xg15

Completely ignoring the extreme inequalities in ownership and just basic control about the general direction that economic development goes to, so I can't really take this seriously.

There is zero objective necessity for the current data center gold rush, and yet it takes away an enormous amount of resources.

Meanwhile, China somehow manages to turn deserts back into fertile land.

I think it's interesting how even inside tech, so many people seem to view neural network weights as "black magic" utterly beyond human comprehension. I guess that contributes to the sense of mystery around LLMs...

(And also the sharp difference in levels of sophistication you see inside neural nets and outside of it: Pipelines that feed text to LLMs so it can detect highly abstract and nuanced semantic differences in the text's contents - but the pipeline still uses regexes for sentence splitting...)

> Altman talks about how they’re working on having robots build “self-replicating data centers” (what?!?)

Company that cautions against the paperclip maximizer already at work building the paperclip-maximizing hardware themselves.

Though I appreciate how this would invalidate the last remaining arguments for the "data centers bring local employment" claim, if the security and construction jobs are automated away as well.

Maybe it will be replaced with a new spin: "Data center? What data center? Oh that one... that's just a wild weed, you can't prove it's one of ours..."

I think there as a difference whether the discovery was done by some greek guy 2500 years ago or by a person who is still living and interested in knowing the consequences of his discovery.

Next up: Wageless employees.

> if you give a human the weights of qwen 0.8B and 5 years it'll still be impossible to edit the weights in any legible manner.

Is that so? There is quite a lot of research going on to identify the "meaning" behind certain directions. I don't see why a human with enough time, perseverance and ability to run experiments with the model shouldn't be able to uncover some of the underlying mechanisms of the model.

Only of course not with the speed with which an LLM can do it.

And here we go...

As others have said, while still absolutely insane, I don't find it completely out-of-the-blue surprising. One of the recurring abilities of (current-generation) LLMs seem to be to rapidly grasp very complex systems, where humans would need months or years to reach the same level of expertise.

You can see this in the way a coding agent can "understand" and explain sprawling legacy codebases in seconds.

You can also see it in the projects that are using them to reverse-engineer proprietary IoT protocols or computer games.

But those were all systems where we had detailed knowledge about what structures are there to be reverse-enginered.

If the models can transfer this ability to some other systems where we don't know the underlying structure then I think we will be in for a lot more craziness in the future. (Protein interaction networks? DNA sequences? Animal languages?)

As for weights, if the models really can analyze the semantics behind specific weights or directions, I have my hopes that they could also give a textual explanation or help designing tools to make "handcrafting" weights more practical in general.

I know the "technical" definition of a token, i.e. an individual feature vector of the input and output sequence of an LLM; usually a word or part of a word.

I have no idea if that's the same definition of "token" that OpenAI's billing department uses.

I think the court viewed this less like a something someone would say to a person (because obviously they'd just respond with "lol no") and more like a hacking attempt that happened to use English words as the exploit vector.

So I think a better comparison would be if they had embedded some byte sequence in the document that was intended to cause a buffer overflow and RCE in the court's reader software.

The article cited the law the court was using, and it's essentially "attorneys must not do anything to sabotage the court proceedings":

> Article 77 of the Code prohibits parties and counsel from obstructing judicial processes, including a duty not to practice illegal innovations in the factual or legal state of the proceeding and not to act in a way that makes judicial activity difficult.

In that sense, I think the fact that it targeted the court's software is more important than that the exploit used human language.

Yeah, I feel this often confuses what "efficiency" means from the company's perspective vs wider society's perspective.

A company that streamlines its operations and archives a huge profit margin is clearly efficient from the PoV of its own management and shareholders. However from the PoV of its customers, it's maximally inefficient, because it tries to extract the highest price for the least amount of work done that people are still willing to pay.

I agree, but would like the same thought applied to the private sector. I read a lot of comments who are extremely critical of every bit of power a state assumes, but have no problem if a private company reaches a practical monopoly or gets access to every intimate detail of peoples' lives.

It's right if you only focus on the resource allocation aspect. (Which would be a bit hypocritical as other comments here are mentioning).

But the elephant in the room is the question of who would control all that RAM and set or change the rules about which usages are allowed, for what price and which aren't allowed.

This against the backdrop of Anthropic just now deciding to block using Claude for reverse-engineering video games.

Or to go with your meme, the person you hate didn't just make a great point, they would also get to control what you do with your computer. Would you be fine with that?

> This is what will happen for all pets registered with us immediately after the Rapture:

- Our non-Christian administrators will activate our rescue plan.

- Volunteers will be alerted immediately by email and telephone that they have been activated.

[...]

I wasn't aware people are making practical plans and building software systems to be activated after the rapture.

Why are wire ends like this not using a loop? Would the loop be too huge?

> The research was tricky because we didn’t want to reach out to Namco for help too soon, instead planning to wow them with a close-to-finished doodle.

...and not at all because they'd see this as a copyright violation or make demands, no sir.

