HNHacker News
TopNewBestAskShowJobs

laichzeit0

2,543 karma · joined February 7, 2011

submissionscomments
laichzeit0··on Is Matrix Multiplication Ugly?
Well function composition f(g(x)) is not the same as g(f(x)) and when you represent f and g as matrices relative to some suitable set of basis functions then obviously AB and BA should be different. If the multiplication was defined any different, that wouldn’t work.
laichzeit0··on Short Little Difficult Books
Rudin’s Principles of Mathematical Analysis aka Baby Rudin. Small, short, terse yet has brought many undergraduates to tears.
laichzeit0··on Iceland reports the presence of mosquitoes as climate warms
There were similar such comments during Covid where people were saddened that people were still debating whether it was a lab-leak or not. Dogmatism, on which ever side, unless maybe in a field like mathematics which is entirely deductive, is not good.
laichzeit0··on Vijaye Raji to become CTO of Applications with acquisition of Statsig
Tell me you've never used Statsig without telling me you've never used Statsig. It's an online controlled experiment platform. You know like when you want to figure out if a blue button gets more clicks than a red button or a green button, and you want to avoid the situation of some tech-bro calculating the average clicks on all 3 groups and going "this one is higher, so it must be better" because they have zero background in statistics and don't know how to correct for multiple comparisons or what power analysis is, etc.
laichzeit0··on IQ tests results for AI
I read the book Bell Curve years ago, but I remember the analysis being that the found statistically significant differences between race and IQ. The authors argued that individual differences in IQ within a population are strongly influenced by genetics (heritability estimates around 40–80%). They emphasized that this doesn’t mean IQ is fixed, but that genes play a large role in explaining why individuals differ. Their ultimate policy argument was less about race per se, and more about what society can realistically do. They argued that large-scale social programs (e.g., Head Start, income redistribution, affirmative action) had limited power to reduce cognitive inequality or close gaps, because much of IQ variation was resistant to environmental manipulation. On the genetic vs. environmental debate about group differences, their ultimate claim was: we don’t know, but genetics might contribute, and pretending otherwise could be harmful to honest policy discussion.

But really, if you can't go about doing more studies on race and IQ, we'll never really know. It's a valid and legitimate scientific question

laichzeit0··on IQ tests results for AI
So were their claims falsified?
laichzeit0··on Python performance myths and fairy tales
I use Python 90% of my day and I can't say I like it or hate it or care about it at all. I use it because it has all the libraries I need, and LLMs seem to know it pretty well too. It's a great language for people that don't actually care about programming languages and just want to get stuff done.
laichzeit0··on Writing an LLM from scratch, part 13 – attention heads are dumb
Neither are most functions, but locally, at a point, a linear approximation works just fine in practice.
laichzeit0··on Derivation and Intuition behind Poisson distribution
But this just gives the definition of the distribution. No intuition about where it might have come from, it just appears magically out of thin air and shows some properties it has in the limit.
laichzeit0··on Adipose tissue retains an epigenetic memory of obesity after weight loss
“Any diet” is still putting you at a net caloric deficit. There is no other principle that can explain weight loss that I know of. Whether you’re counting calories to achieve that deficit or following a “diet” it comes down to the exact same thing. Counting calories just demystifies the process of weight loss.
laichzeit0··on The Agent2Agent Protocol (A2A)
The LLM returns a message called ToolMessage which then describes which function to call and the parameters (you register these functions/tools as part of the initialisation step, like when you pass it temperature, or whatever other options your LLM allows you to set). So think of it as instead of steaming back text, it’s streaming back text to tell you “please call this function with these args” and you can do with that what you want. Ideally you’d call that function and then give the output back to the LLM. Nothing magic really.
laichzeit0··on How University Students Use Claude
I don’t know where you live, but I haven’t used nor carried cash on me for at least 5 years now. Everything either takes card or just tap using your phone/watch. Everything. Parking meters, cashiers, online shopping, filling up your car. I live in a “third world” country too.
laichzeit0··on The Agent2Agent Protocol (A2A)
Take a look at the samples: https://google.github.io/A2A/#/documentation
laichzeit0··on Interviewing a software engineer who prepared with AI
Well I find conversations with people in interviews to be less of a game than giving them “homework” to do, given that unless they’re totally green with no work experience, I’d assume they would actually have some stuff they’ve worked on and would like to talk about.

It’s completely bizarre to me that take home assignments have been normalized as part of an interview with professional working people.

laichzeit0··on Interviewing a software engineer who prepared with AI
Don't see the point of these "take home projects". Just ask them what's the most difficult technical thing they had to do before, and have them walk you through it, probe, ask questions. If you don't like the one they talked about, ask about another one, or another one. You can generally weed out the bullshitters, they talk alot about "we" and hardly ever use "I" meaning they didn't do anything.
laichzeit0··on Evaluating modular RAG with reasoning models
It's not just the retrieval, tool calls entail another call to the LLM (ToolMessage) and possibly the result will then require other tool calls. Massive latency.
laichzeit0··on I got OpenTelemetry to work. But why was it so complicated?
Well let's be fair. You can't get the type of telemetry Dyntrace provides "for free". You have to pay for it somewhere. Pretty sure you can exclude the agent from instrumenting performance critical parts of the code, if that is your concern.
laichzeit0··on Statistical Rethinking (2024 Edition)
I happened to install everything two days ago. R version 4.3.3 (I use RSwitch to switch between R versions on Mac). You should use REnv for dependency management. There were no problems installing the rethinking package, the Cmdrstan package just needed to be installed with devtools instead of install.packages.

