What is Hackers' Pub?

Hackers' Pub is a place for software engineers to share their knowledge and experience with each other. It's also an ActivityPub-enabled social network, so you can follow your favorite hackers in the fediverse and get their latest posts in your feed.

I'm not terribly familiar with Yann LeCun and his work, but his opinion as expressed in this article is one I agree with - llm are a local maximum, and there is no path where this leads to a "genai" or "superintelligence" consequence, regardless how sophisticated you make your models nor however much processing power you throw at them.

futurism.com/artificial-intell

Thinking is not language-based.

Language is an API by which people approximate the concepts in their heads to each other.

The semantic connections between words and concepts are not fixed; they're fairly sloppy, with a high degree of tolerance in how they fit together.

This is a feature; this is how poetry works, for instance.

This is also why, in areas such as law and medicine, the practitioners have fossilized specific semantic connotations and relationships using extremely specific jargon that you don't find outside of those fields, and frequently use Latin - a language not subject to the same forces of semantic drift as English, due to the paucity of normal speakers of it - to ossify those concepts and keep them consistent.

Starting -from- language and working backwards to the underlying conceptual framework is the opposite of how humans learn in the first place; infants learn basic facts about the world during their early life, and then they are taught the external cues that allow for communicating facts about the world with their caretakers through consistent conditioning - same way you teach a dog to sit; you associate the condition with the word 'sit' and thus achieve instruction.

While llms are certainly a clever way to create the impression of "understanding" it is, ultimately, a trick - the only 'understanding' comes from the human side; Clever Hans is not doing math at all, but engaging in a fuzzing of human responses to get the sugarcube.

0
0
0
0

Hot take: llm "guardrails" are worthless and will always be ineffective; they are a throwback to a premodern model of security as a list of prohibitions against actions instead of a more modern, holistic approach where the system as a whole is structured such that impermissible operations fail as a consequence of the system architecture.

The core mechanism of llm systems relies on the random elision and remixing of inputs; all such guardrail systems exist within this milieu, and are thus - architecturally, according to how llms work as a baseline - subject to that same elision; therefore, you can never be assured that a given guardrail directive will be present in the context window for the llm at the time of processing.

I personally think this is blindingly obvious, but I do understand why people who are bought into the tech might not understand that any attempt to 'instruct' an llm as to 'alignment' is going to be subject to an erosion of those 'protections' as an inherent part of the function of the machine.

Bluntly, if you don't want the llm to "do" a thing, you must make that thing impossible for the llm to do. Do not give it access to your filesystem; do not give it access to your production infrastructure; do not give it access to your children; do not give it access to anything unsupervised whatsoever.

And do not use an llm for any system where determinacy of operation is even slightly important, for that matter.

theregister.com/2025/11/14/ai_

0
0
0
0
0
0
0
0
0

Meta has banned us from their monetization program for a comic where our Grim Reaper reassures a kid they won't die until they're very old, and the kid – processing this information with the flawless logic of someone whose prefrontal cortex is yet to develop – leaps off and shouts "Awesome! I'm gonna try heroin!"

@warandpeasWar and Peas 🧿
Step 1: drive the real content away.
Step 2: "train" genAI* on the real content.
Step 3: ai supplies good enough content to keep eyeballs.
Step 4: hey wait, where is the training data for our genAI
Step 5: oh god, what have we done?
Step 6: whew, safe in our Indonesian bunker.
Step 7: the shock collars have been circumvented. We are overrun! All is lost!

* copyright infringement engines

0

Meta has banned us from their monetization program for a comic where our Grim Reaper reassures a kid they won't die until they're very old, and the kid – processing this information with the flawless logic of someone whose prefrontal cortex is yet to develop – leaps off and shouts "Awesome! I'm gonna try heroin!"

The comic contains the word "heroin," which is a drug. We're aware. The joke is obviously not promoting drugs, but Meta's enforcement bots – which seem perfectly content to let right-wing hate speech flourish like mold in a dorm fridge – have decided otherwise. So we recommend that you subscribe to our newsletter before the bot overlords complete their inevitable takeover and we lose each other forever in the algorithmic void.

warandpeas.com/subscribe

0
0
0

What I'm listening to today: "Scandinavia", Billy Woods, ELUCID and The Alchemist

Woods and ELUCID just dropped an album under the name Armand Hammer and it's INCREDIBLE. Woods and clique have this idiom of slowed-down, format-breaking rap music and it gets better and more refined with each release. Parts of "Mercy" sound like early Portishead, other parts sound like nothing I've ever heard before. There's like 4 tracks on here I wanna link but here's some ominous piano

armandhammer.bandcamp.com/trac

What I'm listening to today: "07", ebl (video by Sean Redbeard)

This guy's got a YouTube channel where he picks experimental electronic tracks (good selections too) and makes glitchy TouchDesigner animations for them.

This track's cool, a quiet but relentless onslaught of pattering percussion. Somebody left the kick drum faucet on overnight and now the floor's flooded in kick drums. Old Richard Devine/PBO kinda feel maybe? Manages to feel like an oddity in an odd genre

youtube.com/watch?v=eol20bvTiAw

0

So I grew up next to and this is, well, TERRIFYING.

A story for y’all: I’m from a city called Zhytomyr, 2 hours west of Kyiv in the North of . We were downwind of the Chernobyl power plant when the 1986 disaster happened.

