nsagent

> Ideas are easy.

Let's not get into absolutes. Ideas can be as hard or harder than execution. Not many will claim ideas like Special Relativity are easy to come by.

It really depends on the idea and what you're executing. Often in business both the idea and the execution have to work out to captialize. In this case Jobs might have set the direction and Cook capitalized. Both deserve a share of the credit.

I was previously using Wipr 2, but got tired of the fact that it would block things I didn't want and there was no recourse except disabling it wholesale (for example, I noticed it blocks any element with an id of 'privacy_policy' or 'privacypolicy,' which I ran into when actually trying to read the privacy policy for a service I use).

wBlock fixes that by allowing you to have custom rules. It also includes the ability to run userscripts, so I no longer need a separate extension for that.

You might be interested in reading about the minimum description length (MDL) principle [1]. Despite all the dissenters to your argument, what your positing is quite similar to MDL. It's how you can fairly compare models (I did some research in this area for LLMs during my PhD).

Simply put, to compare models, you describe both the model and training data using a code (usual reported as number of bits). The trained model that represents the data within the fewest number of bits is the more powerful model.

This paper [2] from ICML 2021 shows a practical approach for attempting to estimate MDL for NLP models applied to text datasets.

[1]: http://www.modelselection.org/mdl/

[2]: https://proceedings.mlr.press/v139/perez21a.html

Maybe you mean it's too much effort, because I'm sure you could. I was taught touch typing on a QWERTY keyboard in the summer between 6th and 7th grade. Last year I switched to Colemak after nearly 30 years of QWERTY.

  To every man is given the key to the gates of heaven. The same key opens the gates of hell.

-Richard Feynman

https://www.goodreads.com/quotes/421467-to-every-man-is-give...

https://calteches.library.caltech.edu/1575/1/Science.pdf

I'd be 100% okay with the government stating no grant money can be used for funding DEI, but the actual wording in the grant prevents organizations like the PSF from any DEI, whether or not it is funded by the government grant. That's overly broad and seems likely to be an infringement of the first amendment (though I'm not a lawyer, so this is my lay interpretation).

  do not, and will not during the term of this financial assistance award, operate any programs that advance or promote DEI, or discriminatory equity ideology in violation of Federal anti-discrimination laws.

The government can certainly add restrictions to the use of the grant money, but applying that broadly over any actions the grantee performs during that time is overreach. I wonder about the legality of that condition.

Agree that the snarky reply doesn't help, but the quotation isn't really skewed (though it is a paraphrase). It comes from an internal memo that leaked in 2018 that states:

  Maybe someone dies in a terrorist attack coordinated on our tools.

  And still we connect people.

  The ugly truth is that we believe in connecting people so deeply that anything that allows us to connect more people more often is *de facto* good.

This comment [1] linked to an article [2] with the leaked memo.

[1]: https://news.ycombinator.com/item?id=45721016

[2]: https://techthelead.com/incendiary-leaked-memo-facebook/

I read the report and it actually made me realize how much of a propaganda campaign HRW was engaging in. It reminded me of PETA campaigns and how offputting many of them were (this is coming from someone who was vegan for 7 years in spite of them).

I know that organizations like HRW and SPLC have to draw attention to topics, but I found the bias and lack of nuance in the report very troubling. The report suffers from the same sort of bias that is so prevalent in most reporting these days, which has gotten to be tiresome.

If HRW reviewed over a thousand cases of censorship, why don't they provide the raw, unedited examples? Instead they include categories of examples, like stating the slogan "From the river to the sea Palestine will be free" was frequently censored. For many that is seen as a call for a different type of genocide, one which HRW gives no indication of whatsoever, simply stating:

  For instance, the words in each of these statements on their face do not constitute incitement to violence, discrimination, or hostility.

It's true that other slogans that were censored are neutral, such as "Ceasefire now" and "Stop the genocide." But lumping the first phrase in as if there were no legimate concerns with it is disingenuous at best.

That said, I'll still support HRW in much of its work, but I hate the tactics that mirror the broad cultural shift to inject more and more biased viewpoints. I really want to go back to the time where bias in the media and nonprofit organizations was much less pronounced in general.

I've used fengari and it's a great option for lua in the browser, but it is quite slow.

