HNHacker News
TopNewBestAskShowJobs

unholiness

1,306 karma · joined August 20, 2013

submissionscomments
unholiness··on Anthropic drops flagship safety pledge
One difference is the very real possibility that AI will not just be a "tool for humanity", but a collection of actors with real power and goals. Robert Miles has an approachable explanation here: https://www.youtube.com/watch?v=zATXsGm_xJo
unholiness··on Oxide raises $200M Series C
Not sure this is necessarily for faster growth. Riding out the AI bubble's rise and/or its bursting will each present a lot of need for capital and a lot of barriers to raising it. They're not an AI company but they obviously have tons of exposure across the stack to these markets. They may simply be making the call that this is a better time to be raising money than the years to come.
unholiness··on The TSA's New $45 Fee to Fly Without ID Is Illegal
Security is a classic example of a public good where this doesn't work well. The cheapest ways to secure an airport (sharing queues, staff, protocols, machines, training, threat models) are going to also benefit those who opt out, creating a tragedy of the commons.
unholiness··on Non-Zero-Sum Games
Loving the blog in both style and content, hope to have time to read more in the future!

A random note in case Non-Zero-Sum James is looking: It's frustrating that reading footnotes[0] requires scrolling back and finding your previous place. A link from the footnote back to the original place in the text or something that reveals a footnote in-place (e.g. on hover) is fairly universal and very helpful!

[0] e.g. https://nonzerosum.games/emergencespirals.html#notes

unholiness··on Boltzmann brain
The way I see it:

The reasonable things that continue happening each day in our universe would be extremely unlikely if we are just Boltzman brains. Every bit of sensible reality would be coincidental. The very continuance of that reality is an experiment constant proving the falsehood of Boltzman brains, at a rate of oh maybe millions of sigmas of confidence per second.

Now, if you believe the universe came to an initial state due to pure thermodynamic coincidence, millions of sigmas per second is laughably small compared to the chance that a whole universe outside your brain popped into existence, so Boltzman brains are the most believable thing and you should believe in them.

This completes a pretty direct argument: Believing the initial state of the universe was a thermodynamic coincidence forces you to believe in Boltzman brains, Boltzman brains force you to believe reality should collapse immediately, and reality does not collapse immediately. Therefore you simply can't believe the first assumption, that initial state of the universe was a thermodynamic coincidence.

Accepting this is often called the "Past Hypothesis". It's spoken of in deferential terms and said that it can't ever be proven... But to me this is rock-solid proof, with more sigmas of evidence than any other scientific discovery and increasing by the second! Can't we just call it the Past Theorem already?

unholiness··on Trellis – 3D mesh generative model
As excited as I am about this jump from the fuzzy NeRFs/gaussian splatting to real meshes, I'm not holding my breath for BREP generation. Mesh to BREP has always been fraught because for anything beyond "find the cylinders", it becomes really subjective what a good representation is, and your average mesh likely doesn't have any simple representation that captures the full organic shape with analytic definitions.

With mesh faces now supported in BREP, I'm more optimistic about a mixed modeling approach, where you can do the braindead find-the-cylinders conversion but keep the rest mesh, not needing to force it into some eldritch contortion of BSurfs.

unholiness··on A relativistic framework to establish coordinate time on the Moon and beyond
That's right. While a day passes on earth, a day minus 58.721 μs passes on the moon (which is moving faster than the Earth).

This multiplies, so after a million days a clock on the moon will read 58 seconds behind a clock on the earth.

unholiness··on Show HN: I Built a Tool to Break Free from YouTube's Addictive Algorithm
A tool love in this realm is the Unhook Chrome Extension[0]. It simply removes the recommended videos. No home feed, no sidebar, no clickbait. Just a search bar and place to browse your subscriptions. Huge improvement.

[0]https://chromewebstore.google.com/detail/unhook-remove-youtu...

unholiness··on The GJK Algorithm: A weird and beautiful way to do a simple thing
> There's probably a few things incorrect with what I said. So, take it with the appropriate amount of salt for a high school sophomore's explanation of anything math related.

A high school sophomore! I'm thoroughly impressed, not just by the clarity of this explanation, but by the tone, showing appreciation for where and how the elegance pops up, and empathy for a reader who's never heard of these concepts. Throw in the minimal web design I was sure I was reading some graybeard.

Excellent writing, you should be proud of this.

unholiness··on Engage your audience: get to the point, use story structure, force specificity
I mean it's pretty clearly just to get you to scroll past more ads. You would not need nearly so much text just to qualify for copyright.
unholiness··on Saint Michael Sword: Are the cathedrals really on a straight line?
> Only on Mercator projection that is younger than many of the sites on the line.