I wasn’t born for another 12 years, but my childhood was filled with stories and the aftermath of it all. Things like:

- My grandmother worked as a head doctor in a hospital and rehabilitation facility exclusively for children of Chernobyl victims to treat the extremely high prevalence of Tuberculosis and other severe health complications. (To specify: these were SECOND GENERATION of exposure).

- A lot of the kids in that facility were orphans, because their parents died young from health problems.

- My uncle’s wife was born in Pripyat. She was 1 year old when the disaster happened. Her parents were told to evacuate while given no information about what happened. They had to pack up their things and rush out to an unfamiliar city with their baby, never to see the rest of their belongings, apartment, or hometown again.

- When I was a kid, it became so common to see weirdly mutated animals and insects that even 2-3 year olds would make jokes about “Chernobyl mosquitos” and I wouldn’t even flinch seeing occasional giant bugs, dark frogs, weird-looking dogs.

- We’d frequently hear of nearby farms having issues with their animals being born too mutated to survive or random outbreaks from contaminated water / food. Crops would randomly fail. People would get poisoned on a regular basis. This all got less common as I grew up.

- My mother still remembers being a little girl, 10 years old, and looking outside from their balcony at the clouds blowing over from Chernobyl that day. People were told to not go outside and to shut all the windows, but not given an explanation as to why. My mother swears that the rain looked different. They weren’t able to go and buy more food for the kitchen for multiple days.

Anyway - nuclear safety isn’t a joke. I don’t understand how this level of carelessness can happen after Chernobyl and Fukushima.

404media.co/power-companies-ar

0
0
0
0
0
0
0
1
0
1

A little PSA: if your library of choice is facing funding cuts, don't hold off on using their services because you're worried it'll put pressure on their existing funds.

Take advantage of everything and help them get some lovely stats to help them demonstrate impact as they fight back! If it looks like they're not being useful to folks, they'll get cut!

Don't do the cost cutters' jobs for them!

0
0
0
0
0
0
0
0
0
0

Ich baue gerade Slides für einen Vortrag nächste Woche und frage mich, wie viele Memes ich einbauen kann, ohne dass es zu unprofessionell wirkt.

Was meint ihr? 🤔

0
0

Suppose I wanted to throw a silent rave party some day, and said event would be open to anyone to attend provided they bring their own headphones and receiver.

What would be the best form of legal broadcast such that 1) the event could happen in a small park, 2) the required receiver is relatively easy to obtain, 3) the broadcast is heard relatively synchronously, and 4) doesn't require the attendees have a radio license, and 5) doesn't require me to have an expensive license?

0
0
0

There was a problem in one of my MS-R1 benchmarks last week. One that took many hours of troubleshooting to figure out (Arm SVE vs NEON support in a linear algebra library).

I take you on a little bit of that adventure on Level 2 Jeff today: youtube.com/watch?v=OOZHUqjRlpg

0
0
0
0
0
0
0
0
0
0
1
2
0
0
0
0
0
0
0
0
0
0
0
0

Walmart now uses AI-based surveillance cameras in all its stores and is rolling out body cameras to associates to collect even more video of customers that are fed into its storage. Kroger, Fred Meyer, and QFC are also now using AI-based video surveillance and also store facial data of every customer that enters their stores. Going grocery shopping in America is becoming a privacy nightmare that will only get worse.

0
2
0
0
0
0
0
0
0

@stefanoStefano Marinelli This is indeed a great post. Refreshing to read - the author didn't distro-bash but instead highlighted how needs were met.

I have been leaning toward after 20 years on , which has a special place in my heart. However, FreeBSD's *consistency*, elegant design, and better documentation are driving factors and I just can't look away.

Over these two decades, Linux has definitely changed. It's no longer what it once was: a clone. It's evolved into its own thing. Software does that and that's OK. But a lot of the changes or "improvements" have been needlessly reinventing the wheel with a worsening user experience and convoluted results. The audio subsystems (ALSA -> PulseAudio -> Pipewire and of course the mess that is/was JACK) and are two big examples.

On SystemD, I don't disagree that Linux needed a modern init system. SystemD is faster, but from a human perspective it's worse. I am now typing *more* characters to manage services. Is there a reason why it couldn't be designed to manage services like:

> $SERVICE start/stop/restart

Instead, we are left with:

> systemctl $SERVICE start/stop/restart

But I digress.

I fell in love with FreeBSD when I first started my Linux journey in the mid-2000s, but only dabbled in it as it was never in any production environment I managed.

I have a FreeBSD VM sitting in my lab somewhere but due to other factors rarely do I get to it. I'm thinking it's time I grab a spare laptop, load it up, and use it as a daily driver to really force myself to learn it.

0

& video accounts to follow:

➡️ @heartlandurbanistMatt Caffrey - Urbanist & activist campaigning for livable cities, better transit, especially in Midwestern USA
➡️ @ohtheurbanityOh The Urbanity! - City planning in Canada & rest of world
➡️ @drtcombsTab Combs - Cycling & city planning in USA, esp. road safety
➡️ @ianrbuck - Urban planning & land use in Minnesota USA
➡️ @paigePaige Saunders - Canadian urbanist & activist, also politics, economics, cycling etc
➡️ @reeceReece Martin - In-depth videos on transit in cities

🧵 1/2

0
0
0