Also posted yesterday: https://news.ycombinator.com/item?id=45672467

Using LLMs to review papers isn't going to fix the fundamental issues with publishing in AI/CV/ML/NLP. I can't help but think I got out of the academia rat race just in time.

This is exactly how I read research papers and how I advocate others read them as well. As you read try to figure out how you would solve the problem outlined and what experiments you would need to perform.

You're right that there is no purely mathematical argument; it's almost non-sensical to claim such.

Instead you can simply make the rather reasonable observation that LLMs are a product of their training distribution, which only contains partial coverage of all possible observable states of the world. Some highly regular observable states are thus likely missing, but an embodied agent (like a human) would be able to correctly reason about these states due to other sensory modes that observe these states. It's very clear that the number of possible states far exceeds any text describing those states (the number of states transcribed to text is vanishingly small in comparison, but they are locally coherent and quite regular when observed through other sensory modes like vision). That's the closest you can make to a mathematical argument.

As a result one common failure case for an LLM might involve describing a specific physical environment and associated set of actions that are in an area of low (or no) probability mass in the training data and asking the LLM to reason about it. Here's an example where I asked ChatGPT the following:

  imagine i walked into a room through a doorway, then turned around and walked right back out of the room. from the perspective of a person in the room facing the door, which side would my right arm be? please explain.

It responded by incorrectly stating the following:

  Let’s break this down step by step.

  1. *You walk into the room through the doorway*:

   * As you're entering, the person in the room is facing the door. So, you're entering from the **outside**.
   * Your **right arm** will be on the **right side** of your body, just as it is normally when you're standing.

  2. *You turn around and walk right back out*:

   * When you turn around to walk out, you’ll be facing **away** from the room, looking back toward the doorway.
   * Since you're walking **back out**, the person inside the room is still facing the door, while you are walking away from them, so your **right arm** is still on the right side of your body.

  So, from the perspective of the person in the room, as you’re exiting, your *right arm* will still be on their *left side*. That’s because they’re facing the door, and your right side is to their left when you’re turning around and leaving.

  To visualize:

  * Imagine a door in front of them.
  * You walk through and turn around, so you're now facing away from them.
  * Your right arm is on your right side, but to them, it's on their left side since you’re now moving in the opposite direction.

Thanks. I've never heard it framed this way. I've mentioned my appetite diminishes after aerobic exercise, but a common response is that I'm weird.

Interestingly my appetite sometimes skyrockets after certain activities like bouldering or weight lifting, so the diminished hunger response must be a bit more complex of a phenomenon.

I also used to submit to *ACL conferences exclusively, and was even a SAC, but got out of the academic game altogether. I'm still conducting research, but as an independent researcher for my own startup. The system seems to be more and more bananas over time and no one is willing to just force a change. It's really a tragedy of the commons.

That said, why don't conferences work like journals: if you're rejected you cannot resubmit. Find a new conference. That gets rid of the queuing problem. Yes, you'll have some amazing papers that will not be accepted by a top conference. So what, it happens to everyone. Plenty of influential papers were not published in a top conference in ML/AI/NLP.

Bought the game when it came out, but still haven't had the time to play. Just flew out for a three week vacation with my Steam Deck in tow. Unfortunately, I left it on the plane and I haven't heard back from lost and found yet (seems unlikely I'll get it back considering it was an international flight). Oh well.

I don't think this is any better of an argument.

Maybe yours is a more apt analogy, but as a very competent driver I can't tell you how often lane assist has driven me crazy.

If I could simply rely on it in all situations, then it would be fine. It's the death of a thousand cuts each and every time it behaves less than ideally that gets to me and I've had to turn it off in every single car a I've driven that has it.

I initially tried that and thought it improved some things, but it increased contrast across the OS, such that some webpages, including stock HN became too blinding. I instead switched to "Reduce Transparency," but that has its own issues.

Overall not pleased. I really did not want to care about the UI changes at all. But having experienced it now, I'm so annoyed I upgraded to iOS 26 and I'm having trouble focusing on the screen. I want WebGPU support, but I'm very hesitant to upgrade to macOS 26 (which is required for WebGPU in Safari).

