HN user
Once a model is built, the biggest cost is inference
Something i cant find any reliable data for, but would help for a sense of scale: How much use before its equal to training?
I.e. assuming you have the training data and setup, and we only care for compute - How hours of using eg Kimi K3 / Fable, for it to equal the compute required to train it?
People infringe on Anthropics IP
No.
Authors do not infringe on IP when they read another's book, nor should the lumber company be able to dictate how I use planks and if I can resell them if i'm done with them.
You're framing it as if the added value of the author or lumber company, awards them consideration when somebody uses the products to create more value.
IP law was always a big mess, and these questions cross far into ideology instead of law; but I do not understand people who think we need an ideology where more IP-law is good for society.
I'm not sure your complaint works in this case.
You only write a report if you want the exemption.
The hypothetical business you're imaging shouldn't be looking for an exemption.
The law's effect on small business is conceptually not too different from eg laws against dumping toxic sludge.
I came to the formalisms of software engineering through painful experience rather than academic instruction. If anything, that made me take those principles more seriously once I understood them.
Not related to the article, but I've seen this thought before and I think its wrong.
This isn't what good academic instructions gets you. Instead, they provide a systematic approach to learning foundational/core formalisms which let you recognize other problems as being of the same kind.
An academic background should let the person reason from a place of pre-explored essential complexity, instead of first having to rediscover & deconstruct the accidental complexity.
Building scar tissue about why things are a certain way is practical experience for (non)academics alike.
I think you misunderstood my previous comment.
Game theorists, like any academic, prefer models with the least number of assumptions and (transitive) dependencies.
My issue is with thinking this presumed fixed point was created with an elegant model (and thus conclude it's relatively interesting to study).
If you deconstruct the model - i.e. look through its full dependency tree to state the assumptions - It's not simple or elegant.
Other assumptions like modeling without irrational behavior are purer in that sense.
This is too dismissive of the impact.
Datacenter builders now have to add security so it doesn't happen a second time, perhaps even add it in more places around the world, and the overall attractiveness of building a datacenter in the region go down.
The CO2 not emitted by opening a later easily offsets curing by orders of magnitude.
To fully model it you'd have to account for the demand being moved as other centers will pick up the load and try to model either the reduced output and reduced future-demand at the temporarily higher cost.
That's too much effort for me, but "concrete curing causes more CO2" is jumping to a conclusion.
The elegance of the conjecture seems overstated, and looks like dressed-up ideology.
The conjecture requires a theory of time in the form of periods/rounds to reduce price to zero - but wants to ignore time when it comes to information spread and product value.
Pinning a variable to any 1 extreme can be very informative, but when the variable has to both exists to justify the presumption, but also be ignored in the model - you're not going to find an elegant basis for a grand-theory-structure to add corrective terms onto.
Instead, you'll just get a mess of variables from over-fitted data.
Glad I got off Anthropic when they decided they'd be better off building a walled-garden for subscription users "to improve UX".
So first off, phenomenal stuff to see a 1bit model at 90% capability.
However, this is the 5th product post in 2 weeks that proclaims that AI use is shifting, and why [insert tradeoffs] are the perfect fit.
Paradigms shift don't happen in the release announcements.
I suspect this is an AI-ism, making all the release posts sound so paradigmshiftery.
I bought a bag of chips without having to show my ID.
There is a big difference between: Government demands every website to have age verification, and government supported scheme by which service can opt into age verification.
As of now, American private spyware is actively filling the demand.
I get the feeling some privacy advocates are approaching the choice as a tier system, with government being the worst case.
I don't see it.
The only viable solution for the future of privacy is to not be dependent on the giant platforms in the first place.
In my mind, it's pretty simple: I'm a human, LLMs are not. If a human writes a novel, it's inherently worth more because it's hard-earned and anchored to experiences we share.
.... What?
You open up with the calculator argument. Great. Then state "on the flip side". Ok.
And then declare it very simple in your mind when there is _only_ the flip side.
The whole issue is the not-simple gray area and where each of us believes the line between empowered human and brainless idiot is drawn.
I'm a heavy Rust user who doesn't like Zig all that much.
I browsed through the Bun code following Kelly's post, and decided to have Codex replace all my Bun usage with Deno.
One of his policies is to build 1 affordable house - which is brilliant.
I think you wildly misunderstand the situation.
If what you think is true, why did Farage step down call the election in the first place?
It seems rather clear it was a PR stunt he knows he'll win in this district, which lets him distract from his own corruption scandals.
But if you think there is some other reason I'd be interesting to hear the theory.
This is kind of a strange comment as it implies a false dichotomy.
Its not 'nefarious' in that its in their best business interests.
But it'd be difficult to take anyone serious who thinks Anthropic's motivation was to improve the UX, and the other effect were by accident. At the time they specifically started blocking based on openclaw prompt text. Its a walled-garden tactic.
A walled garden is nefarious to people who do not want to be inside one.
I bailed on Anthropic the moment they started blocking alternative harnesses like pi on their subscription plans.
If two people agreed on that statement, its entirely unclear if they agree with each other and if they found something profound in the first place.
The imo much more foundational relationship not everybody is aware of is https://en.wikipedia.org/wiki/Curry%E2%80%93Howard_correspon...
