jmalicki

20 ms·

511.
by jmalicki 7mo ago
We recently had a lot more probationary ATCs cut because of specific action by Elon Musk's DOGE, which has since then resulted in two major air disasters due to poor ATC handling, which was virtually unheard of before. This isn't
512.
by jmalicki 7mo ago
Did they? If there's evidence great!
513.
by jmalicki 7mo ago
Okay, so then what is? Most jobs have this failure mode because there's a tendency to strip funding until disaster happens, even when it was clearly foreseeable.
514.
by jmalicki 7mo ago
Maybe the ppl they hired had a lower attrition rate! Maybe the people hired required less supervision and training than the CTI graduates would have! Maybe this had rippling effects on increasing their hiring pipeline as people of color w
515.
by jmalicki 7mo ago
It's been awhile, but I once spent four hours trying to get Opus 4.1 to change the color palette of a bar graph, that would've taken me 5 minutes. The only reason I wasted so much time was that given all of my other positive exper
516.
by jmalicki 7mo ago
You can just shortcut all of that if you're a Bayesian and just plain say "p-values are posterior probabilities under a uniform (improper) prior" and save everyone a lot of time. And if you're doing that, don't care
517.
by jmalicki 7mo ago
It's not the hard part, just the tedious part that takes the most time.
518.
by jmalicki 7mo ago
Strongly agreed! I think getting good code organization out of an LLM is one of the subtler things - I've learned quite a bit about what sort of things need to be specified, realizing that the LLM isn't actively learning my prefer
519.
by jmalicki 7mo ago
It depends - there are some very very difficult things that can still be easily verifiable! For instance, if you are working on a compiler and have a huge test database of code to compile that all has tests itself, "all sample code mus
520.
by jmalicki 7mo ago
"I also encounter edge cases where they fail surprisingly badly multiple times per day. " If 80% of the time they 10x my output, and the other 20% I can say "well they failed, I guess this one I have to do manually" - th
521.
by jmalicki 7mo ago
It's not even just 3d printing skill, but understanding the physiology of how the bone and teeth will restructure themselves over time to know the right amount of pressure to apply to encourage that remodeling, without so much pressure
522.
by jmalicki 7mo ago
Most people I've seen complain say things like "I asked it for code and it didn't compile." The real magic of LLMs comes when they iterate until completion until the code compiles and the test passes, and you don't
523.
by jmalicki 7mo ago
I do always laugh when I see a Bayesian object to p-values, then use a Bayesian procedure that is mathematically identical to treating p values as posterior probabilities. Just saying the word "Bayesian" doesn't actually make
524.
by jmalicki 7mo ago
There is some frequentist procedure there, but it seems hard to not recognize the deep connection to Bayesian statistics and wonder if you should begin to question your baseline assumptions. Since the entire justification for using a shrin
525.
by jmalicki 7mo ago
I'm talking about, for something simple, the negative sampling in word2vec. Or the temperature setting for an LLM etc.
526.
by jmalicki 7mo ago
I guess you have me thinking more... things like Parzen window estimators or other KDEs are frequentist... But while it's a probability distribution, to a frequentist they are estimating the fixed parameters of a distribution. The dist
527.
by jmalicki 7mo ago
I feel like I'm a polyglot here but primarily a native frequentist thinker. I've found Bayesian methods shine in cases of an "intractible partition function". Cases such as language models, where the cardinality of your
528.
by jmalicki 7mo ago
I agree... I feel like "The Elements of Statistical Learning" was possibly one of the first "postmodern" things where "well, frequentist and Bayesian are just tools in the toolbox, we now know they're not so in
529.
by jmalicki 7mo ago
They do! https://arxiv.org/pdf/2510.18777 But that doesn't mean a frequentist views a VAE as a generative model! Putting it another way, Gaussian processes originated as a frequentist technique! But to a frequent
530.
by jmalicki 7mo ago
Thank you for Lindley's paradox! TIL
531.
by jmalicki 7mo ago
Huh? Are there really any pure frequentists post Stein's paradox? At least ones that are aware of it and maintain objections to fusing the fields?
532.
by jmalicki 7mo ago
I'm not well versed enough, but what would a frequentist generative model even mean? The entire generative concept implicitly assumes that parameters have probability distributions themselves that naturally give rise to generative mode
533.
by jmalicki 7mo ago
Given that, shouldn't you be even more concerned about people YOLOing it, if even highly trained orthodontists are regularly screwing this up?
534.
by jmalicki 7mo ago
These kinds of things can slowly move your bone structure over time. After all, that is their entire point . You don't want to accidentally mess up your teeth and jaw even more.
535.
by jmalicki 7mo ago
The README.md doesn't really explain what it is or why I'd want it, just directory structure and how to install
536.
by jmalicki 7mo ago
420ppm seems to ignore how bad it actually is for a lot of humans. A lot of buildings have over 1000ppm of CO2 due to being poorly ventilated, levels where humans start to have noticeable cognitive decline. Spending so much time at home pos
537.
by jmalicki 7mo ago
Ok wow. I mean the default model being Grok, whatever - that everyone sets to their favorite. But the hidden use of a different model is wow.
538.
by jmalicki 7mo ago
That is exactly why I posted it.
539.
by jmalicki 7mo ago
I want to love this project - but can you edit that to give a more succinct summary of what it is without the hype building?
540.
by jmalicki 7mo ago
I wish this had a README.md written by a human so I could understand what the project was about, since it sounds cool on the surface.

More ›