> The Michael Bay of doodles

...still no software engineer though.

I like how the cutesy haha-but-actually-no style we know too well of tech companies today was already bleeding through back then...

Not to be confused with training a model on Demoscenes, the little-known cousin 3rd degree of Greek statesman Demosthenes.

There is some irony there. So asking the AI to mod or reverse-engineer an existing game is a no-go, but asking it to create a "new" game using stolen assets and stolen code is fine...

This is true, but I think building an entire technology on top of the question of who can consume the resource as inefficiently as possible is still something else.

Weirdly, underutilization of hardware was never a problem before. In fact, wastefulness is so ingrained in programming culture that the previous hype - blockchain - was about an explicit race over who could waste the most computing power on nonsensical calculations.

Also, I'd be very surprised if the current rush to build datacenters everywhere is optimal usage of resources. How many companies are building datacenter commons and sharing with other companies to optimize resource usage?

> Two statements can be true

As I general point, because I keep seeing this line on HN: Of course they can in the general case, but more often than not, the reason why this line is written is because there is some tension or contradiction between the statements. In that case, understanding where the tension lies is IMO more useful than shrugging and deciding to "hold two contradictory thoughts in your head" like in 1984.

Sometimes it feels like a rhetoric device to "rescue" an opinion when the opposite opinion is hard to argue against.

> Explorer sets the current directory of the program to the directory containing the thing that you double-clicked

I found the linked article to informative and confirming my suspicion: "The thing that you clicked" doesn't have to be the executable - it could also be a document with the executable registered as launch handler, in which case the current directory will be the directory of the document.

Would be interesting to know how links and items in shell namespaces are handled.

> for(size_t i = count; i--; ) { /* find needle */ }

As someone who only writes C or C++ occasionally, I'd consider this a good submission for an obfuscated C contest.

It really does look simple at first glance.

Then you start to wonder what the stop condition is.

Then you realize the i-- is the stop condition.

Then you wonder how the whole thing works at all.

(The description explained the trick well, and the construct makes sense once you understand it, but I still think this is one example of something that is "too simple" and becomes cryptic)

After all the stuff you hear about undefined behavior, I'm also honestly surprised that wraparound on unsigned types is not undefined behavior.

Saw a TV panel on this a few weeks ago.

You know the "pro-datacenter" person was getting desperate when he tried to sell the construction work of building the datacenter as the "job growth".

On the moderator's remark that such jobs would be temporary, he got back with "oh no, datacenter construction can take several years, so it's not temporary!"

No further questions...

Still remembering the Kellerautomat from my CS class...

EDV - maybe a less glitzy term for "tech" - is still used occasionally by Gen-Xers.

> The President is the chief executive in charge of government operations, not the courts. If he’s acting outside his authority, you can sue him and then the courts can figure it out.

I love this. Yes, and then several years later, the courts can declare his order was in fact illegal when it's already far, far too late to reverse any of it.

That kind of "oversight" is no oversight at all. If you believe this is just how things should be, then you want a king with unrestricted power, not a president.

> There’s also this thing called the International System of Units, a scientific measurement deal that goes by its French acronym: SI.

That would probably one more reason for Trump and his supporters to choose that acronym...

I think in this case, it was rather cynical. The "free time" the workers got was because they were unemployed. The workers who were still employed of course had no additional leisure time, the added productivity went to the employer.

At this point I wonder if those aren't simply marketing shills or the usual kind of "pro-business" press articles that always push ridiculously optimistic takes on new technologies.

Reminds me of the old press pieces along "with the new machines, the workers won't have to do this tedious and dangerous work anymore and will have so much more free time!"

And free time they had...

Is there meaningful difference between handing back a wiped phone and handing back a new phone of the same model? So in that case, why not keep the un-wiped phone in possession of the court and allow attorneys access to it (with proper documentation) - and give the convict a new phone instead?

> avoid a giant prefix miss

I was wondering how feasible it is to modify the context, instead of just appending - e.g. to remove malformed tool calls and their error messages that would pollute the context otherwise.

Sure, you'd have to rebuild the KV cache and take a performance loss, but that might be acceptable if it happens rarely enough.

> especially with the newly proposed Dell bank account for kids that can only be used to invest in stocks

The ol' "The real evil overlord is just you and me" retort is a bit undermined by the existence of products like this one that actively constrain people's actions and pressure them to perpetuate the system. Like, who gave the order to design that product?

It's also not true. "Companies" on their own don't do anything. The operational decisions are made by CEOs and board members who do have a lot of leeway to decide how they want - and often do have an ideology they follow.

Shareholders are also not "you and me" in the common case - especially not with index funds, where you have zero agency to influence the market even despite being a shareholder on paper.

Significant amounts of shares are held by wealthy individuals or other companies with humans at the top - people who are in the same social group (or often the same individuals) as those in the C-suite.

At the level of wealth concentration we are in, I think talking about the "billionaire class" is more constructive than talking about "shareholders" and then noticing full of surprise that we all are shareholders.