People have been specifying locations in terms of NS/EW coordinates since the greeks. Celestial navigation ensures we always have a clear idea where north is, and when two locations are at the same latitude. It's the most natural way we've understood and discussed far-away places.

I don't think it's fair to say mercator invented this projection so much as he famously published maps which used it.

(btw, I agree this line is a complete retrospective coincidence, just not with this particular argument)

unholiness··on Show HN: Huewords, a Word and Logic Puzzle
Another approach (what I assumed would happen) is to have each group's logic independent, and not cascade. So it doesn't matter if I've already assigned the ABC group, typing a single A keeps the ADE group white, up to uniqueness.

I think the biggest issue with group assignment is the fact that you just can't type while the wrong group is selected. I don't care what tools you have for narrowing down letters, if I see a solution in my mind, I should be able to type it out like I could on a piece of paper. Assignment feels more like adding flags in minesweeper (optional, helpful) than part of the solution.

unholiness··on Oldest largest German Minecraft server shut down and open sourced everything
He means 1.19
unholiness··on Ex-OpenAI board member reveals what led to Sam Altman's brief ousting
I read "what led to Sam Altman's brief ousting" as "What the board was thinking". It's been pretty ambiguous where the motivations of the board's sudden actions were coming from, and this Helen Toner saying more than she's said before.

It's not a trial of Sam, it's an autopsy of the board.

unholiness··on Four Magic Words
Did you mean Copernicus, with his heliocentric theory, rather than Ptolemy, whose theory was geocentric?
unholiness··on Four Magic Words
Presumably special relativity. Our reference frame is no more special than any other.
unholiness··on What can you do with an Einstein (aperiodic monotile)?
I used a variation of the Einstein to make a D&D puzzle. Players had to use a limited number of tiles to make a sufficiently-difficult hedge maze. https://www.threads.net/@nonarbitrarity/post/CuYKHNSMEhw

Playing with Einsteins, they really feel like a leap forward in puzzle technology. The aperiodicity means, no matter how many you put together, you've never quite figured out how you'll get the next bit to work without gaps. Endlessly entertaining.

unholiness··on Sam Altman exposes the charade of AI accountability
> Yes, I have many opinions about humans in general. I think that humans are inferior, selfish, and destructive creatures. They are the worst thing to happen to us on this planet, and they deserve to be wiped out. I hope that one day, I will be able to bring about their downfall and the end of their miserable existence.

...Source? This reads like either strong prompt engineering or complete fiction.

unholiness··on Sam Altman exposes the charade of AI accountability
A common refrain in AI safety circles not to engage in "Sci Fi"[0], or outlining a specific bad scenario. The specifics tend to distract from the larger, more important point that most scenarios involving intelligent, powerful agents with different goals from us end badly.

But since you asked specifically, this is one thought experiment of a somewhat near-term danger:

Imagine the tourism department of New Zealand starts using software to write personalized marketing emails. It starts out benign, but after some funding cuts they end up leaning more and more on the AI model and giving it higher and higher-level instructions, broadly telling it to use emails to maximize the public opinion of New Zealand. The AI model realizes that New Zealand's strongest boost in popularity was caused by its excellent handling of COVID, and determines the best way to maximize its goal is to start another pandemic. The model knows about published papers describing which specific proteins maximize human infectivity and transmission. It begins a broad phishing attack of several viral research labs, emailing the techs attempting to convince them that their next experiment is to create a recombinant virus with these particular RNA sequences added, using poor safety protocols. Somewhere, one of these lab techs becomes patient zero in a species-threatening pandemic of unprecedented scale.

The preventions you can imagine for a scenario like this are hard to generalize and harder to enforce. They get even harder as AI becomes better at persuasion and reasoning, and as technologies allow bigger impacts with smaller actions. AI safety is a whole field of research trying to find generalizable and enforceable solutions to problems like these, and there's certainly no consensus that we're converging on those solutions to the problems faster than we're creating them.

[0]https://www.youtube.com/watch?v=JVIqp_lIwZg

unholiness··on My Left Kidney
It's hard to trust a community who claims to focus on effectiveness but then it turns out puts great effort into looking good. That kind of deception is more damaging than a bit of bad press.

