jlebar

This is literally the opposite of psychopathic behavior.

Shameless plug for my writeup about convolutions: https://jlebar.com/2023/9/11/convolutions.html

No one seems sufficiently outraged that human drivers kill 40,000 people a year in the US.

It's approximately one 9/11 a month. And that's just the deaths.

Worldwide, 1.2m people die from vehicle accidents every year; car/motorcycle crashes are the leading cause of death for people aged 5-29 worldwide.

https://www.transportation.gov/NRSS/SafetyProblem

https://www.who.int/news-room/fact-sheets/detail/road-traffi...

As a former LLVM developer and reviewer, I want to say:

1. Good for you.

2. Ignore the haters in the comments.

> my latest PR is my second-ever to LLVM and is an entire linter check.

That is so awesome.

> The code is of terrible quality and I am at 100+ comments on my latest PR.

The LLVM reviewers are big kids. They know how to ignore a PR if they don't want to review it. Don't feel bad about wasting people's time. They'll let you know.

You might be surprised how many PRs even pre-LLMs had 100+ comments. There's a lot to learn. You clearly want to learn, so you'll get there and will soon be offering a net-positive contribution to this community (or the next one you join), if you aren't already.

Best of luck on your journey.

They introduced tap-to-pay with your credit card a few weeks ago.

fastmath is absolutely not the default on any GPU compiler I have worked with (including the one I wrote).

If you want fast sqrt (or more generally, if you care at all about not getting garbage), I would recommend using an explicit approx sqrt function in your programming language rather than turning on fastmath.

Just want to suggest: Ask an LLM about it! If you have access to a reasoning model like o3, I've found it to be very helpful.

I think this answer is as good as any of the human-generated ones in the thread so far, but the real power is that you can ask it follow-up questions. https://chatgpt.com/share/6894504f-4458-8008-a8c9-f371588259...

> [Trawling around online for information] trained and sharpened invaluable skills involving critical thinking and grit.

Here's what Socrates had to say about the invention of writing.

> "For this invention will produce forgetfulness in the minds of those who learn to use it, because they will not practice their memory. Their trust in writing, produced by external characters which are no part of themselves, will discourage the use of their own memory within them. You have invented an elixir not of memory, but of reminding; and you offer your pupils the appearance of wisdom, not true wisdom, for they will read many things without instruction and will therefore seem [275b] to know many things, when they are for the most part ignorant and hard to get along with, since they are not wise, but only appear wise."

https://www.historyofinformation.com/detail.php?id=3439

I mean, he wasn't wrong! But nonetheless I think most of us communicating on an online forum would probably prefer not to go back to a world without writing. :)

You could say similar things about the internet (getting your ass to the library taught the importance of learning), calculators (you'll be worse at doing arithmetic in your head), pencil erasers (https://www.theguardian.com/commentisfree/2015/may/28/pencil...), you name it.

> Its too shallow. The deeper I go, the less it seems to be useful. This happens quick for me.

You must be using a free model like GPT-4o (or the equivalent from another provider)?

I find that o3 is consistently able to go deeper than me in anything I'm a nonexpert in, and usually can keep up with me in those areas where I am an expert.

If that's not the case for you I'd be very curious to see a full conversation transcript (in chatgpt you can share these directly from the UI).

> How can we know if his view of reality is accurate and unbiased? Without seeing the full discussion and judging for ourselves, we can't.

I agree with that of course.

> coming from someone who has an anonymous profile how do we know it's true

My profile is trivially connected to my real identity, I am not anonymous here.

I think that leaks like this have negative information value to the public.

I work at OAI, but I'm speaking for myself here. Sam talks to the company, sometimes via slack, more often in company-wide meetings, all the time. Way more than any other CEO I have worked for. This leaked message is one part of a long, continuing conversation within the company.

The vast majority of what he and others say doesn't get leaked. So you're eavesdropping on a tiny portion of a conversation. It's impossible not to take it out of context.

What's worse, you think you learned something from reading this article, even though you probably didn't, making you more confident in your conclusions when you should be less confident.

I hope everyone here gets to have the experience of seeing HN discuss something that you're an expert in. It's eye-opening to see how confidently wrong most poasters are. It certainly has humbled my own reactions to news. (In this particular instance I don't think there's so much right and wrong but more that I think if you had actually been in the room for more of the conversation you'd probably feel different.)

Btw Sam has tweeted about an open source model. Stay tuned... https://x.com/sama/status/1932573231199707168

Okay but that's not what this is about. This is saying that a foreigner cannot express private thoughts online at any point before they enter the United States.

I assume someone who goes by "15155" would believe that having private conversations online can be useful. Or do you want to post your identifying information?

Who gets to define what is anti-American?

Perhaps you think it's anti-American to believe that Israel is committing war crimes in Gaza. Perhaps I think it's anti-American to believe that the Jan 6 rioters should have been pardoned.

Whose purity test should we apply?

I think the author has misunderstood things here.

This technique is orthogonal to integer mod. Indeed the author multiplies by their magic constant and then does an integer mod to map into their hashtable's buckets.

This technique is actually just applying a fast integer hash on the input keys to the hashtable before mapping the keys to buckets. You can then map to buckets however you want.

The additional hash is useful if and only if the input hash function for your table's keys doesn't appear to be a random function, i.e. it doesn't mix its bits for whatever reason. If your input hash functions are indeed random then this is a (small but perhaps measurable) waste of time.

Using prime-numbered table sizes is another way to accomplish basically the same thing. Dividing the input hash key by a prime forces you to look at all the bits of the input. In practice these are written as division by a constant, so they use multiplies and shifts. It's basically a hash function. (Though I'd use multiply by a magic number over divide by a prime, mul alone should be faster.)