They blocked Claude from being used in a different harness as well squeezed the usage like crazy. Switched to Codex and haven't cared since.
Between the two the biggest difference by far is ... getting your harness / AGENTS.md / skills / tools set up right.
What shows up in your news feed and what the politicians are spending time on are wildly different things.
There are other (voices in other) parties that want to expel/restrict foreigners. There is a spectrum, and the AfD holds a rather large portion of the people tentatively agreeing with what you call misrepresentation.
How many people of their party must make such statements while being welcome in the party, for it to not be misinformation/misrepresentation.
Is that true per se?
I was under the impression certain dedicated single-algorithm quantum computers might be much easier to build; allowing you to attack some construct but not yet do full Shor.
PS I'm not saying that's whats happening. Just trying to nail down the scope of what is possible (not plausible).
That is rather disturbing but this had me lol:
Spain is “making a big mistake,” said Bart Groothuis [...] “Spain is now dependent on the country with the largest and most sophisticated offensive espionage program directed against us.”
I highly doubt he's naive enough to believe the "against us" qualifier exempts the operator of the largest and most sophisticated offensive espionage program ever.
The two opening arguments are rather weak.
- European group could not be infiltrated by a state-actor with 100billion/y budget and a history of doing so?
- NOBUS today would not be secret in the algorithm but a quantum algorithm/device. Just a month ago HN was getting flooded with "PQC is probably required by 2030".
Whats the revenue model for this NBPaaS? (No Bugs Please As A Service)
I suspect the mindset that makes you write these nonconstructive questions trying to be right instead of making sure we're on the same page, is the same one that is preventing you from entertaining the idea that people are able to collaborate, understand, and write out a set of rules they can agree on are in their own interests.
I did not say a system that benefits everyone.
I meant the difference between "everyone" as in Bob living on the street everyone, and "everyone" that is using it.
I thought maybe that was the thing that you were not understanding, so i keep having to go to your position and guess at what you mean. But sure take it as a moment to gloat with your superior intellect.
If you think 5 comments in and bringing up "But what if China ignores it" is scrutiny then I'm sorry to tell you you're just oblivious, not critical. You keep bringing up taxes and tolls. I've already addressed them 2.5 times in this thread.
is typically achieved
well at least i know you couldn't be bothered to comprehend any of my comments before wasting my time.
Sorry but i'm good on trying to explain things again - this isn't going to work. Clearly i made the mistake of taking to much for granted and we set off on the wrong foot. Its not you its me. I'm sure you understand.
Maybe try again in a week.
Pay what? Its a 'pure' system where nothing has to be denominated in money but instead we care about effects:
taking-up-orbits, applying-force, and danger-pollution.
The system requires the biggest player calculates the cost of non-compliance being too high either by:
- the scheme triggering a clause where other members start to apply force to their non-sanctioned orbits and trashing their satellites, add in the additional threat of earthly economic sanctions like we do with so many companies.
- the scheme breaking apart and becoming a wild west with _extreme_ danger-pollution.
Clearly, at the moment, the biggest users do not care because the danger is low enough, and they'll try to delay a bit so they get more 'grandfathered' in.
I'm not saying its going to be perfect or fair. I'm arguing its dumb to fear a mythical global government, because that's not required to create stable solutions for these kinds of issues.
You're not really good at asking anything. I did not say a system that benefits everyone.
I did not say the dutch solved this exact problem, and my previous response was a best effort after you prompting me because i assume you've just never thought through or dealt with non-mega-government solutions before.
I brought up the comparison because nba456_ 19 was already complaining about a required mythical global government - as if that's relevant for the 50y time frame we're planning for.
I also dont understand why you keep bringing up tax and tolls. I'll try to answer that again in the other comment.
You wish to take up a specific space, and with the right to use it come obligations.
There was no tax or toll to consider.
People naturally self-assembled under the idea that if you benefited, then you had to contribute. The calculation was: how much land did you have, how valuable was it, how much benefit you'd get from some waterworks (it wasnt just dikes but also rerouting rivers etc). Obligations were denominated in labor.
My point is not that this was some perfect idyllic corruption-free scheme - it wasn't - but it was very transparent.
All you should need for a stable system is for the majority of interests to align. One addition you can add is to have the labor be turned against individuals/groups that decide to unalign - i.e. instead of trashing space debris your next labor is to de-orbit a satellite if 1 company decides to try their luck at tyranny over cooperation.
Things hinge on a shared understanding that its always possible to go back to the salted earth solution.
Anything you want to censor has at least 5 stages:
Creator, first share (direct), second order sharing (public-ish website), third order sharing (indexed resharing), and finally the consumer wanting it presented.
In the way there are things we clearly want to censor for being awful, there are things we must never allow to be censored. Eg knowledge of a genocide.
But the solution kind of rights itself. To censor something you need as many actors as possible in that enormous graph of sharing nodes to clearly want to censor that thing we all agree we clearly want to censor. I.e. a public library doesn't censorship child porn because they are required to.
Isnt taking down illegal content censorship?
So yes this is censorship, but 'illegal' content is too vague.
We want to know about the censorship beyond the natural baseline.
Censorship, usually, means the extraordinary request for powers to control the web of communications - in the context of what and why.