  Idk what to tell you. Email is mostly a solved problem for most cases

Idk what to tell you. Considering email is mostly a solved problem, Proton must be extra incompetent for inadvertently deleting people's emails due to multiple different bugs in their code that took them far too long to address (multiple years in some cases).

https://old.reddit.com/r/ProtonMail/comments/t8vwhf/deleting... https://news.ycombinator.com/item?id=33432296 https://old.reddit.com/r/ProtonMail/comments/yjz3yu/proton_b... https://old.reddit.com/r/ProtonMail/comments/1j79x7j/has_the...

(The temerity of the customer service response on that last one, saying they have no clue about the bug being asked about is galling, but par for the course for them).

BTW, make flippant responses, get responses in kind. Normally I'd ignore this idiocy, but today was your lucky day. Anyway, it's clear you're just a troll and I've indulged you enough.

The person I was responding to literally said they were "a paying user for a very long time" and "never encountered a bug". No software is bug free. I can't think of a single software service I've used for as long as Proton (7 years now) where I haven't encountered a single issue over that time. I take their statement to be so incredibly unlikely as to be facetious or intentionally duplicitous.

So I responded in kind, because I've definitely seen company cheerleaders, and I'll have no part of it. I'm glad you all are happy with Proton. I'm not telling you to leave.

And if you really want to see complaints, you don't have to look far. Read the other comments on this thread. I don't have to spell everything out for you.

Proton seems to have a lot of cheerleaders that come out of the woodwork when anyone complains. I'm happy that somehow their code is magically bug free for you, since you've somehow never encountered any bugs whatsoever in their code (despite their release notes mentioning literal bugs they've fixed).

I'm glad it works for you, but their offering is frequently buggy and broken for me.

To be honest, I've found Proton's public customer service representatives to be very duplicitous, so it's hard to take their word at face value. It's pretty ridiculous to see their response to legitimate concerns start with: "That doesn't sound right..." 80-90% of the time.

I've need a paying subscriber to Proton since 2018, but I recently canceled my subscription (which ends in November). I just got fed up with the constant bugginess and jankiness of their offerings.

Any suggestions for mail hosting and VPN? I hear good things about Fastmail and mailbox.org (I see they very recently rebranded to just mailbox and revamped their offering).

Also, I've been a heavy user of the SimpleLogin alias service. Any suggestions for easily porting all those accounts to a new provider? Manually changing each and every account to a new email seems painful.

I was previously a long time listener and donator to NPR (similarly for NYT), but their progressive bias for the last decade has seriously degraded the quality of their reporting. I remember when their articles and radio coverage was much more balanced.

For that, I think The Economist is much better. It has more direct reporting, with seemingly less editorializing. Try their news in brief to keep up.

I think the main reason I've pulled back from the consuming mainstream media is directly tied with the change in reporting style rather than the news fundamentally being more depressing or anxiety-inducing.

For example, I was listening to Left Right and Center until a few weeks ago when Sarah Isgur departed. The show really should have been called Left Left and Center, because if anything Sarah Isgur was more center leaning while Steve Inskeep is definitely quite progressive. Now the show feels even more lopsided. It's as if journalists are so entrenched with their point of view that they can't see the wider landscape. I truly wonder if social media has clouded journalists' perception as well, which might be contributing to this phenomenon.

I really do want balanced coverage. I want to know what each side of the political debate actually thinks, from their own mouth. It turns out that a lot of the people I was indoctrinated to vilify, were in fact people who believed differently than I did, but certainly weren't so toxic as to be simply pilloried for their beliefs. That approach is tiresome and I've lost hope that such reporting will return. That's why I've given up.

Then those people deserve what they get as a result. People should have autonomy to make less than optimal decisions if they want.

And to provide a counterpoint, my dad can barely navigate his iPhone. I literally spent an hour on the phone with him when he was lost and needed directions; it took 20 minutes to guide him to open the messages app so he could read the address I sent. Someone that clueless isn't searching the internet to figure out workarounds for installing anything.

Amen Jack! That's why I quit my postdoc and I'm working on my own AGI research.

It seems that neither academia nor industry has much appetite for pursuing interesting, but risky, research ideas.

You don't win people over by singling them out.

Yeah, he joined Trion soon after we started development.

Interesting! I didn't realize you all experimented with an MMORTS. I was a lead engineer on a never released MMORTS [1] that was developed after WoW became big. Seems that's a tough genre to convert to MMO.

[1]: https://youtu.be/2JRkogAL3Uo