HNHacker News
TopNewBestAskShowJobs

heisenzombie

1,607 karma · joined June 19, 2012

submissionscomments
heisenzombie··on The first nuclear clock will test if fundamental constants change
Of course you’re right about photons!
heisenzombie··on The first nuclear clock will test if fundamental constants change
I think you might mean the one _electron_ conjecture. It’s fun because you have anti-electrons whose Feynman diagrams look like electrons going backwards in time. So you could conceivably be observing the tangled world line of a single electron bouncing back and forward in time — sometimes observing it as an antielectron.

Doesn’t work with photons because there’s not an anti-photon.

Anyway it’s sort of a fun “woah!” moment that Feynman was so good at producing, but I don’t think it’s taken particularly seriously as a theory.

heisenzombie··on For those who hear voices, the ‘broken brain’ explanation is harmful
I think this is a very enlightened way to understand things.

We have many models of reality available to us (whether we're talking "statistical mechanics" or "quantum chromodynamics" or "chakra energies"). Categorizing them as "True" or "False" is not very useful (is Newtonian physics "True" or "False"?). Rather, we should be content with judging our models by whether they are _useful_, and that judgement must be made contextually.

Having a mechanistic view of your experiences grounded in neuroscience is useful: It suggests that you should take your medication and that probably helps.

Having a spiritual understanding of your relationship to your experiences caused you to do things that helped you and made your life better. Does that mean that model is "True"? Who cares. It's sort of a meaningless question. It was a useful model, and that's all there is.

Somewhat relatedly, I was quite influenced by this (unfinished) work: https://metarationality.com/

heisenzombie··on I learned the language of computer programming in my 50s – what I discovered
An infinitely long lever? So in theory you can move absolutely anything, but in practice it takes an infinite amount of time to make even the smallest change happen?

Yeah, OK, could be right.

heisenzombie··on Jax-FEM: A differentiable GPU-accelerated 3D FEM solver
Paper describing this and what it’s for here:

https://arxiv.org/pdf/2212.00964 (PDF)

We emphasize the following three features that differentiate JAX-FEM from other FEM libraries:

1. Efficient solution to forward PDE with GPU acceleration;

2. Differentiable simulation for automatic inverse design;

3. Seamless integration with machine learning

heisenzombie··on Joe Biden stands down as Democratic candidate
Yeah, it certainly /can/ work. For example in New Zealand in 2017: The labor party (our big centre-left party) had an unpopular/uninspiring leader and was polling terribly. The existing leader stepped down and Jacinda Ardern took over just seven weeks before the election, and rolled to a win.

YMMV, we have very different political systems and cultures, etc. etc.

heisenzombie··on Show HN: What Beats Rock – AI Rock Paper Scissors
Absolutely! Also, I tried this and the game agrees:

Hope -> Anti-life -> universe -> a nova -> a world -> butcher bacterium -> bird of prey -> serpent -> hunter -> dire wolf -> water -> rock

heisenzombie··on Potatoes Are the Perfect Vegetable–But You're Eating Them Wrong
I’m right there with you (and also a late T1 here). Pizza is sooometimes manageable for me now with extended boluses on a pump and a closed loop to “mop up” what’s left. It’s definitely a “vibes” based dose calculation though haha.

My vote for biggest food bastard is Sushi. That sugary rice inevitably sends me to the moon (and crashing back to earth a few hours later…)

heisenzombie··on Potatoes Are the Perfect Vegetable–But You're Eating Them Wrong
Haha, I think you have mildly violated your own rule 2! Anyway, you're right that almost any statement about nutrition and diabetes can be disagreed with and picked apart. Ultimately, nutrition is complex and highly personal and exponentially more so when metabolic disorders are at play!

Taking this specific example:

"...the specific reason T1 diabetics are advised to avoid them is because of how slowly your body breaks down carbohydrates from these foods after consuming."

In my experience most medical professionals lump all potato preparations together as "high-GI foods" and therefore advise that they will raise blood sugar much more quickly than other sources of starchy carbohydrates such as oats, whole-wheat bread, or brown rice. This is backed up by evidence that, for example, baked potatoes can have a GI of higher than 100 (i.e. they raise blood sugar faster than the same number of grams of pure glucose) [1].

Your experience seems to be the opposite, since you claim a "slow" break down. And that's backed by research also! It turns out that the GI of a potato can be drastically changed by cooking method. [2]. Boiling and then cooling a potato causes starches to first gelatinize (raising GI) and then undergo retrogradation, in which hard-to-digest "resistant starch" is reformed (lowering GI again). For those reasons, a cold potato salad (with cooked-and-cooled potato) made with new potatoes (lower GI to begin with due to higher amylose) and mixed with fats (which further reduce GI), might have a shockingly low GI compared to a fresh fluffy baked potato.

All of that to nit-pick the details of your post while agreeing with your overall thesis. This topic is extremely nuanced!