Relatedly see this post by Daniel Lemire about an alternative to integer mod, https://lemire.me/blog/2016/06/27/a-fast-alternative-to-the-... which is interesting if your number of buckets is not a power of 2 for some reason.

> As long as you don't advocate for violence, I'm sure you're fine.

This is known as the "Shirley Exception", https://vocal.media/theSwamp/the-shirley-exception

The reality is that people are being arrested and deported with no evidence of advocating violence. Here are some examples. https://www.axios.com/newsletters/axios-denver-8d5a7cc0-04fd...

I agree that people don't like being called out for their views (on race, lgbtq, women, whatever). They would rather be left to believe what they believe in peace and not face the disapprobation of others.

Calling individuals may even further radicalize them, as you say. I am not convinced on this point, I sort of think their mind is not changing either way, but maybe I am wrong.

What I am sure of is, it is not the responsibility of people whose rights are being taken away to be polite to their oppressors for the sake of rhetorical strategy.

Your argument is, "Don't say 'I believe X and if you disagree with me you're bad'. Just say, 'I believe X.'"

But then literally in the same sentence, you say, "If you do the thing I don't like (in this case, calling people bigots because they don't support lgbtq) *then you are self-aggrandizing."

"Nobody should be called a bigot for their views on lgbtq, but it's virtuous to call people self-aggrandizing for calling people bigots."

Either name-calling is okay or it's not. You can't have it both ways.

I think you may be thinking of the question, "how much greenhouse gas does it take to move a human X miles if they are self-powered versus using an ICE?"

For example this study https://www.nature.com/articles/s41598-020-66170-y.

> our estimates suggest that the net emissions associated with the ‘fuel’ required for driving, walking, and cycling may be comparable in some settings

With the caveat that they're looking at *per-mile* emissions, and nobody in the developed world is walking 60 miles to and from work every day.

However there's an important clarification to your recollection.

- You're correct that eating meat (and dairy) is very carbon-intensive.

- You're not correct that agricultural transportation is significant. It doesn't even make it into the paper above, as far as I could find. Agricultural transportation, even things like flying blueberries to California from Chile, is negligible compared to the carbon cost of food.

You can see that in the following graph. Red is transport. You have to look closely even to see it. https://ourworldindata.org/food-choice-vs-eating-local

A human burns say 2000 kcal / day. That's about 2.3kW hours / day.

An iPhone 16 pro battery appears to have about 18 watt hours. So approximately 100x less then a human uses.

I never said iMessage needs to support anything, I was merely answering a question that I thought was asked in good faith.

The US government claims that Apple is engaging in anticompetitive practices by degrading the behavior of iMessage when communicating to non-Apple devices.

Your stance seems to be, this should not be something for the government to be involved in, let the market decide.

This is ambiguous. Perhaps you believe that US antitrust laws shouldn't exist, or should be changed so they don't apply to this case, or actually don't apply to this case (ie the government is wrong that Apple's behavior violates the law).

Those are all coherent stances you could have, though I think it would be helpful if you identified which of them you hold if you want to engage in meaningful discourse with others.

When friends with iPhones send me images or videos using iMessage, they are very low-quality compared to what iPhone users receive. But when Android users send me the same, they are higher quality.

So I think the specific answer to your question is "iMessage and its lack of support for <protocol (RCS?)>".

Citation needed?

Here's the 2023 report from Roche, which owns Genentech.

https://assets.roche.com/f/176343/x/0ef2047502/ar23-roche-ho...

Page 17 shows that R&D costs for pharma is 10-20x larger than equity-based compensation. That's all stock-based compensation, for execs and regular people. Your claim is that it's the opposite.

> It's amazing and sad how quickly the hacker joy has been sucked out of this company.

I get the aversion to polished presentations like this, but I dunno, I work here and still feel like there's a lot of hacker joy still. You clearly aren't the audience for this presentation, but one audience doesn't exclude another.

Shameless plug for a recent blog post of mine where I try to explain ML convolutions, https://jlebar.com/2023/9/11/convolutions.html

Reminds me of this network routing technique: Send a packet to a random node, and then have that node forward it to its final destination.

Sounds dumb but actually it's useful sometimes? I dunno, IANA networking person.

A Scheme for Fast Parallel Communication https://ldhulipala.github.io/readings/ValiantPermutationRout...

Two classics as relevant today as ever:

End to end principle: https://web.mit.edu/Saltzer/www/publications/endtoend/endtoe...

The rise of "worse is better": https://www.jwz.org/doc/worse-is-better.html

Oh wow, I think this translates PTX (nvidia's high-level assembly code) to SPIR-V? Am I reading this right? That's...a lot.

A note to any systems hackers who might try something like this, you can also retarget clang's CUDA support to SPIR-V via a LLVM-to-SPIR-V translator. I can say with confidence that this works. :)

Because reduced demand for home purchases presumably means increased demand for renting. People have to live one way or another.

I don't think I agree with the principle of, if you can ignore the hate speech it's actually a pretty egalitarian place.

Like you say, it's not egalitarian towards those who don't have an easy time ignoring the hate speech. And, okay, in principle if the strength of one's stomach for hate speech was randomly distributed to humans with no correlation to anything else, maybe that would not disqualify it as egalitarian.

But it's not. People who face discrimination and hate speech IRL will, on average, surely have less tolerance for it online than people who don't. I have to believe that women, racial minorities, religious minorities, and so on, feel much less welcome on a place like 4chan. They've already spent their "ignoring hate speech points" ignoring someone harassing them on the subway.

So whatever it's virtues -- and I'm willing to believe you there are some -- I don't think that a place with this much toxicity can rightly be called egalitarian.