating millions of hours of music nobody likes or has time to hear
that we think the former is world-changing while the latter is just
slop?
I got a UGREEN DXP6800 about two weeks ago on which I installed
FreeBSD, and was just noticing today how slow the USB is when I saw
this article mentioning a possible explanation (like ten times slower
than when I plug the same external SSD into another machine). Have you
noticed USB bandwidth issues in your setup or is it likely just some
misconfiguration on my part?
> This has even extended to fictional CSAM such as AI generated stories and pictures.
Peripheral to your comment, there is an argument
against AI generated CSAM that no one seems to be
making so allow me. AI will do the same for CSAM as it has already
done for copyright laundering. That is, it will enable its
distributors to train their models on real CSAM while obscuring the
training material. A legal regime that permits unrestricted
distribution of AI generated CSAM incentivizes actual abuse as a
source of training material. If one opposes legal prohibitions on free
speech grounds, then at a minimum there should be an audit requirement
incumbent on AI generated CSAM distributors to document their process
in sufficient detail as to establish that they haven't used actual CSAM for
training, similarly to the way porn distributors are required to
document that their models are of legal age.
The problem as stated seems ill posed to me. If Santa wakes up when
either quorum is ready, then it's never true that both are waiting
unless they arrive simultaneously. Either that's a zero probability
event that can be safely ignored, or it needs to be handled by
arbitration in some form. Attempting to resolve contention by giving
preference to one or the other shifts the problem to that of
determining whether they're close enough to be considered
simultaneous, with edge cases needing to be resolved by arbitration,
and so on.
Learning stuff can be done on your own any time. The best use of your
time in college is to cultivate professional contacts. Keep an eye out
for professors who appear to be in charge of research projects with
outside funding and volunteer to help out by doing whatever nobody
wants to do, even if it's unpaid and even if it cuts into your study
time. Be the first person that comes to mind when the professor's high
flying industry friends tell him they wish they could find some bright
highly motivated intern. If you don't think you need to try this hard
this soon to find a job I hope you're independently wealthy.
Not to be alarmist, but suppose a galactic federation judges that
humans in their current state of development will pose a danger to
other civilizations when they imminently attain warp capability, so as
a safety precaution they need to be confined to their planet for at
least a millennium. An agent of the federation posing as human
manipulates the population into allowing 100,000 satellites to be
deployed. With that done, federation scientists solve the many-body
problem for the exact necessary speed and trajectory of a small meteor
to shatter one of the satellites such that some of its fragments
precisely target its neighboring satellites, and so on, while the rest
get kicked up into higher orbits. Life goes on but any enterprise that
depends on penetrating the debris field becomes infeasible.
I blame it all on one guy. What if that plonker who broke ranks on
the English side during the battle of Hastings after the English had
successfully held the high ground against the French all day stayed in
bed that morning, and William the Conqueror just had to bugger off
back where he came from? The British Isles turn away from Europe and
develop stronger ties with Scandinavian countries over the ensuing
centuries. A kingdom of the north eventually extends to encompass
Iceland, Greenland, Newfoundland, and the Iroquois confederation well
before Columbus. With first contact happening at a time when neither
side has any technological advantage, the relationship becomes more of
an alliance than a colonization. I don't know how to extrapolate
further except to imagine that the world looks very different.
Story time. In college 40+ years ago my housemate wanted to impress
this girl by inviting her to write an important paper of hers with a
looming deadline on his computer with WordStar instead of her usual
methodology involving a typewriter. She was using WordStar comfortably
in less than five minutes but being completely new to it was unaware
of the practice of periodically saving one's work. Around three in the
morning it crashed and took all of her work with it. I was told she
burst into tears and had left when I woke up the next morning to find
my housemate busily trying to retype it from her handwritten notes.
Inform them that you've put the answer key to their upcoming final
exam on a poorly secured EC2 instance or s3 bucket. This technique
serves a variety of students in different ways. Those who have no idea
what you mean get a valuable introduction to independent study. Those
who have the chops might up their game enough to crack it. Those who
are rich enough to hire a smart kid to do it for them impart a
valuable lesson to the group on the nature of privilege. Those who are
neither smart nor rich are incentivized to stay in school. To top it
off, make the final exam consist of thought provoking essay questions
whose answer key takes the form of detailed notes and citations.
