mnl

kellogh, 4 months ago to random

i’ve become disappointed with the lobste.rs community lately. it used to be the best technical discussion on the internet, to the point that i would skip the article just to read the comments, they were that good.

but the comments on my latest blog were 100% about packaging stuff, like appearance, linking, etc. it’s weird but hacker news is now where the real conversation happens and lobste.rs became a cesspool

https://lobste.rs/s/xnzvea/htmx_is_composable

mnl, 4 months ago to random

elasticsearch, after I complained for the nth time about it's $#@$!% lack of schemas and documentation:

{"dynamic": "false"}

"false" <- a string, not a bool

FML

hrefna, 4 months ago to random

I keep thinking about that diagram that showed the execs directing AIs rather than workers.

It's such a dystopian nightmare of a future, but also it's just so laughable. That's not how this works, that's not how any of this works.

It's still a dystopian nightmare of a future, but replacing CEOs with AIs is so much more doable than replacing rank-and-file in so so many cases

CEOs make inexplicable decisions and people respond by saying "well surely they just know best." Perfect for an AI.

mnl, 4 months ago to random

I find it weird that this "copilot assisted code is less secure than manually written code, especially in the hand of beginners" gets boosted around like some gotcha.

Any code written by a human is insecure unless proven so. The dismal result in the control group show that manually handcrafted code is almost just as problematic.

The problem here isn't LLM or not LLM, it's the software development workflow around security critical software.

1/

mnl, 4 months ago to llm

You can't both complain about AI not meaning anything and then keep complaining about "AI" as if it is a thing...

I really wish I understood the psychological triggers that lead to such lack of clarity. Say "OpenAI" or "LLMs" or "diffusion models" or "classifiers" or "training corpus" or "LLM backed search" or whatever.

There's plenty to argue and debate about there. Else we conflate applications and research and companies and ethics and a conversation can't be had.

#llm #llms #ai

1/

mnl, 4 months ago to LLMs

"with llms all conversation is going to be bullet points to email back to bullet points".

Yo when I write an email, I first make myself a list of bullet points and then write them out.

When I receive an email, I read it and if I'm actually conscientious I will study it and write out some bullet points for my notes.. Otherwise I'll just let it wash over me and whatever sticks sticks.

Human communication is not some magic "i beam my thoughts into yours" stuff.

#llms

1/

hazelweakly, 4 months ago to random

Currently addicted to this song

https://youtu.be/h5jz8xdpR0M?si=vy73IOFih2yqzFQv

Anyone got similar song recommendations? The things I'm particularly vibing with are:

• powerful lyrics
• the emotional strength and compression in the vocal style
• the reverb style and echoing of the vocals
• I'm an absolute slut for halftime shuffle style beats and this one is straight delicious

I want more 👀

(bonus points if it's sapphic or witchy vibes)

kellogh, 4 months ago to random

i find the AI criticisms of Emily Bender & friends to be valuable, but they’re misquoted frequently to support arguments that AI is worthless. the papers themselves essentially just argue that the terms researchers use to talk about AI are misconstrued by the general public — basically that AI went big too soon. they acknowledge in just about every paper that AI is quite valuable

hrefna, 4 months ago to random

Me: <minding my own business, winding down for bed>

Me: <exhausted from lack of sleep>

My brain, unprompted:

"Okay, but what if we actually model RDF graphs in datalog? Can we do that? What does it look like? If we do that, what kind of properties fall out?"

mnl, 4 months ago to LLMs

i've spent the whole day procrastihacking on things, and I"m just amazed by how comfortable the whole "how the heck do I do this", "oh no I have to do $tediousStuff", "ohh... I wish I could learn more about X but I don't have the time" has become.

Like I legit feel I kind of wasted my day, but...

#llms

mnl, 4 months ago to random

it's wild to me how many people in my professional peer group, reaching the ripe ages of 40-50, start to complain about the youth, about how human intelligence and creativity is going to shreds, how kids don't know anything.

Things I try to do to avoid that:

• remember what it was like to be 20 and indeed not know anything
• hang around teens and 20 years and remind myself how wildly creative and quick learners they are

...

mnl, 4 months ago to random

This hits pretty hard... https://www.aporiamagazine.com/p/a-tale-of-two-teenagers

I was the only technical person in my family (but got plenty of exposure to the life of the mind, if not direct stimulation, because I was into computers and not art or literature), but I failed school so bad (or rather, school failed me? I'm still getting used to phrasing it that way).

I made it to a good place now at 40, but there's a lingering feeling of grief about 10 year old Manuel getting into the education grinder.

1/

daviwil, 4 months ago to random

It turns out that you can extend OBS with Lua scripts.

Let's see what kind of magic can be made when you throw Fennel in the mix...

kellogh, 4 months ago to python

update on #fossil: last night i pushed changes to make algorithms pluggable. it uses #python entry points for discovery, so you can develop your own plugin without cloning fossil. it gives you a few hooks:

• train: e.g. if you need to train an ML model, or just calculate aggregate statistics

• render: format toots into an object that’s passed to a jinja template. you can make the jinja template or just use one that already exists

• HTML forms: e.g. to capture hyper-parameters