[1] Miller et al. (1992) - American Journal of Clinical Nutrition: Found baked potato to have GI of 121 ± 9

[2] Ek et al. (2012) - Food & Function: Reported GI values ranging from 56 to 94 for different preparations

heisenzombie··on Woman set up a table to share her collection of washers. Nothing was for sale
Yeah, a split ring washer there is probably not doing much. Possibly even making things worse.

Given it's more or less impossible to sufficiently torque a wing nut, there are actually not many great solutions...

A dab of low strength loctite would definitely help, but that's one-time-only and you've got to carry the loctite around if you want to re-do the connection... At which point why not just carry a spanner instead and just use a properly-torqued normal nut?

Probably the best would be to swap out the plain wingnuts for nylock wing nuts, which should at least help a bit.

heisenzombie··on Woman set up a table to share her collection of washers. Nothing was for sale
If you're talking about split-ring lock washers, I would think again. They do not accomplish any kind of locking, and may make vibration-induced loosening _worse_ by providing a small opening force. They are useless for almost all purposes.

If you mean something like a Nord-Lock washer, then carry on.

heisenzombie··on Heretical thoughts about science and society (2005)
Yes, there is certainly an archetype of "Accomplished physicist who has ventured outside their area of expertise and is patronizing the locals with confident proclamations that `this would all be trivial if you knew math better and made these obvious simplifying assumptions`".

Occasionally a figure like that is actually correct — for example, Schrödinger bringing the concept of entropy over to biology and thereby influencing Francis Crick to search for the gene.

But many more are, to put it politely, not Schrödinger.

heisenzombie··on Golden Gate Claude
It's like someone gave Claude media training and now it's desparately trying to keep to its talking points...
heisenzombie··on How to keep Earth from being cooked by the ever-hotter Sun
P
heisenzombie··on Brain Uses Quantum Effects, New Study Finds [video]
Hmm, we have actually ruled out quite a lot of that sort of thing... For example we're pretty sure there's nothing like a "local hidden variable", e.g. a little deterministic process in an electron determining if it's going to be spin up or spin down when measured.

See: Bell's theorem.

heisenzombie··on TORAX is a differentiable tokamak core transport simulator
Cool, thanks for the link it's always interesting to see what's out there (there's so much interesting code around!).

My quick look at that makes me think it's more like a modern SIMION than a modern TRANSPORT.

I'm more interested at the level of accelerator beamlines (ensembles of magnets with high energy beams), and how to reduce the "impedence mismatch" between that process and the physical design optimization of individual magnets.

heisenzombie··on TORAX is a differentiable tokamak core transport simulator
Yup! Still used all the time! And both (IMHO) super un-ergonomic.
heisenzombie··on TORAX is a differentiable tokamak core transport simulator
Unfortunately it’s for my work so not currently public. One chunk of it which I would love to open source sometime, is basically just TRANSPORT but modern Python. Despite being a million years old, just second order matrix-style beam optics is actually still a really useful place to start for a lot of stuff. And with it all differentiable, one can do way more kinds of interesting optimisations.

What’s interesting is that the old Fortran source code for TRANSPORT is still out there if one digs, and most of it is taken up by thousands of lines of code that (1) implement a bunch of matrix multiplication that is now just one call to “jnp.einsum()“, and (2) implementing a tedious by-hand differentiation of those same calculations, in a few parameters, which is now basically inferior to just one call to “jax.grad()”. It’s amazing how modern tools can make hard stuff truly trivial.

heisenzombie··on TORAX is a differentiable tokamak core transport simulator
I recently started using JAX for some ion-optics work in accelerator physics. I have found it very very good. The autodiff stuff is magical for doing optimisation work, but even just as a compiled-numpy, I have found it very easy to get highly performant code. For reference, I previously tried roughly the same thing in “numba”, and wasn’t able to get anywhere near the same performance as JAX, even running on the CPU, which I understand is JAX’s weakest backend. By and large I have just written basically idiomatic Python/numpy code — sprinkled a few “vmap”s and “scan”s around, and got great results. I’m very pleased with JAX.
heisenzombie··on I keep a WTF notebook (2021)
You say the genuine motivation would be: "I acknowledge my flaws and I'm willing to put in the effort to change myself for the benefit of my wife"

But... How do I know which actions will "benefit" my wife? I argue that one of the best ways to know is to ask myself: "Will this action make her feel positively about me?". That way, I'm not going to do things that are important to me but not her, or that I think she SHOULD appreciate but she doesn't actually care about, or whatever.

Of course, to answer that question accurately requires plenty of listening, understanding and empathy.

In the past, I thought more like you. But I think it harmed me. Ultimately I came to the conclusion that intentionally doing things so that other people like to be around you isn't "disingenuous", it's a wonderful thing to work towards!

heisenzombie··on I keep a WTF notebook (2021)
I think your reaction is common, your mindset is one that I recognize in myself and causes me many insecurities in relationships both personal and professional.

However, it's worth saying that: Being intentional about relationships is not manipulation.

If I decide "I want to be a better husband" and then spend time noticing and writing down a list of things that my wife says bother her or would make her happy or she thinks would be romantic, and then I go through and choose some of them and set myself reminders in my calendar to do them... Am I "manipulating" my wife into "thinking" I'm a better husband? Or am I just plain being a better husband?