A small leak can sink a ship. The fake dollars weren't knowingly accepted.
If public confidence in the value of money is lost, we're all in big trouble.
The Secret Service was right to pursue the case zealously.
I don't know about y'all, but this paradox was resolved to my complete
satisfaction in a blog post some years ago, I believe by Scott
Aaronson, though I can't find the link. If the predictor has such a
good success rate, then it must be simulating people's brains, but
since it's not always right, the simulation isn't perfect. The best
strategy for playing this game therefore is to look for indications as
to whether I'm the real me or the simulation when the question is
posed to me, and choose accordingly. Am I floating in a sensory
deprivation tank being asked my choice by a disembodied voice with no
recollection of how I got there and no memory of my childhood? In that
case maybe I'm the simulation, so my answer is that I'll choose just
one box. Is it an ordinary day of my life and a plausible setting with
all of my faculties and recollections intact? Then I'll assume
simulated me had my back and take both boxes.
Take it to a charity shop or thrift store near you that accepts
donations of electronics so that it finds its way to some smart
kid who wants to teach himself or herself about computers.
For another example of betrayal, one of the cronies in Katherine the
Great's court always gave a dog to his girlfriends whenever he started
a new relationship. Then if the dog ever greeted some other guy
familiarly, he inferred he was falling out of favor. He probably
learned that trick when someone did it to him, because he would let
the other guy know how he was rumbled before graciously bowing out.
I worked for a company that made dedicated FPGA based hardware for
high frequency trading by deep-pocketed customers. You can certainly
implement trading strategies running directly on the board with crazy
fast turnaround times, but if a retail trader could benefit from them
we all would have quit our jobs and become independent high frequency
traders. Interestingly, at that point the Linux network stack becomes
a bottleneck so you'll want to go for a proprietary alternative whose
name I don't remember any more (not cheap), which I imagine might also
apply to a Mac, but that's neither here nor there unless you have a
server in close physical proximity to the exchange. Meta-advice: If
you want to learn about FPGAs as a matter of interest, more power to
you. Otherwise, skip the bikeshedding and learn enough about financial
markets to find a niche where you might have an edge.
FreeTube [1], and yt-dlp [2], especially in combination with a
ready supply of VPNs. Switching them around to avoid being
blocked by Google reminds me of adjusting the tuner for better
reception on an old analog tv. Infant me might have imagined a
malevolent being who inhabits the airwaves deliberately causing
interference, and in the world we've created since then that's
not far from the truth. Many thanks to the developers
tirelessly compensating for Google's frequent deliberate
breakage.
It happened to me too, and I was unable to verify myself by any
acceptable means due to being based in a country other than that of my
passport. Having been redirected somewhere else for the identity
verification onboarding, I think the process is outsourced by
Hetzner to a firm of security specialists apparently oblivious to
edge cases. Nice work if you can get it.
There's some other specific character besides spaces that's also not
permitted in passwords. It's a normal printable ascii character but I
can't remember what it is any more, and sometimes it's not caught.
Let's hope nobody signs up with it by mistake.
I have, and the technical support representative at Proton confirmed
it, but not without implying that it was my fault for using rclone. I
asked the official recommendation for Linux users to do automated or
scriptable backups onto a Proton drive and the answer was that some
kind of SDK was planned for the future. Proton drive stopped working
completely with rclone shortly after that, which was about two months
ago.
I know everyone says SK combinators can express any computable
function, but I don't get it. How do we write this function foo in
terms of SK combinators alone? Is there some obvious programming trick
I'm missing that makes it trivial? (It wouldn't be the first time.)
Former backblaze customer here, the trick is that you can't copy and
paste your password as one might do with a password manager. You have
to type it manually so that the web page can interactively tell you
how strong the password is getting as you type it.
Backblaze lost me as a customer due to the new password and 2FA
requirements, which would lock me out if I were to lose my devices,
the exact scenario I'm trying to mitigate. Not affiliated, I'm now
trying my luck with pixeldrain, mega, and koofr (having quit proton
lately as well since it broke rclone compatibility a few weeks ago).
If a computer could have an intelligent conversation, then a person
could manually execute the same program to the same effect, and
since that person could do so without understanding the
conversation, computers aren't sentient.
Analogously, some day I might be on life support. The life support
machines won't understand what I'm saying. Therefore I won't mean
it.