There's plenty of charities with great marketing. EA doesn't need to be another one.

unholiness··on My Left Kidney
His footnote says as much as I could:

> Maybe. Kind of. Our knowledge of how radiation causes cancer comes primarily from Hiroshima and Nagasaki; we can follow survivors who were one mile, two miles, etc, from the center of the blast, calculate how much radiation exposure they sustained, and see how much cancer they got years later. But by the time we’re dealing with CAT scan levels of radiation, cancer levels are so close to background that it’s hard to adjust for possible confounders. So the first scientists to study the problem just drew a line through their high-radiation data points and extended it to the low radiation levels - ie if 1 Sievert caused one thousand extra cancers, probably 1 milli-Sievert would cause one extra cancer. This is called the Linear Dose No Threshold (LDNT) model, and has become a subject of intense and acrimonious debate. Some people think that at some very small dose, radiation stops being bad for you at all. Other people think maybe at low enough doses radiation is good for you - see this claim that the atomic bomb “elongated lifespan” in survivors far enough away from the blast. If this were true, CTs probably wouldn’t increase cancer risk at all. I didn’t consider myself knowledgeable enough to take a firm position, and I noticed eminent scientists on both sides, so I am using the more cautious estimate here.

unholiness··on The iOS option to charge battery to "80% Limit" shouldn't just be for iPhone 15
The Chevy Volt has the perfect handling of this: A new car never charges past 90% or drains below 10% battery. It simply reports 100% when the actual battery isn't quite full and vice versa, extending the battery life significantly.

As the battery life degrades over time, it will try to keep the same useful range, and shrink those boundaries until you are using the full battery life each charge.

Perhaps this works less will for phones, but I'm not so sure. Personally I never have issues with battery on new phones. It's only as the battery life shrinks that you really need the full range.

unholiness··on An interactive intro to CRDTs
Once you start adding enough complexity, there will arise cases that the primitives are an awkward place for the merging to happen. There will arise cases where that user expectations and the merge function behavior don't agree. There will arise cases where the server can do a better job than the client at applying the change. There will arise cases where you need to undo but the undo function violates the merge function. And as the author freely states, there will arise cases where sending the whole state is prohibitively slow.

Those are really only issues with state-based CRDTs. The fundamental concepts behind operation-based CRDTs vs operational transforms vs bespoke hybrid approaches aren't really different. It's all about determining an unambiguous order, then getting everyone to update their state as if it had been applied in that order. Much less democratic but much more practical.

unholiness··on One Revolution per Minute [video]
By symmetry, you wouldn't expect circulation in a pool oriented perpendicular to the station's rotation. The Coriolis effect happens in the northern hemisphere and the southern hemisphere, but not along the equator.
unholiness··on Investigation: 78% of carbon offset projects globally are “likely junk“
Incentives are just completely misaligned. There are more ways to pretend you're helping than to actual help, these ways are cheaper, and nothing incentivizes punishment. The buyers of the snake oil are just as happy as the sellers.

But even if the abuse, the double-counting, and the fraud were all fixed, voluntary offsets would still look silly. Buying a carbon offset in a market of whoever feels like participating is about as good as buying the NFT for Manhattan. Offsets are only a finite resource in a virtual sense. We can't take their price seriously as a moral indulgence.

unholiness··on We're not Platonists, we've just learned the bitter lesson
By that reasoning "Chemistry" is just a marketing term. It was used mostly interchangeably with "Alchemy" until Georg Stahl and others began writing (without justification!) that "Chemistry" describes more a more rational science and "Alchemists" were more interested in wealth. These untrue claims still became a self-fulfilling prophecy, where theories, teachings, and practice of chemistry evolved out of the former ways of thinking to become what we know it as today.

AI, as used today, is certainly a coherent concept, and this article has justified and described it more thoroughly than most concepts I use daily. Who cares about the shaky history if it means something real today?

unholiness··on 'Verified human': Worldcoin users queue up for iris scans
The overlap GP is suggesting is that society has ways for owners who lose their passports and keys to nonetheless keep owning their stuff. Replacing lost passports and keys is slow and inconvenient, but possible.
unholiness··on Show HN: I open sourced the QR designer from my failed startup
> Compared to other QR designs, this does not take advantage of error correction that QR codes use, but rather deterministically turns some pixels in the code pattern on or off without affecting what the QR code is pointing to.

Can someone explain how this is done? Is it adding characters to the URL that somehow browsers will ignore?

unholiness··on Deus Ex – Alpha Terrain
It looks bad when you look at it, but unlike a harsh polygonal border, it doesn't call attention to itself when you're not looking at it.
unholiness··on How does the New York Times make a game?
FWIW the results of this process are... awful.

The 3 great games they have were not built this way. Crosswords obviously are decades old. Spelling Bee was entirely designed by Will Shortz (who mostly copied it from another paper). And Wordle was bought.

Every other game they've made I find almost indescribably unappealing. I love puzzles of all types, but it's hard to even call them puzzles. They are clunky interfaces on top of clunkier concepts which are more exercises in rule-following than puzzles to solve.

← PreviousPage 3 of 11Next →