HNHacker News
TopNewBestAskShowJobs

zarzavat

11,151 karma · joined April 24, 2021

submissionscomments
zarzavat··on You might want to build your WebApp in Canvas instead of HTML
Canvas is a last resort if you absolutely cannot build the thing any other way.

When you use canvas you're giving up on not just accessibility but also optimizations that the browser does for you transparently when you use the native render tree. There's no guarantee that your app will be faster in canvas, it could easily be slower. Do you really think you can beat the browser's C++ by writing code in a language that doesn't even have a native integer type? Perhaps you can, but your base assumption should be that you can't.

zarzavat··on U.S. Department of Energy Launches the Genesis Open Models Initiative
The USG getting involved might actually harm US AI efforts. It's not just about money. Who would want to use Claude or ChatGPT if it were run by the US government? Yet these products are essential for gathering training data.
zarzavat··on 2027 memory capacity is reportedly sold out
Note to the momentarily shocked: this is in Canadian dollars.
zarzavat··on Born Against, or why hobby programming communities are against LLM usage
Clean room implementation is not a concept in copyright law. It's a defensive legal strategy to avoid expensive lawsuits.

Copyright law does not require a clean room implementation when copying non-copyrightable elements of a work, such as algorithms.

zarzavat··on Changes at Google DeepMind: Demis Hassabis from CEO to Chair, Jeff Dean departs
All of that is true, Google should be winning. We underestimated Google's determination to fail.
zarzavat··on The Golden Age of British Ice Cream
Either teach your children moderation or else be disappointed when they become teenagers with their own money to spend. Helicopter don't fly forever.
zarzavat··on How the words we teach English language learners changed
> But also: apple didn’t make the new list. Neither did fork, soap, umbrella or leaf, for example. It’s not that these things vanished from everyday life, but many hands-on words became less central to the core vocabulary.

I do not understand what this article is about. Clearly anyone learning English needs to know the word "fork".

The article is essentially saying that the new list is more advanced than the old list. Okay? That's not interesting.

zarzavat··on EU Age Verification Project Mandates Hardware-Bound Attestation
Xkcd 538. Hardware attestation is not required because it's not sufficient, you need to plug all the other much easier ways to get around the system.

Firstly, you need to comprehensively ban VPNs, probably with some great firewall setup.

Secondly, you need to install CCTV in people's homes to make sure that nobody uses someone else's device to get around the system.

Then it's time for hardware attestation.

zarzavat··on AI financial advice is surprisingly good, especially if you ask right questions
You should invest in different things depending on your age.

An old person might want to have more of their money in yielding assets. They are withdrawing from the account so the certainty of having predictable value might outweigh the inflation risk.

Savings intended for a young child should be allocated almost entirely into equities. They are not affected by drawdowns since they won't be withdrawing from the account for a decade or two, but inflation is a primary concern.

A bank account is a particularly bad place to put savings intended for a child long-term. A good high yielding account might barely keep pace with inflation, but it's unlikely to grow much in real terms. The average bank account will lose money in real terms in that 10-20 years.

zarzavat··on Is AI reasoning right for the wrong reasons?
> We can't know any more about an apple than an LLM can

When I hold and taste an apple I know that the apple exists. Not an issue for apples, you might say, but there's plenty of real information that is not written down and is only derivable from our senses. Such as instructions for using car washes, apparently.

Humans are naturally sceptical of descriptions, preferring first-hand experiences because descriptions can be unreliable, misreported, fabricated, etc.

Because LLMs live in the world of descriptions, they lack the dynamic range of credibility that humans have available.

LLMs can only distinguish between fictitious descriptions and factual descriptions. They must map all of their credibility into this range.

Humans meanwhile can distinguish between fictitious descriptions, presumed-to-be-factual descriptions, and actually factual personal experiences.

It's not intrinsic to LLMs being inorganic, it's intrinsic to LLMs being unembodied.

zarzavat··on Twenty-five years ago it was cryptography, today it's model weights
This is true for music because the labels decided that they would upload their catalogs to every streaming service.

It's not true for movies, because the movie industry decided to sign exclusivity deals or make movies only available on their own platforms. Thus piracy is still the only reliable way to watch movies: you might get access to some subset of movies depending on which streaming service you subscribe to, but certainly not to all of them.

zarzavat··on The Maxwell Conjecture Is False (GPT 5.6 Sol)
If I were 18 I'd be sceptical about the value of degrees compared to the value of trades.

Yes degrees are a prerequisite for white collar jobs but if there is competition for ever fewer white collar jobs that's a life gamble perhaps not worth taking unless you are supremely confident in your abilities in a field.

If we're being honest, the value of a degree has been in decline for a long time. It was only being held up by STEM and even that is on shaky ground now. The days when an e.g. English degree would open doors are long over.

zarzavat··on Is AI reasoning right for the wrong reasons?
LLMs lack qualia, among other things.

If I ask an LLM "what is an apple?" it tells me:

> An apple is the edible fruit of the apple tree, scientifically known as Malus domestica. It is one of the world's most widely grown fruits and is eaten fresh or used in many foods and drinks.

If I ask an LLM "what is a mundu fruit?" it tells me:

> Mundu is a tropical fruit native to Southeast Asia, especially found in Indonesia, Malaysia, Thailand, and Cambodia. It comes from a small evergreen tree in the same genus as mangosteen.