Would it be worse if I got the idea from a book titled Would it be better if, instead of being so intentional, I just let my passions and romance sweep me into doing romantic things without any conscious thought? Why?

To make my point clear: Being very intentional about relationships (how others perceive and feel about you — and what actions you take to make them feel and perceive you that way) is not manipulation. If I act in a way that makes my coworkers think that I'm a good coworker, then I AM a good coworker! The fact that it was on purpose and not accidental is...?

Manipulation happens when you develop your "be-a-good-coworker" skills (which is good) and then use those skills in a way that intentionally hurts your coworkers or makes them act against their interests (which is bad).

I see evidence in the article of the first but not the second.

heisenzombie··on Google will start showing AI-powered search results for users who didn't opt-in
I mean, that’s just completely wrong though! (1) It attempts to wrap and then substitute, which is nonsense (2) it doesn’t even expand the string you give it! That f-string in there is basically a no-op and (3) it truncates to only the first line, and throws away the rest.
heisenzombie··on Apple Is in Talks to Let Google's Gemini Power iPhone Generative AI Features
I feel like it will certainly need to be a hybrid approach for the moment. They’ll get a long way running small highly specific and tuned models locally on the neural engine for a lot of stuff. But, at some point, people are going to expect to ask questions and have something in the league of a Mistral/Claude/ChatGPT talk back and that’s just not possible with today’s hardware. Over time I expect more and more of that will get moved locally, and less will hit the “escape hatch” of throwing things to a giant LLM.

If they end up renting rather than building that capability, it says to me that they’re betting on being able to move most of that locally to the phone on a relatively short time horizon, which is interesting.

heisenzombie··on Apple announces ability to download apps directly from websites in EU
No, that misunderstands things. The technical iOS features are a relatively small part of it (Yes, there's a user dialog you have to pop up to get an IDFA, that's great!). But apps COULD pretty trivially work around that. There are a million and one ways to fingerprint a user that don't rely on an IDFA and it's impossible to restrict that kind of activity through OS-level features alone (see: https://amiunique.org/ ).

So why don't apps just work around it? Because they have to agree to terms with Apple and SAY that they won't. And then Apple has the power to kick them off the store if they do it anyway.

In addition to that, they have to list what data they collect and what they do with it when they get it. Basically they have to make a bunch of pinky-promises. Some of those pinky-promises Apple just shows to the user as a list of "here's how this company is going to use your data". Other things are just forbidden outright.

Whether a company can collect and sell certain kinds of my data to a third party is not something that can ever be controlled through an OS level feature.

heisenzombie··on Apple announces ability to download apps directly from websites in EU
Apple implemented strict rules about tracking and privacy that cost Facebook 12 billion dollars:

https://www.forbes.com/sites/kateoflahertyuk/2022/04/23/appl...

I, as a user, think that's great.

The Google play store has no similar anti-tracking rules. Why? Well, partly because those rules would negatively impact Google, but also because Meta could credibly decide to move their apps off the play store and onto their own "Meta Store: The only place to get Facebook, Instagram, Threads, and WhatsApp".

If it was easy for Meta to set up their own store for iOS, they absolutely would because currently Apple kneecaps their ability to track and sell user data in ways that costs Meta REAL money.

Google does not exert the same kind of control over the play store, so Meta is quite happy to keep Facebook there.

So, as things stand (with a closed iOS ecosystem) I can use an Instagram app on my phone that's subject to a bunch of pro-consumer rules about tracking and data protection. If iOS allowed third party stores then I would almost certainly not be able to do that any more.

heisenzombie··on Latency numbers every programmer should know
I mostly agree, and certainly a list of numbers or maybe a log plot would be better if the goal was communicating the raw data. Certainly the click-about jumpy interface is pretty janky. However there’s one thing I think this does better than a list of numbers would: Most people (me included) have a hard time getting an intuitive feel for things like just how much smaller 1ns is compared to 1ms or truly how much a billion dollars is. SI prefixes or a log scale can give the wrong /feeling/ even when they’re giving the right /information/.

Sometimes, the inconvenience of a linear scale is the point.

Pages that I think use this technique to really good effect:

https://xkcd.com/1732/ https://mkorostoff.github.io/1-pixel-wealth/

heisenzombie··on Fasting-mimicking diet causes markers indicating reduced age and disease risk
Well… Nature communications. But still.
heisenzombie··on In Mongolia, a killer winter is ravaging herds and a way of life
Perhaps there was more to it, but as told — that’s a story of incredible hubris! Truly amazing to fly across the world to tell Mongolians how to live on the steppe.
heisenzombie··on Deleting and destroying finished movies
Curiously it’s almost exactly half an order of magnitude (30*10^0.5 = 94.9)
heisenzombie··on Vision Pro teardown part 2: What’s the display resolution?
That brings up a question for me: Are there people with wildly different near and far prescriptions whose vision would be improved by looking at the world through this device which reprojects it at a fixed optical distance such that a single prescription can be used?
← PreviousPage 3 of 12Next →