eadwu

I think it is pretty much agreed that the OP has a perfectly good claim starting a legal event. Whether it amounts to the 1B settlement or not is not something anyone here can decide.

Legally, the grant contract is the "Casus belli" - it is the reason for which you have a claim to something, whether it was intent or not doesn't matter, you have a valid claim to take them to court, there is no need to really talk about that. However it gets resolved, is really not something I can tell, not a lawyer.

Whether I remain friends or on trustworthy terms is a personal matter, which is unrelated to legal outcomes in this case.

I'm just differentiating between the legal correctness and personal opinion/viewpoints on the matter.

Regardless of whether you are legally correct or not isn't really the crux of the issue for me, heck maybe Jensen doesn't even care about a billion given everything's he's minting (who knows though).

I was only talking from personal standing, in your stated cases or if I was in this position (I made some assumptions regarding the case), it would be a breaches of good faith and I would find it hard to trust them in a partnership ever again.

Only my POV, maybe the folks at the B level don't care about this as much since they are starting to fling Ts around. I can see how some people wouldn't really care, i.e. treating them as a freelancer and main objective is getting something done, but on a personal basis probably wouldn't trust them (say on personal matters that are company defining).

Of course, contracts are legally binding, but that's completely different from my perception / evaluation of someone - if the mutual understanding was over 4 years and then 30 years later you find the typo and come back to bite me then yes it is 100% an irreparable relationship.

Yes I would think better of them if they came back to bite if NVIDIA was a penny stock but probably nobody would bite them back if they found out it was worth $100 instead of $1B.

To be clear, the grant is the legally binding document regardless of intent (initial offer + vesting seems to imply 4 years), I would eat the typo if it was 30 years ago and chalk it off as a learned mistake (assuming it wasn't company altering), although I would still think less of them for fleecing.

It seems like both you didn't caring about it during vesting or maybe they did a typo since it seems like you were vested over 4 years normally and the offer is over 4 years?

Anyway if someone did this to me this is pretty much grounds for an irreparable relationship. Though I guess it was worth the attempt for $1b.

What it does do is that since it provides a rigid structure, it compels law abiding citizens to do those actions, not forcing still.

You would have a hard time proving liability if you didn't have separated finances.

LLMs still are better than Jev at the task, just across the board slower.

Anyone who had a reason to try this already tried it (ads/recommendations) - back in 2023/2024 during the first fine tuning wave and it was accurately determined that it was not worth the effort, the results were more bogus than just using CoT, so frankly parallelism meant nothing if bogus * parallel = bogus.

So thrown into the dumpster and nobody really cared to revisit because it was already tried.

Pretty much sometime between then and now it somehow became the state where the tradeoff makes sense now.

NixOS is not suited for general workstations, to even get a semblance of rationality for general use you would need to develop the underlying infrastructure which you are not going to get from a government unless they are diehard sponsoring it (spoiler alert, in the current climate, pretty much as nil as it can get).

Pretty stupid choice of action - seems to be a more nationalistic play than a rationale one (the creator and the Foundation are based in the Netherlands which is what they are referring to).

You're comparing a local GPU to network hops? Wouldn't be surprised if Jev was actually similar in runtime and their is just a great deal of network latency.

The evaluation is quite interesting though - I'd actually say the raw answer is correct in the absence of detail and prior knowledge (Who wrote the play Romeo and Juliet).

This is normal (specific to pyproject.toml overwriting authors). I am not sure about pyproject.toml conventions but I would have done the same. This is the equivalent of telling you who the maintainers of the project are.

Usually what you see in the code is the retention of Copyright statements on source code and LICENSE with some reference on README.

You can't not reinvent something if you don't know it exists in the first place.

The thing about software is that it slots nicely into every other field, making it a _very_ good base to work off of to get concepts (aka the modern bootleg polymath) that have likely been invented in other fields (with different names).

    cap_add:
      - SYS_ADMIN
    security_opt:
      - seccomp:unconfined
      - apparmor:unconfined

Stop coping and deluding yourself mate.

To begin with, whether AI is the one doing the discovering or not makes no difference. Any "pure math" person would aim to understand regardless - and would be quite glad that they have a longer paved path.

Any mathematician in academic or industry is more than likely not a "pure math" person (tainted by capitalism).

You complain so much to the point that nationalism is becoming prevalent but when someone tries to force you to be independent you complain every way possible to not be independent? But then when you are dependent you complain about not being equivalent.

It doesn't work both ways.

However, this isn't a plug to be using AI for coding everything, but a more general plug that AI should be integrated to a lot more things outside of the mainstay of chatbots.

There is a lot of merit to using AI to establish a new abstraction layer.

Consider the combination of senile + mirages.