I’m mostly a Python guy, and didn’t find it particularly hard to get this going. Although I’m always left scratching my head when using RStudio/Renv/R. It’s such a horrible environment (always hanging, crashing, slow, the tooling sucks ass). I refuse to believe that I’m the only person who has RStudio hang and require a restart or get stuck on some uninterruptible process and requires forcing killing it at least once a day.

laichzeit0··on Trump wins presidency for second time
Did you reply to the wrong comment? Nothing what you said addresses illegal immigration. Are you saying illegal immigration is something good and if you’re against it you’re for child rape?
laichzeit0··on Trump wins presidency for second time
What is the problem with deporting people who are there illegally? As someone who doesn’t live on the border of the United States do you know how incredibly hard it is to legally immigrate there? I don’t see why other people should be allowed to jump the line. There’s a legal way to get in, follow it like everyone else.
laichzeit0··on Boston Dynamics robot Atlas goes hands on [video]
The BD demos are cool and I’ve been seeing them for probably a decade or more now, but what exactly do these robots do? Besides for cool demos, where are they actually used?
laichzeit0··on What sank the Bayesian superyacht in Italy?
What would the advantage to not have it lowered be? Like you're sitting there, sails down, anchored. Why would you go "Nope, better if keel isn't lowered"?
laichzeit0··on Survival Analysis Part I: Basic concepts and first analyses (2003)
https://autonlab.org/auton-survival is also pretty good and is built on PyTorch.
laichzeit0··on You are not dumb, you just lack the prerequisites
I always sucked at math, even though I did it in undergrad. I basically did this over the course of the last five years to try get better. It went something like this:

Spivak - Calculus. This was a bad idea. Got maybe 30% of it. Gave up at Taylor series.

Hammack - Book of Proof. Finally understood how to prove things, and induction arguments.

Abbot - Understanding analysis. Got far, things fell apart around the Gamma function.

Apostol - Volume I. Got better at calculus. Also trigonometry. Exercises were easy. Skipped differential equations. It was too hard.

Hoffman/Kunze - Linear Algebra. Gave up after a few chapters, too hard.

Friedman/Insel - Linear Algebra. Much better, got to the Spectral Theorem and gave up.

Rudin - Principals of Mathematical Analysis. Absolutely brutal, probably got 30% of it.

Abbot, round 2. Much easier this time, got through the whole book.

Spivak, round 2. Much better, got through the whole book. Actually found it easy.

Hubbard - Vector Calculus. Gave up early, it was too hard.

Apostol - Volume 2. Much better. Stopped somewhere in the middle when it got too focused on differential equations and physics stuff.

Back to Friedberg / Insel - Made it through the spectral theorem.

In between I was doing a lot of mathematical statistics and probabilty stuff like Casell-Berger (I did this book twice, each time going back to the math where I floundered). I’ve worked through just about every exercise in the above books and watched YouTube video lectures where they exist (there is a good one for Rudin). Solution manuals sometimes exist, sometimes you have to find university courses based on the books and look for homework assignments where they have posted solutions, Quizlet has ok solutions, some are buggy. Apostol volume I some dude worked through and posted online.

Anyway point is I refused to accept how stupid I am and I brutally forced myself to become better at math. My attitude was I don’t give a fuck how long it takes, I will keep going until I get better.

I think I’m better now, although I’m still shit. It’s true what von Neumann said: In mathematics you don’t understand things, you just get used to them.

laichzeit0··on You are not dumb, you just lack the prerequisites
I’m bilingual since a child but didn’t really use one of the languages much after age 12 to about 30. When I do speak the second language I can fool locals into thinking it’s my first language, but it takes exactly 1 mistake or mispronunciation and they ask “oh are you actually English?”. The bar is that that high for passing native fluency in another language, if you somehow could fake the accent perfectly as well (there is absolutely no way).
laichzeit0··on Artificial intelligence is losing hype
That's great. Then don't use it? I however find it immensely useful and will continue to use it.
laichzeit0··on Uncovered Euripides fragments are 'kind of a big deal'
TLG is pretty easy to get hold of. You also want Diogenes, an app that allows you to use that database. It’s very useful if you study Latin or Greek, even as a hobby.
laichzeit0··on Bayesian Statistics: The three cultures
What about automated predictive modeling pipelines? In other words, I want the best possible point estimates only on future data. I’d think, regardless of the model selection process, I want to reestimate the parameters on the entire dataset before I deploy it, so as not to “waste” data? I.e. I want to use the hold out test data in the final model. Is this valid?
laichzeit0··on What John von Neumann did at Los Alamos (2020)
Yeah it’s like people thinking anyone can be a Micheal Jordan or Tiger Woods if they just tried hard enough, started young enough, were born in the right place to the right parents enough. Some people just have the gift, it’s like neural networks: some random weight initialisations are just better, some maybe even so close to optimum that almost no training is needed at all.
laichzeit0··on Is Tableau Dead?
I'm the opposite. I love Tableau Creator and hate working in the browser. There's also no Power BI creator desktop app for Mac (nor will there ever be, I guess), so I refuse to use it.
← PreviousPage 3 of 34Next →