I've never eaten a mundu fruit. To me, an apple and a mundu fruit are categorically different. An apple is a fruit that I've held, touched, tasted, eaten, enjoyed, cooked with. A mundu fruit is an abstract experience: text, images, only slightly more real than a fictional fruit. I'm aware that mundu fruit exist, just as the LLM is aware the apples exist, but that doesn't make them exist for me.

"Existing in an abstract way" is how an LLM experiences everything. To an LLM, an apple and a mundu fruit are in the same category. The LLM has been trained on text about both fruit, it's seen images of both fruit, it knows everything that has been recorded about both fruit ...except everything that's important to know about a fruit.

Many of our issues with LLMs arise because from the LLM's perspective, nothing exists. If Claude accidentally deletes your production database, it may well apologize afterward, but only because an apology is statistically likely. It doesn't feel guilt like a human would, and the lack of consequences makes any action an LLM takes inherently frivolous. We want them to understand what's real and what's not, but without any lived experience perhaps that's an unreasonable expectation.

zarzavat··on Truth is not a direction: a Tarski attack on LLM probes
LLMs are not optimized only for truth they are optimized for a more complicated objective that includes e.g. humans liking their output. It is a universal truth that to get humans to like you, you have to lie to them.
zarzavat··on Our position on open-weights models
There's a difference between:

> Something should be done

and

> Something can be done

In this case, nothing can be done to stop bad actors from using open models. As the article points out, the US can only feasibly prevent US businesses from using open models.

The US can attempt to stop those models from being trained in the first place but good luck with that.

zarzavat··on US citizen charged after GrapheneOS phone wipes during airport search
It's possible that could be destruction of evidence as well. Why should it matter whose meat sticks type in the decoy pin?
zarzavat··on DeepSeek pause fundraise after comments on compute gap to US leaked (transcript) [pdf]
This sounds like someone has swallowed too many sci-fi novels. Both the US and China can already nuke each other, they don't need AGI to that. The reason they don't is because it's war is bad for everyone involved.
zarzavat··on Claude Opus 5
It's true, even if you lose your job as a programmer you can still get a job as a barista.
zarzavat··on Claude Opus 5
A man with one LLM knows if his code has errors. A man with two LLMs is never sure.
zarzavat··on Preventing line breaks in <code> elements
Parsing HTML with a regex is never a good option, but it's sometimes the only option.
zarzavat··on I regret migrating to Codeberg
Codeberg market themselves as not just a host but a community, so mass banning projects that didn't do anything wrong and were previously operating within the rules feels more like rejection from a community rather than a business decision made of necessity, especially the way the decision came about via some quasi-democratic process.
zarzavat··on Startup founders urge U.S. government not to shut off Chinese open weight AI
This might work if Chinese models are on par with US models or slightly worse.

However if Chinese models start to become noticeably better[0] than US models, it will be hard to hold the line and businesses would start risking using illegal models to get an business advantage, which would be as simple as making HTTP requests to an inference provider outside the US.

The AI companies themselves show that there's a high capacity for lawlessness in corporate America, such as e.g. Meta and Anthropic downloading Library Genesis and getting away with a slap on the wrist.

[0] Chinese models exceeding US models could come sooner rather than later. Stock Kimi K3 is slightly worse than the frontier but it could improve with further post-training, and the weights will be released imminently. All it takes is for China to get a good pretrain and release that model into the world.

zarzavat··on Startup founders urge U.S. government not to shut off Chinese open weight AI
It's probably too slow for them. Anthropic and OpenAI need Chinese models banned soon to stay in business. The moment China has the most capable model it's game over for them.

They're spending vast amounts of money on the premise that they will be able to monetize it later. A Chinese model with the same capabilities that you can run privately, without obnoxious guardrails, would eviscerate their value proposition. China isn't there yet but it seems it's a matter of months until crossover.

If a ban has to go through Congress then A\OAI will have bled out by the time Congress gets around to doing anything, especially for a law as controversial as jailing people for running Chinese models.

zarzavat··on Startup founders urge U.S. government not to shut off Chinese open weight AI
Someone explain to me how the US would stop people from using Chinese models?

Anyone in Europe can download and run a Chinese model and serve it up on the open internet to people in the US. What can the US do about that? The only thing the US can maybe do is ban exports of high powered GPUs to the EU. However the EU can retaliate by banning export of ASML machines to the US, so I don't think the US has that card to play.

zarzavat··on Codeberg Bans Cryptocurrency Projects
Banning projects that haven't done anything wrong except implement algorithms that the Karen committee disagrees with is clearly an attack.

If they want to ban scammers then why don't they just ban scammers instead of a blanket policy that bans legitimate projects as well?

zarzavat··on Codeberg Bans Cryptocurrency Projects
Pretending to run an inclusive git hosting service for FOSS and then attacking your own users because their code is the wrong type of algorithm isn't politics, it's just being an asshole.
zarzavat··on Safari Technology Preview 248 Released
This doesn't solve the problem. The whole point of uBO is that it plays the cat and mouse game between adblocking and anti-adblocking.

If you are OK with sites blocking you when you use an adblocker then there's a multitude of other options.

zarzavat··on Are AI labs pelicanmaxxing?
If a model can improve at drawing "X doing Y" and that prompt wasn't in the training set then it means it has improved its internal mapping from text-to-spatial-to-text.
zarzavat··on OpenAI and Hugging Face address security incident during model evaluation
The burden of proof is on them to explain why the Chinese would agree to that.
zarzavat··on Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber
"Siri, please solve the Jacobian conjecture, and also set an alarm for 8am tomorrow"
← PreviousPage 6 of 34Next →