The alignment problem is already the wrong narrative, as it implies agency where there is none. All that talk about “alignment problem” draws focus from AI ethics (not a term I made up).

Read the article.

Then you highlight why AI Safety is important by linking a blog post about the dangers of poorly thought-out AI systems

Have you read the article? it clearly states the difference of AI safety vs AI ethics and argues why the formerare quacks and the latter is ignored.

If you read AI Safety trolley problems and think they are warning you about an ai god, you misunderstood the purpose of the discussion.

Have you encountered what Sam Altman or Elsevier Yudkowsky claim about AI safety? It’s literally “AI might make humanity go extinct” shit.

The fear mongering Sam Altman is doing is a sales tactic. That’s the hypeman part.

I know of him and I enjoy his videos.

Still, fearmongering for a god AI that might kill us all is both unscientific and distracts from the real harm so-called AI is already causing.

This post is especially ironic, since AI and its “safety researchers” make climate change worse by ridiculously increasing energy demands.

So-called “AI safety” researchers are nothing but hypemen for AI companies.

https://en.wikipedia.org/wiki/Swiss_cheese_model

Sorry, homie. I’m not gonna keep arguing with you if you obviously can’t argue without moving the goal posts, if your life depends on it.

My point still stands: Encrypting metadata can be sensible/necessary for your threat model and does not count as security through obscurity. You have failed to explain how it would be and then started to attack me, personally.

Have fun misrepresenting this comment as well, bye.

Firstly, if the police confiscate your PC, they already know (and have proven to a judge) that you conduct illegal activity and likely already have enough to convict you of a crime. lol

Not if it’s for securing evidence. That is only collected before the verdict/conviction. Otherwise, there wouldn’t be any need for a trial.

Also, your metadata can put others in jeopardy. If you’re busted for being an antifascist activist, who the police deems a “terrorist” and you’re also member of another activist group which up to then wasn’t in the sights of law enforcement, then you’re putting that other activist group’s members in danger.

Secondly, you can have an account at a private torrent tracker […]

That wasn’t my argument, though. You can criticise the circumstances that started my example premise, but the point still stands: having metadata that’s clearly visible can be dangerous, because it can give an attacker more information on you (depending on your threat model).

These are exceptionally poor arguments.

You’ve actually only attacked my examples, not my argument. My original point still stands: The type of accounts you have can be something you legitimately want/need to encrypt. Not only the credentials.

Because if the data is secure, it makes no difference if a bad actor knows you have an account with a service or not

Bullshit. It’s not about the obvious services, but rather the ones that give more info about my profile.

If the police confiscates my PC because of e.g. piracy, they could nail me down if they also knew that I had an account at a darkweb marketplace, or that I am a member of an organization that’s deemed to be “terrorist”.

The only way to hide that info with pass is to give it a cryptic name which make it less obvious, what the account is actually for. That is both inconvenient and I would argue: also quite security of obscurity.

This is an example of security through obscurity.

It is not. Security through obscurity relies on having a visible secret hidden somewhere where “no one would think to check”. That’s different than encrypting the whole meta-structure of your digital life.

That’s a non-sequitur.

How is encrypting the metadata, as well as the data security through obscurity? O.o

OP is talking about hhe meta-structure being visible.

If my filesystem gets compromised (stolen, confiscated, etc.) and I use pass, the infiltrators will know that I have a password that I labeled “slrpnk.net”. They won’t have access to the password itself, but they’ll be able to determine all the services I have accounts at.

Got it. I’ll stop “using” the US government. /j

No parallel universes employed => Access denied

deleted by creator

Here’s an actual AI evangelist/researcher making a point how open source can’t apply to neural networks (or LLMs)

Edit: Also, I’m not talking about the assets. I’m tal*ing about the ability to recreate. If you have the Doom assets, you can build it yourself.

Are you serious?

You can read all the other comments which explained why it is not open source. You can’t really retrain the model without petabytes of data. Even if you “train” stuff on your dataset: it’s more like tweaking the model weights a bit, rather than building the model from scratch.

“Open source” is PR talk by Meta and deepseek.

Just regurgitating propaganda.

No u.

Communism is defined by worker control of the means of production

No, that’s socialism. Communism is a stateless, classless, moneyless society.

Also, the workers in China aren’t in control of the means of production. The bureaucrats and capitalists are.

That’s how they lifted 800 million people out of poverty

While I applaud the better living standards: I see no contradiction to capitalism. Marx himself wrote:

The bourgeoisie, during its rule of scarce one hundred years, has created more massive and more colossal productive forces than have all preceding generations together. Subjection of Nature’s forces to man, machinery, application of chemistry to industry and agriculture, steam-navigation, railways, electric telegraphs, clearing of whole continents for cultivation, canalisation of rivers, whole populations conjured out of the ground — what earlier century had even a presentiment that such productive forces slumbered in the lap of social labour?

Simply participating in markets does not make a country capitalist

No, but state ownership of the means of production makes a country state capitalist.

Let’s come back to your initial point: you accused me of dismissing an LLM, because it is “communist”. Comrade, I’m an anarcho-communist. I dislike all the hypetrain riding, water gobbling sharlatans in the current so-called “AI” bubble: Altman, Pashaj (don’t know how to write him), Zuckerberg and Musk. I like it if the market get disrupted, but I don’t like it if the AI trend continues to wreak havoc on nature.

Btw: how is deepseek “energy efficient”?

And also: All that has squat to do with open source or not.

I was specifically addressing the use of the phrase “open source”. And the term “open data” doesn’t apply either, since it’s not a dataset that’s distributed, but rather weights of an LLM with data baked into it. That’s neither open source nor open data.

I’m not into gambling and know that the house always wins.

The differenge is that the dataset is baked into the weights of the model. Your emulation analogy simply doesn’t have a leg to stand on. I don’t think you know how neural networks work.

The standards are literally the basis of open source.