To me this is a major problem of everyone saying security through obscurity is bad. But then those same people reinforcing encryption as a gospel of security.

As far as I know, there are no secrets in the world. Encryption is not providing security to anything. It only gives you guarantees wrt to a certain interpretation/perspective.

Modern encryption is underpinned by, no common folk (not no one or even the people who would have the ability to which are probably the ones that should be worried about) should be able to decrypt your contents _within your lifetime_ - which in and of itself is a pragmatic goal, but does not ensure secrets remain secrets.

If I open source something, I am not your parent, if you decide to wing it just because it is open source that is your problem not mine.

You are not obligated to use anything anyone open sources. If you can't take responsibility for your own actions that's on you, not the person who open sourced it - that is what generally all the licenses state to begin with.

The difference between "craft lovers" and "doers" is that one operates at a better fitting abstraction (that is more aligned to the values of capitalism).

You can say "doers" are just "craft lovers" in and of itself - there is little distinction between them - this is just reiterating the change from binary to high level languages.

I've tried explaining the implementation word and word and it still prefers to create a whole new implementation reimplementing some parts instead of just doing what I tell it to. The only time it works is if I actually give it the code but at that point there's no reason to use it.

There's nothing wrong with this approach if it actually had guarantees, but current models are an extremely bad fit for it.

The reason why LLMs are still restricted to higher level programming languages is because there are no guarantees of correctness - any guarantee needs to be provided by a human - and it is already difficult for humans to review other human's code.

If there comes a time where LLMs can generate code - whether some term slop or not - that has a guarantee of correctness - it is indeed probably a correct move to probably have a more token-efficient language, or at least a different abstraction compared to the programming abstractions of humans.

Personally, I think in the coming years there will be a subset of programming that LLMs can probably perform while providing a guarantee of correctness - likely using other tools, such as Lean.

I believe this capability can be stated as - LLMs should be able to obfuscate any program code - which is pretty decent guarantee.

For LLMs to have replaced Tailwind - in part by using it themselves - this does not have mean that there _will_ be another library to reuse. In the context of LLMs, it becomes so "cheap" to customize the webpage that a library is no longer needed.

Tailwind in and of itself can be considered a "highly structured LLM" - if they so took it that far.

Bose's brand is built on audio quality. There is close to little negative impact open sourcing the API (server) in this case will bring to their brand.

For a game, open sourcing the server generally means anyone can basically mess it up and with the internet make it available to everyone to see. Then the responsibility is on the developer to protect their "brand".

The plethora of WoW private servers is not a good example. These are from individuals, or groups of people who willfully reverse engineered it on their own. This is different from a company expressly permissing and implicitly giving a grant on allowing a similar product to exist - the difference is that one gives credibility, which the other does not.

If you wish to hide something, why have you leaked it in the first place?

Why not ask the other question? Why are you trying to hide public information to begin with? Why are you introducing encryption on top of an underlying public interface.

This is intentionally different from are there things one would generally not be to be widely accessible or generally public.

There is nothing to hide if it is already public, because it is already public, you can't hide it even if you want to, you're only making it more difficult for a general member of the public to access that data. Even if you consider that "hiding", the source is still public.

To be fair, this wouldn't be an issue if Open Source stuck with "Debian Free Software". If you really want to call it a bait and switch, open source did it first.

I haven't exactly bisected the issue but I'm pretty sure convolutions are broken on sm_121 after a certain size, getting 20x memory blowup from a convolution from a 2x batch size increase _only_ on the DGX Spark.

I haven't had any problems with inference, but I also don't use the transformers library that much.

llama.cpp was working for openai-oss last time I checked and on release, not sure if something broke along the way.

I don't exactly know if memory fragmentation is something fixable on the driver side - this might just be the problem with kernel's policy and GPL, it prevents them from automatically interfering with the memory subsystem to the granularity they'd like - see zfs and their page table antics - or so my thoughts on it is.

If you've done stuff on WSL, you have similar issues and you can fix it by running a service that normally compacts and clean memory, I have it run every hour. Note that this does impact at the very least CPU performance and memory allocation speeds, but I have not have any issue with long training runs with it (24hr+, assuming that is the issue, I have never tried without it and put that service in place since getting it due to my experience on WSL).

If SOCAMM2 is used it will still probably be at most near the range of 512/768 GB/s bandwidth, unless LPDDR6X / LPDDR7X or SOCAMM2 is that much better, SOCAMM on the DGX Station is just 384 GB/s w/ LPDDR5X.

Form factor will be neutered for the near future, but will probably retain the highest compute for the form factor.

The only way there will be a difference is if Intel or AMD pump their foot on the gas, which this makes maybe 2/3 years of it, with another 2 years unless they have something cooking it isn't going to happen.